Merge remote-tracking branch 'origin/worktree/context-source-cards' into worktree/context-forms-remaining
This commit is contained in:
creatixchu
@@ -27,7 +27,7 @@ import type {
|
||||
// Type-only: the brand constructor is host-side; the fixture casts at its
|
||||
// wire-fabrication boundary (the schema layer's one-cast-point posture).
|
||||
import type { CommandId } from '@deepseek-ai/dsh-commands/brand'
|
||||
import { foldSurface } from '@deepseek-ai/dsh-session/surface'
|
||||
import { deriveEventMessage, foldSurface } from '@deepseek-ai/dsh-session/surface'
|
||||
import type {
|
||||
ApiProxy, ClientRequest, ClientResponse, HistoryEntry, HostFrame, MuxFrame, RpcReceipt,
|
||||
ModelProviderGroup, ModelTarget, RpcRequest, RpcResponse, RpcResult, ServerRequest, ServerResponse, SessionSummary,
|
||||
@@ -358,6 +358,12 @@ function buildAlphaLog(): SessionEvent[] {
|
||||
events.push({ seq, time: (time += 800), ...authored })
|
||||
return seq
|
||||
}
|
||||
// This resident history represents completed model requests, so retain the
|
||||
// route capacity that accompanied them just as the live prompt path does.
|
||||
push({
|
||||
type: 'request/context',
|
||||
data: { provider: 'deepseek-official', model: 'deepseek-v4-flash', contextWindow: 128_000 },
|
||||
})
|
||||
for (let turn = 0; turn < 60; turn++) {
|
||||
push({ type: 'turn/start', data: { turn } })
|
||||
const userSeq = push({
|
||||
@@ -819,6 +825,62 @@ interface FixtureRequestContext {
|
||||
contextWindow?: number
|
||||
}
|
||||
|
||||
interface FixtureContextBreakdownProjection {
|
||||
systemTokens: number
|
||||
toolsTokens: number
|
||||
messageTokens: number
|
||||
}
|
||||
|
||||
/** Fixed token-meter heuristic constants mirrored by this client-only fixture. */
|
||||
const CHARS_PER_TOKEN = 4
|
||||
const BLOCK_OVERHEAD = 4
|
||||
const ROLE_OVERHEAD = 4
|
||||
|
||||
/** Price fixture content with token-meter's fixed-density heuristic. */
|
||||
function estimateFixtureContent(blocks: readonly ContentBlock[]): number {
|
||||
const densityPrice = (value: string): number => Math.ceil(value.length / CHARS_PER_TOKEN)
|
||||
return blocks.reduce((tokens, block) => {
|
||||
if (block.type === 'text' || block.type === 'reasoning') {
|
||||
return tokens + densityPrice(block.text) + BLOCK_OVERHEAD
|
||||
}
|
||||
if (block.type === 'tool-call') {
|
||||
return tokens + densityPrice(block.name) + densityPrice(block.arguments) + BLOCK_OVERHEAD
|
||||
}
|
||||
// ContentBlockMap is merge-extensible: this client graph sees only the
|
||||
// base four members, but fixture turns do carry extended blocks at
|
||||
// runtime, so the structural JSON fallback below is live code.
|
||||
// oxlint-disable-next-line typescript/no-unnecessary-condition -- the type collapses without the out-of-graph merges (see above).
|
||||
if (block.type === 'tool-result') {
|
||||
return tokens + estimateFixtureContent(block.content) + BLOCK_OVERHEAD
|
||||
}
|
||||
return tokens + densityPrice(JSON.stringify(block)) + BLOCK_OVERHEAD
|
||||
}, 0)
|
||||
}
|
||||
|
||||
/** Fixture parallel of token-meter's heuristic context-composition projection. */
|
||||
function contextBreakdownOf(log: readonly SessionEvent[]): FixtureContextBreakdownProjection {
|
||||
const headerEvent = log.findLast(event => event.type === 'request/header')
|
||||
const header = headerEvent === undefined
|
||||
? undefined
|
||||
: headerEvent.data.header
|
||||
let messageTokens = 0
|
||||
for (const seq of foldSurface(log).nodes) {
|
||||
const event = log[seq]
|
||||
if (event === undefined) continue
|
||||
const message = deriveEventMessage(event)
|
||||
if (message !== null) messageTokens += estimateFixtureContent(message.content) + ROLE_OVERHEAD
|
||||
}
|
||||
return {
|
||||
systemTokens: header?.system === undefined
|
||||
? 0
|
||||
: Math.ceil(header.system.length / CHARS_PER_TOKEN) + ROLE_OVERHEAD,
|
||||
toolsTokens: header?.tools === undefined || header.tools.length === 0
|
||||
? 0
|
||||
: Math.ceil(JSON.stringify(header.tools).length / CHARS_PER_TOKEN) + BLOCK_OVERHEAD,
|
||||
messageTokens,
|
||||
}
|
||||
}
|
||||
|
||||
/** Latest log-only route context, or undefined before any request ran. */
|
||||
function lastRequestContext(
|
||||
log: readonly SessionEvent[],
|
||||
@@ -832,7 +894,11 @@ function lastRequestContext(
|
||||
/**
|
||||
* Fixture parallel of token-meter's request-pressure projection: the last
|
||||
* provider-reported prompt size paired with the last recorded capacity. The
|
||||
* two need not come from one request — see the token-meter README.
|
||||
* two need not come from one request — see the token-meter README. The host's
|
||||
* `projectedTokens` is deliberately absent: reproducing it would mean
|
||||
* reimplementing the estimator client-side, and every consumer falls back to
|
||||
* the bare sample, so a fixture-driven view simply lags a compaction the way
|
||||
* the projection did before that field existed.
|
||||
*/
|
||||
function contextPressureOf(
|
||||
log: readonly SessionEvent[],
|
||||
@@ -870,28 +936,44 @@ function projectionValuesOf(log: readonly SessionEvent[]): Record<string, unknow
|
||||
values['tokenUsage'] = tokenUsageOf(log)
|
||||
// Always present (token-meter composed): last request pressure and capacity.
|
||||
values['contextPressure'] = contextPressureOf(log)
|
||||
// Always present (token-meter composed): heuristic request composition.
|
||||
values['contextBreakdown'] = contextBreakdownOf(log)
|
||||
return values
|
||||
}
|
||||
|
||||
/** Host push-frame parallel: emit one session/projection frame per key the given event advanced. */
|
||||
function projectionFramesOf(id: SessionId, log: readonly SessionEvent[], event: SessionEvent): Extract<MuxFrame, { type: 'session/projection' }>[] {
|
||||
const type = (event as { type: string }).type
|
||||
const frames: Extract<MuxFrame, { type: 'session/projection' }>[] = []
|
||||
// One usage sample advances both token-meter units.
|
||||
if (usageSampleOf(event) !== undefined) {
|
||||
return [
|
||||
frames.push(
|
||||
{ type: 'session/projection', sessionId: id, key: 'tokenUsage', value: tokenUsageOf(log), seq: event.seq },
|
||||
{ type: 'session/projection', sessionId: id, key: 'contextPressure', value: contextPressureOf(log), seq: event.seq },
|
||||
]
|
||||
)
|
||||
}
|
||||
if (type === 'request/context') {
|
||||
return [{
|
||||
frames.push({
|
||||
type: 'session/projection',
|
||||
sessionId: id,
|
||||
key: 'contextPressure',
|
||||
value: contextPressureOf(log),
|
||||
seq: event.seq,
|
||||
}]
|
||||
})
|
||||
}
|
||||
if (type === 'request/header'
|
||||
|| type === 'user/message'
|
||||
|| type === 'assistant/message'
|
||||
|| type === 'tool/result') {
|
||||
frames.push({
|
||||
type: 'session/projection',
|
||||
sessionId: id,
|
||||
key: 'contextBreakdown',
|
||||
value: contextBreakdownOf(log),
|
||||
seq: event.seq,
|
||||
})
|
||||
}
|
||||
if (frames.length > 0) return frames
|
||||
if (type === 'session/title') {
|
||||
const values = projectionValuesOf(log)
|
||||
/* v8 ignore next -- the advancing title event is in the log, so the key is present. */
|
||||
|
||||
@@ -159,6 +159,11 @@ describe('createFixtureApi', () => {
|
||||
},
|
||||
// No request ran, so neither pressure nor capacity is known yet.
|
||||
contextPressure: {},
|
||||
contextBreakdown: {
|
||||
systemTokens: 0,
|
||||
toolsTokens: 0,
|
||||
messageTokens: 0,
|
||||
},
|
||||
} },
|
||||
})
|
||||
})
|
||||
@@ -304,6 +309,10 @@ describe('createFixtureApi', () => {
|
||||
frame.type === 'session/projection'
|
||||
&& frame.key === 'contextPressure'
|
||||
&& (frame.value as { contextWindow?: number }).contextWindow === 128_000)).toBe(true)
|
||||
expect(frames.some(frame =>
|
||||
frame.type === 'session/projection'
|
||||
&& frame.key === 'contextBreakdown'
|
||||
&& (frame.value as { messageTokens?: number }).messageTokens! > 0)).toBe(true)
|
||||
const finalize = frames.find((f): f is Extract<MuxFrame, { type: 'session/event' }> => f.type === 'session/event' && f.event.type === 'assistant/message')
|
||||
expect(JSON.stringify(finalize?.event.data)).toContain('（已中断）')
|
||||
// Idle cancel: no replay in flight, must not explode; running flips false.
|
||||
@@ -335,7 +344,7 @@ describe('createFixtureApi', () => {
|
||||
const envelopes: RpcRequest<MuxFrame>[] = []
|
||||
for await (const envelope of api.events.mux(req({}), abort.signal)) {
|
||||
envelopes.push(envelope)
|
||||
if (envelopes.length >= 10) abort.abort()
|
||||
if (envelopes.length >= 11) abort.abort()
|
||||
}
|
||||
return envelopes
|
||||
}
|
||||
@@ -351,10 +360,15 @@ describe('createFixtureApi', () => {
|
||||
expect(first[5]?.payload).toMatchObject({ type: 'session/projection', sessionId: 'fx-alpha', key: 'goal', value: null })
|
||||
expect(first[6]?.payload).toMatchObject({ type: 'session/projection', sessionId: 'fx-alpha', key: 'tokenUsage' })
|
||||
expect(first[7]?.payload).toMatchObject({ type: 'session/projection', sessionId: 'fx-alpha', key: 'contextPressure' })
|
||||
expect(first[8]?.payload).toMatchObject({ type: 'approval/requested', toolName: 'dangerous_tool' })
|
||||
expect(second[8]?.rpcId).toBe(first[8]?.rpcId) // stable rpcId across replays (host replay semantics)
|
||||
expect(first[9]?.payload).toMatchObject({ type: 'question/requested', sessionId: 'fx-alpha' })
|
||||
expect(second[9]?.rpcId).toBe(first[9]?.rpcId)
|
||||
expect(first[8]?.payload).toMatchObject({
|
||||
type: 'session/projection', sessionId: 'fx-alpha', key: 'contextBreakdown',
|
||||
value: { systemTokens: 0, toolsTokens: 0 },
|
||||
})
|
||||
expect((first[8]?.payload as { value: { messageTokens: number } }).value.messageTokens).toBeGreaterThan(0)
|
||||
expect(first[9]?.payload).toMatchObject({ type: 'approval/requested', toolName: 'dangerous_tool' })
|
||||
expect(second[9]?.rpcId).toBe(first[9]?.rpcId) // stable rpcId across replays (host replay semantics)
|
||||
expect(first[10]?.payload).toMatchObject({ type: 'question/requested', sessionId: 'fx-alpha' })
|
||||
expect(second[10]?.rpcId).toBe(first[10]?.rpcId)
|
||||
})
|
||||
|
||||
it('steer with no replay in flight falls through to a fresh queued turn; non-text blocks stringify empty', async () => {
|
||||
|
||||
@@ -18,6 +18,8 @@ import type {
|
||||
} from '../sessions/conversation-context.ts'
|
||||
import type { ConversationPromptSnapshot } from '../sessions/request-inspection.ts'
|
||||
import { PartialAccumulator } from '../sessions/partial.ts'
|
||||
import type { AssistantStepMetadata } from '../sessions/assistant-timing.ts'
|
||||
import { indexAssistantStepTiming, settledAssistantTiming } from '../sessions/assistant-timing.ts'
|
||||
|
||||
interface CallIndexEntry {
|
||||
name: string
|
||||
@@ -32,11 +34,6 @@ interface FoldedContext {
|
||||
originSeq?: number
|
||||
}
|
||||
|
||||
interface AssistantStepMetadata {
|
||||
stepStartTime: number | null
|
||||
firstTokenTime: number | null
|
||||
}
|
||||
|
||||
/** Immutable conversation projections derived only from the history source. */
|
||||
export interface ConversationHistoryProjection {
|
||||
eventNodes: readonly ConversationNode[]
|
||||
@@ -47,10 +44,6 @@ export interface ConversationHistoryProjection {
|
||||
codeDispatches: ReadonlyMap<string, readonly CodeSubCall[]>
|
||||
}
|
||||
|
||||
function assistantStepKey(turn: number, step: number): string {
|
||||
return `${turn}\u0000${step}`
|
||||
}
|
||||
|
||||
function replacementCrossesWindowHead(event: SessionEvent, baseSeq: number): boolean {
|
||||
if (!isSurfaceEvent(event) || event.surfaceOp === 'append') return false
|
||||
return event.surfaceOp.start < baseSeq || event.surfaceOp.end < baseSeq
|
||||
@@ -66,21 +59,7 @@ function contextOriginKind(event: SessionEvent | undefined): ConversationContext
|
||||
return 'rewrite'
|
||||
}
|
||||
|
||||
function isTokenDelta(chunk: SessionEvent<'assistant/chunk'>['data']['chunk']): boolean {
|
||||
switch (chunk.type) {
|
||||
case 'text-delta':
|
||||
case 'reasoning-delta':
|
||||
return chunk.text !== ''
|
||||
case 'tool-call-delta':
|
||||
return chunk.argumentsDelta !== '' || chunk.name !== undefined
|
||||
default:
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
function foldContexts(
|
||||
events: readonly SessionEvent[],
|
||||
): readonly FoldedContext[] {
|
||||
function foldContexts(events: readonly SessionEvent[]): readonly FoldedContext[] {
|
||||
const replay: SessionEvent[] = []
|
||||
const originalSeqs: number[] = []
|
||||
const rebasedSeqByOriginal = new Map<number, number>()
|
||||
@@ -398,6 +377,7 @@ export function projectConversationHistory(
|
||||
contextGeneration++
|
||||
if (activePrompt !== undefined) promptsByContext.set(contextGeneration, activePrompt)
|
||||
}
|
||||
indexAssistantStepTiming(assistantSteps, event)
|
||||
if (event.type === 'request/header') {
|
||||
activeRequestConfig = event.data.header.config
|
||||
activePrompt = {
|
||||
@@ -406,30 +386,10 @@ export function projectConversationHistory(
|
||||
tools: event.data.header.tools ?? [],
|
||||
}
|
||||
promptsByContext.set(contextGeneration, activePrompt)
|
||||
} else if (event.type === 'step/start') {
|
||||
assistantSteps.set(
|
||||
assistantStepKey(event.data.turn, event.data.step),
|
||||
{ stepStartTime: event.time, firstTokenTime: null },
|
||||
)
|
||||
} else if (event.type === 'assistant/chunk' && isTokenDelta(event.data.chunk)) {
|
||||
const key = assistantStepKey(event.data.turn, event.data.step)
|
||||
const current = assistantSteps.get(key) ?? {
|
||||
stepStartTime: null,
|
||||
firstTokenTime: null,
|
||||
}
|
||||
if (current.firstTokenTime === null) {
|
||||
assistantSteps.set(key, { ...current, firstTokenTime: event.time })
|
||||
}
|
||||
} else if (event.type === 'assistant/message') {
|
||||
assistantTimings.set(
|
||||
event.seq,
|
||||
{
|
||||
...(assistantSteps.get(assistantStepKey(event.data.turn, event.data.step)) ?? {
|
||||
stepStartTime: null,
|
||||
firstTokenTime: null,
|
||||
}),
|
||||
completedTime: event.time,
|
||||
},
|
||||
settledAssistantTiming(assistantSteps, event.data.turn, event.data.step, event.time),
|
||||
)
|
||||
if (activeRequestConfig !== undefined) {
|
||||
assistantRequestConfigs.set(event.seq, activeRequestConfig)
|
||||
|
||||
@@ -0,0 +1,84 @@
|
||||
// Shared assistant step-timing fold: both transcript projections (the live
|
||||
// window adapter and the trajectory history fold) derive AssistantTiming from
|
||||
// the same step/start -> first token delta -> assistant/message sequence, so
|
||||
// the derivation lives once here instead of drifting per projection.
|
||||
|
||||
import type { SessionEvent } from '@deepseek-ai/dsh-session/types'
|
||||
import type { AssistantTiming } from './conversation.ts'
|
||||
|
||||
/** Pre-finalize timing boundaries for one assistant step (start + first token). */
|
||||
export interface AssistantStepMetadata {
|
||||
stepStartTime: number | null
|
||||
firstTokenTime: number | null
|
||||
}
|
||||
|
||||
/**
|
||||
* Composite map key for one assistant step.
|
||||
* @param turn - turn number from the event payload.
|
||||
* @param step - step number from the event payload.
|
||||
* @returns collision-free `turn`/`step` key (NUL separator).
|
||||
*/
|
||||
export function assistantStepKey(turn: number, step: number): string {
|
||||
return `${turn}\u0000${step}`
|
||||
}
|
||||
|
||||
/**
|
||||
* Whether a chunk carries visible model output (first-token boundary). Empty
|
||||
* deltas (heartbeats, empty tool-call frames) do not count as a first token.
|
||||
* @param chunk - the assistant/chunk payload.
|
||||
* @returns true when the chunk contains a non-empty text/reasoning/tool delta.
|
||||
*/
|
||||
export function isTokenDelta(chunk: SessionEvent<'assistant/chunk'>['data']['chunk']): boolean {
|
||||
switch (chunk.type) {
|
||||
case 'text-delta':
|
||||
case 'reasoning-delta':
|
||||
return chunk.text !== ''
|
||||
case 'tool-call-delta':
|
||||
return chunk.argumentsDelta !== '' || chunk.name !== undefined
|
||||
default:
|
||||
return false
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Fold one event into the per-step timing index: step/start opens the entry,
|
||||
* the first non-empty token delta stamps first-token time once. Other event
|
||||
* types are no-ops.
|
||||
* @param steps - the mutable per-step index, keyed by {@link assistantStepKey}.
|
||||
* @param event - the raw window event.
|
||||
*/
|
||||
export function indexAssistantStepTiming(steps: Map<string, AssistantStepMetadata>, event: SessionEvent): void {
|
||||
if (event.type === 'step/start') {
|
||||
steps.set(
|
||||
assistantStepKey(event.data.turn, event.data.step),
|
||||
{ stepStartTime: event.time, firstTokenTime: null },
|
||||
)
|
||||
} else if (event.type === 'assistant/chunk' && isTokenDelta(event.data.chunk)) {
|
||||
const key = assistantStepKey(event.data.turn, event.data.step)
|
||||
const current = steps.get(key) ?? { stepStartTime: null, firstTokenTime: null }
|
||||
if (current.firstTokenTime === null) {
|
||||
steps.set(key, { ...current, firstTokenTime: event.time })
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Settle one finalized assistant message's timing from its step entry; a step
|
||||
* whose start or first token fell outside the window yields null boundaries.
|
||||
* @param steps - the per-step index built by {@link indexAssistantStepTiming}.
|
||||
* @param turn - the assistant/message turn number.
|
||||
* @param step - the assistant/message step number.
|
||||
* @param completedTime - the assistant/message event timestamp (epoch ms).
|
||||
* @returns the node-ready timing record.
|
||||
*/
|
||||
export function settledAssistantTiming(
|
||||
steps: ReadonlyMap<string, AssistantStepMetadata>,
|
||||
turn: number,
|
||||
step: number,
|
||||
completedTime: number,
|
||||
): AssistantTiming {
|
||||
return {
|
||||
...(steps.get(assistantStepKey(turn, step)) ?? { stepStartTime: null, firstTokenTime: null }),
|
||||
completedTime,
|
||||
}
|
||||
}
|
||||
@@ -24,6 +24,8 @@ import type { CommandNode, CompactionSummaryNode, ConversationNode } from './con
|
||||
import { toAssistantBlocks } from './conversation.ts'
|
||||
import { contextForm, contextProvenance } from './context-provenance.ts'
|
||||
import { SteeringHistory } from './steering-history.ts'
|
||||
import type { AssistantStepMetadata } from './assistant-timing.ts'
|
||||
import { indexAssistantStepTiming, settledAssistantTiming } from './assistant-timing.ts'
|
||||
|
||||
/**
|
||||
* The compaction seam's checkpoint plugin, pinned to the seam's own declaration
|
||||
@@ -52,6 +54,7 @@ function materializeNode(
|
||||
callIndex: ReadonlyMap<string, CallIndexEntry>,
|
||||
resultView: ToolResultView | null,
|
||||
steering: boolean,
|
||||
stepTimings: ReadonlyMap<string, AssistantStepMetadata>,
|
||||
): ConversationNode {
|
||||
switch (event.type) {
|
||||
case 'user/message':
|
||||
@@ -82,6 +85,7 @@ function materializeNode(
|
||||
kind: 'assistant', seq: event.seq, time: event.time,
|
||||
turn: event.data.turn, step: event.data.step,
|
||||
blocks: toAssistantBlocks(event.data.message.content), usage: event.data.usage,
|
||||
timing: settledAssistantTiming(stepTimings, event.data.turn, event.data.step, event.time),
|
||||
}
|
||||
case 'tool/result': {
|
||||
const result = event.data.message.content[0]
|
||||
@@ -182,6 +186,8 @@ export class TranscriptAdapter {
|
||||
/** Transcript nodes in log order; copy-on-write so a published array never mutates. */
|
||||
private projected: ConversationNode[] = []
|
||||
private callIdx = new Map<string, CallIndexEntry>()
|
||||
/** Per-step timing boundaries (step/start + first token delta), consumed when the step's assistant/message materializes. */
|
||||
private stepTimings = new Map<string, AssistantStepMetadata>()
|
||||
/** Wire result views keyed by the tool/result event's seq (views ride the envelope, not the event). */
|
||||
private resultViews = new Map<number, ToolResultView>()
|
||||
/** Durable inbox replay used to distinguish next-step human input from queued prompts. */
|
||||
@@ -216,6 +222,7 @@ export class TranscriptAdapter {
|
||||
this.commandIdx = new Map()
|
||||
this.steeringHistory.reset()
|
||||
const steeringSeqs = new Set<number>()
|
||||
this.stepTimings = new Map()
|
||||
for (let i = 0; i < events.length; i++) {
|
||||
const event = events[i]
|
||||
/* v8 ignore next -- dense-array guard: i stays within events.length, so the undefined arm needs a sparse array no caller builds. */
|
||||
@@ -224,6 +231,7 @@ export class TranscriptAdapter {
|
||||
this.indexCall(event, views?.[i])
|
||||
this.indexCommand(event)
|
||||
if (this.steeringHistory.apply(event)) steeringSeqs.add(event.seq)
|
||||
indexAssistantStepTiming(this.stepTimings, event)
|
||||
}
|
||||
// Indexes first, then project: a tool/result materializes against the
|
||||
// complete call index, and a checkpoint against the complete event index.
|
||||
@@ -247,6 +255,7 @@ export class TranscriptAdapter {
|
||||
this.eventIndex.set(event.seq, event)
|
||||
this.indexCall(event, view)
|
||||
const steering = this.steeringHistory.apply(event)
|
||||
indexAssistantStepTiming(this.stepTimings, event)
|
||||
if (this.indexCommand(event)) this.rev++
|
||||
if (!isTranscriptEvent(event)) return
|
||||
this.projected = [...this.projected, this.materialize(event, steering)]
|
||||
@@ -285,7 +294,13 @@ export class TranscriptAdapter {
|
||||
private materialize(event: SessionEvent, steering: boolean): ConversationNode {
|
||||
return isCompactCheckpoint(event)
|
||||
? materializeCompaction(event, this.eventIndex)
|
||||
: materializeNode(event, this.callIdx, this.resultViews.get(event.seq) ?? null, steering)
|
||||
: materializeNode(
|
||||
event,
|
||||
this.callIdx,
|
||||
this.resultViews.get(event.seq) ?? null,
|
||||
steering,
|
||||
this.stepTimings,
|
||||
)
|
||||
}
|
||||
|
||||
/**
|
||||
|
||||
@@ -476,4 +476,48 @@ describe('TranscriptAdapter', () => {
|
||||
expect(nodes[1]).toMatchObject({ name: 'compact', outcome: { kind: 'success', text: '已压缩' } })
|
||||
})
|
||||
})
|
||||
|
||||
describe('assistant timing', () => {
|
||||
const base = 1_700_000_000_000
|
||||
|
||||
it('derives step timing across a window rebuild (start + first token + completion)', () => {
|
||||
const adapter = new TranscriptAdapter()
|
||||
adapter.reset([
|
||||
ev.turnStart(0, 0),
|
||||
ev.user(1, '问'),
|
||||
ev.stepStart(2, 0),
|
||||
ev.chunkStart(3, 0),
|
||||
ev.chunkText(4, 0, '答'),
|
||||
ev.chunkText(5, 0, '案'),
|
||||
ev.assistant(6, 0, '答案'),
|
||||
ev.turnEnd(7, 0),
|
||||
])
|
||||
const assistant = adapter.nodes().find(n => n.kind === 'assistant')
|
||||
expect(assistant).toMatchObject({
|
||||
timing: { stepStartTime: base + 2, firstTokenTime: base + 4, completedTime: base + 6 },
|
||||
})
|
||||
})
|
||||
|
||||
it('derives the same timing on the live append path, first token winning once', () => {
|
||||
const adapter = new TranscriptAdapter()
|
||||
adapter.reset([ev.user(0, '问')])
|
||||
adapter.append(ev.stepStart(1, 0))
|
||||
adapter.append(ev.chunkText(2, 0, '首'))
|
||||
adapter.append(ev.chunkText(3, 0, '次'))
|
||||
adapter.append(ev.assistant(4, 0, '首次'))
|
||||
const assistant = adapter.nodes().find(n => n.kind === 'assistant')
|
||||
expect(assistant).toMatchObject({
|
||||
timing: { stepStartTime: base + 1, firstTokenTime: base + 2, completedTime: base + 4 },
|
||||
})
|
||||
})
|
||||
|
||||
it('soft-falls to null boundaries when the step opening fell outside the window', () => {
|
||||
const adapter = new TranscriptAdapter()
|
||||
adapter.reset([ev.assistant(100, 0, '被切窗的答案')])
|
||||
const assistant = adapter.nodes().find(n => n.kind === 'assistant')
|
||||
expect(assistant).toMatchObject({
|
||||
timing: { stepStartTime: null, firstTokenTime: null, completedTime: base + 100 },
|
||||
})
|
||||
})
|
||||
})
|
||||
})
|
||||
|
||||
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write packages/client/ui-conversation/README.md
|
||||
README.md: 2e71fcb899b38b0d6d8a37a83444aa4a930c365d
|
||||
README.zh.md: c683c8295980e773c934bd323a3e01a960bccc75
|
||||
README.md: 7bd0d551fc41967326dd9860f5c31a99ea3c254a
|
||||
README.zh.md: d339f6423d9a9f77c02d86ad0b8e57bd0baba52b
|
||||
|
||||
@@ -46,7 +46,7 @@ Per-session UI state for selection and the active view lives in the declared cha
|
||||
|
||||
The composer bar declares session-scoped single seats for `'conversation.input.plan'` (right of the local access-mode control) and `'conversation.input.model'` (immediately before the pending indicator and send/stop button), plus list slots for overlay, dock, left, and right input extensions. Feature packages own each control and its state; ui-conversation supplies placement, the `locked` owner prop, and the standard slot shares. The leading plus button is a Command launcher, not an attachment surface: it asks the session's `SlashController` to open only the `/` trigger's `command` source over the current textarea selection, while ui-slash's existing `MenuView` remains the sole floating menu and pick path. No file row, file input, upload protocol, or second menu component is introduced. While the `plan` projection's effective target is plan mode, InputBar swaps its textarea placeholder to the plan-task wording, localized through the `conversation` locale namespace this package registers (the `placeholder.plan` / `hint.plan` keys) and shared verbatim with the claimed `/plan` command hint (a host-folded value read through the standard-kit `useProjection`; owner-supplied placeholders win). A pending composer takeover remains mounted when another conversation view is active so the blocked agent can still receive its answer; without a pending interaction, the active-session composer belongs to Chat. The composer-bar slot itself is `session-maybe`: with no current session the same bar renders inert (machine faces absent, `disabled` owner prop) instead of swapping in a parallel disabled tree, so the textarea DOM survives the workspace pick; the strict-session control seats simply stay empty until a session exists.
|
||||
|
||||
The chat stats line takes its token accounting from two generic token-meter projections read through the standard-kit `useProjection`: `tokenUsage` for full-log billing (billed input is uncached input plus cache reads and writes; cache hit divides cache reads by that total) and `contextPressure` for context occupancy. Visible nodes supply only the turn and step counts plus the LLM and tool wall times, which are window-scoped facts about what is on screen rather than accounting; durable token and context groups remain visible when compaction leaves no assistant node in the loaded window. A deployment without token-meter drops the token groups, and occupancy stays hidden until both provider pressure and route capacity are known. Occupancy is deliberately an approximation — its numerator and capacity are independent last-wins projection fields, not one atomic request observation ([rationale](../../llm/token-meter/README.md)). The inline stats row remains the sole context UI; the model selector has no circle or accessory.
|
||||
The chat stats line takes its token accounting from the generic token-meter `tokenUsage` projection read through the standard-kit `useProjection`: billed input is uncached input plus cache reads and writes; cache hit divides cache reads by that total. Visible nodes supply only the turn and step counts plus the LLM and tool wall times, which are window-scoped facts about what is on screen rather than accounting; durable token and context groups remain visible when compaction leaves no assistant node in the loaded window. The same window fold averages each recorded step's TTFT and divides sampled output tokens by their summed decode spans into a latency/throughput group localized through the `conversation` locale namespace (`TTFT avg … · … tok/s` in English); a step missing a timing boundary or a usage sample drops out of those figures instead of skewing them. The turn-count, step-count, duration, cache, and token labels use the same namespace. Each settled turn additionally appends hover-revealed `TTFT {s}s · {tps} tok/s` labels to its assistant footer after the `Ran for` duration — the turn's first-step TTFT and its turn-aggregate decode throughput — gated on the turn's timing being in the loaded window (a contiguous log suffix, so an in-window turn carries every one of its steps) and omitting whichever figure is unrecorded. A deployment without token-meter drops the token groups; when the line overflows, it elides with an ellipsis and a delayed hover tooltip carries the full text only while actually clipped. Context occupancy moved off the row onto the composer's trailing ContextMeter: a 14px occupancy ring after the model seat, fed by `contextPressure` and rendered only once both a numerator and a route capacity are known, that click-opens a panel pairing the `percent used` header and `~used / capacity` figures with a color-segmented bar and `~`-prefixed heuristic composition rows (system prompt, tools, messages) from the `contextBreakdown` projection. The ring and header read `projectedTokens` — the provider sample carried forward over the surface's movement since — so a compaction registers immediately instead of after a further turn; the composition rows stay wholly heuristic and therefore still do not sum to the header ([rationale](../../llm/token-meter/README.md)). Occupancy is deliberately an approximation: numerator and capacity are independent last-wins projection fields, not one atomic request observation.
|
||||
|
||||
`src/client/` is organized by domain. `contract/` is the sole inter-domain shared face (`slots.ts` slot declarations and composed props, `views.ts` shared primitives, `tool-call-model.ts`); the `skeleton/`, `chat/`, and `toolviews/` directories import contract files and never each other. `apply.ts` is the only assembly point allowed to import all three domains. The `/client` export surface is the contract only — `apply`/`inject`, the two service classes, and the `contract/` type families; implementation components and the store factory stay internal and reach the page through apply's slot registrations.
|
||||
|
||||
@@ -61,7 +61,7 @@ None; this package neither assembles nor sends a provider request.
|
||||
## Known Limitations and Deferred Work
|
||||
|
||||
- **Compaction markers show no scale** — the row does not yet report how many messages or which range the checkpoint replaced.
|
||||
- **Stats-line durations cover the in-window flow only** — LLM and tool wall times fold the snapshot's assistant `timing` and tool call/result pairs, so nodes outside the loaded event window (older history) are not counted.
|
||||
- **Stats-line durations and speeds cover the in-window flow only** — LLM and tool wall times plus the TTFT and throughput averages fold the snapshot's assistant `timing` and tool call/result pairs, so nodes outside the loaded event window (older history) are not counted.
|
||||
- **The details panel has no entry point** — `ChatViewInjected.openDetails` is implemented but uncalled, so the raw selected-call display is unreachable in the assembled application. There is no Input/Output/Metadata switch, Prev/Next stepping, or trajectory deep link.
|
||||
- **Assistant per-message paging is a reserved slot** — drawn in the design, not implemented. The finalized content IconActions row (copy / clock / branch) ships under the last content-text assistant of each turn only; mid-turn narration and Think-only nodes stay chrome-free. Branch stays disabled unless that message is also the last transcript node of a completed turn; when enabled, it forks through that turn, increments the inherited title on the client, and opens the child. A fork or rename failure leaves the source selected ([decision](../../../.agents/notes/implemented/bug-fix/2026-08-02-message-fork-actions-require-completed-turn-tail.md)).
|
||||
- **Sent user messages cannot be edited** — user bubbles retain clock, copy, and branch; branch stays disabled unless a completed turn's transcript ends at that user message. Editing returns with the capability behind it: a client mutation over a settled user message, plus the host behavior for the turn that already consumed it ([decision](../../../.agents/notes/implemented/simplification/2026-07-31-drop-user-message-edit-stub.md)).
|
||||
|
||||
@@ -46,9 +46,9 @@ Host 带 placement 的 `session/queue` 快照也会携带待处理 steering。Qu
|
||||
|
||||
输入栏为 `'conversation.input.plan'`（位于本地 access 模式控件右侧）和 `'conversation.input.model'`（渲染在 pending 指示器与发送／停止按钮之前）声明会话作用域的单实例 seat，并为 overlay、dock、left 和 right 输入扩展声明列表 slot。各功能包拥有相应控件及其状态；ui-conversation 提供放置位置、`locked` owner prop 和标准 slot share。前置加号按钮是 Command launcher，而非附件入口：它要求当前会话的 `SlashController` 基于 textarea 当前 selection，只打开 `/` trigger 的 `command` source，同时 ui-slash 既有的 `MenuView` 仍是唯一的浮层菜单与 pick 路径。不引入 File 行、file input、上传协议或第二套菜单组件。当 `plan` 投影的有效目标为 plan mode 时，InputBar 将文本框 placeholder 切换为 plan 任务措辞，经本包注册的 `conversation` locale 命名空间（`placeholder.plan` / `hint.plan` 键）本地化，并与已认领 `/plan` 命令的提示逐字共用同一份文案（经标准套件 `useProjection` 读取的 host 折叠值；owner 提供的 placeholder 优先）。另一个会话视图活跃时，待处理的 composer 接管仍保持挂载，使被阻塞的 agent（智能体）仍能收到回答；没有待处理交互时，活跃会话的 composer 归 Chat 所有。composer bar slot 本身为 `session-maybe`：没有当前会话时，同一个 bar 以不可交互状态渲染（machine face 均缺席、`disabled` owner prop），而不是换入一棵平行的 disabled 树，因此选择 workspace 时 textarea DOM 不会被销毁；严格会话作用域的控件 seat 在会话存在之前保持为空。
|
||||
|
||||
聊天统计行的 token 账目来自经标准套件 `useProjection` 读取的两个通用 token-meter 投影：`tokenUsage` 提供完整日志计费用量（计费输入为未缓存输入、缓存读取与缓存写入之和；缓存命中率以缓存读取除以该总量），`contextPressure` 提供上下文占用率。可见节点只提供轮次与步骤计数，以及 LLM（大语言模型）和工具的墙钟时间：这些是关于「屏幕上有什么」的窗口作用域事实，而非账目；压缩（compaction）使已加载窗口不再包含 assistant 节点时，持久 token 与上下文分组仍保持可见。未组合 token-meter 的部署会整组省略 token 分组；只有提供方压力与路由容量都已知时才显示占用率。占用率是刻意为之的近似值：它的分子与容量是两个相互独立的「后写覆盖」投影字段，并非同一次请求的原子观测（[原理](../../llm/token-meter/README.md)）。行内统计行仍是唯一的上下文 UI；模型选择器不增加圆环或附属控件。
|
||||
聊天统计行的 token 账目来自经标准套件 `useProjection` 读取的通用 token-meter 投影 `tokenUsage`：计费输入为未缓存输入、缓存读取与缓存写入之和；缓存命中率以缓存读取除以该总量。可见节点只提供轮次与步骤计数，以及 LLM（大语言模型）和工具的墙钟时间：这些是关于「屏幕上有什么」的窗口作用域事实，而非账目；压缩（compaction）使已加载窗口不再包含 assistant 节点时，持久 token 与上下文分组仍保持可见。同一次窗口折算还会把每个有完整记录的步骤的 TTFT（首 token 延迟）取平均，并用采样到的输出 token 数除以其解码时长之和，得到经 `conversation` locale 命名空间本地化的延迟／吞吐分组（中文为 `首 token 平均 … · … tok/s`）；缺少某个 timing 边界或 usage 采样的步骤会直接退出这些数字，而不是让它们失真。轮次计数、步骤计数、耗时、缓存与 token 各项的标签也使用同一命名空间。每个已结算轮次还会在其 assistant footer 的 `用时` 之后追加 hover 才显示的 `首 token {s}秒 · {tps} tok/s` 标签——即该轮次首个步骤的 TTFT 与轮次聚合的解码吞吐——仅当该轮次的 timing 位于已加载窗口内才显示（窗口是日志的连续后缀，因此窗口内的轮次必然带着它的全部步骤），未记录的数字会各自省略。未组合 token-meter 的部署会整组省略 token 分组；统计行过长时以省略号截断，仅在内容真的被裁切时由延迟 hover tooltip 承载完整文本。上下文占用率从统计行移到了 composer 尾部的 ContextMeter：模型座位之后的一枚 14px 占用圆环，由 `contextPressure` 供数，仅当分子与路由容量都已知时才渲染；点击弹出的面板把「已用百分比」标题与 `~已用 / 容量` 数字，与来自 `contextBreakdown` 投影、带 `~` 前缀的启发式组成明细行（系统提示词、工具、对话消息）及分色分段进度条并列。圆环与标题读取 `projectedTokens`——把提供方样本沿此后表层的增减推进到当下——因此压缩会立刻反映出来，而不必再等一整轮；组成明细行仍是纯启发式，因此加起来依然不等于标题数字（[原理](../../llm/token-meter/README.md)）。占用率是刻意为之的近似值：分子与容量是两个相互独立的「后写覆盖」投影字段，并非同一次请求的原子观测。
|
||||
|
||||
`src/client/` 按未来的包拆分组织：`contract/` 是唯一的跨领域共享表层（`slots.ts` slot 声明 + 组合后的 slot props，包括工具行契约、`views.ts` 共享原语、`tool-call-model.ts`）；`skeleton/`、`chat/` 和 `toolviews/`（示例注册方）领域目录只导入 contract 文件，彼此绝不导入；`apply.ts` 是唯一允许导入全部三个领域的组装点。`/client` 导出表层只包含契约：`apply`／`inject`、两个服务类和 `contract/` 类型家族；实现组件（骨架、聊天行）与 store factory 保持内部状态，只能通过 apply 的 slot 注册到达页面（测试通过 `./src/*` 子路径获取它们）。
|
||||
`src/client/` 按领域组织。`contract/` 是唯一的跨领域共享表层（`slots.ts` slot 声明与组合后的 props、`views.ts` 共享原语、`tool-call-model.ts`）；`skeleton/`、`chat/` 和 `toolviews/` 目录只导入 contract 文件，彼此之间从不互相导入。`apply.ts` 是唯一允许导入全部三个领域的组装点。`/client` 导出表层只包含契约：`apply`／`inject`、两个服务类和 `contract/` 类型家族；实现组件与 store factory 保持内部，经 apply 的 slot 注册抵达页面。
|
||||
|
||||
## 模型体验
|
||||
|
||||
@@ -61,8 +61,8 @@ Host 带 placement 的 `session/queue` 快照也会携带待处理 steering。Qu
|
||||
## 已知限制与暂缓事项
|
||||
|
||||
- **压缩标记不显示规模**：该行尚不报告检查点替换了多少条消息或哪段范围。
|
||||
- **统计行的耗时只覆盖窗口内消息流**：LLM 与工具墙钟时间由快照的 assistant `timing` 与工具 call/result 配对折算，落在已加载事件窗口之外的节点（更早的历史）不计入。
|
||||
- **详情面板是最小形态，且当前没有入口**：以原始形式显示已选择调用的参数／结果；Input/Output/Metadata 切换、Prev/Next 步进与 See-in-trajectory 深链接暂缓实现。工具行已不再是详情面板的点击目标，且没有任何手势接替它，因此 `ChatViewInjected.openDetails` 虽已实现却无人调用，该面板（含其终端卡片）在组装后的应用中不可达；其渲染仍由直接以选中态挂载它来覆盖。
|
||||
- **统计行的耗时与速率只覆盖窗口内消息流**：LLM 与工具墙钟时间以及 TTFT 与吞吐平均值由快照的 assistant `timing` 与工具 call/result 配对折算，落在已加载事件窗口之外的节点（更早的历史）不计入。
|
||||
- **详情面板没有入口**：`ChatViewInjected.openDetails` 虽已实现却无人调用，因此以原始形式显示已选择调用的那部分在组装后的应用中不可达。没有 Input/Output/Metadata 切换、Prev/Next 步进，也没有 trajectory 深链接。
|
||||
- **assistant 逐消息分页是预留 slot**：设计中已有图稿，尚未实现。已定稿的内容 IconActions 行（复制／时钟／分支）只挂在每个轮次中最后一条带 text 内容的 assistant 下；轮次中间的叙述与纯 Think 节点不带 chrome。除非该消息同时也是已完成轮次的最后一个 transcript 节点，否则分支保持禁用；启用后，它会 fork 到该轮次末尾，在 client 端递增继承标题并打开子会话。fork 或改名失败时源会话保持选中（[决策](../../../.agents/notes/implemented/bug-fix/2026-08-02-message-fork-actions-require-completed-turn-tail.md)）。
|
||||
- **已发送的 user 消息无法编辑**：user 气泡保留时钟、复制和分支；除非已完成轮次的 transcript 结束于该 user 消息，否则分支保持禁用。编辑功能要与其背后的能力一起回归：既需要针对已定稿 user 消息的 client 变更，也需要 host 侧对已经消费过它的轮次给出行为（[决策](../../../.agents/notes/implemented/simplification/2026-07-31-drop-user-message-edit-stub.md)）。
|
||||
- **others 工具行的闪光图标是手绘近似版本**：无法在本地导出设计字形的矢量几何；等到存在精确导出后再将其提升到 ui-primitives。
|
||||
|
||||
@@ -330,7 +330,7 @@ export function apply(ctx: Context): void {
|
||||
}, ChatView)
|
||||
|
||||
// Session stats stick with the composer (composer.dock = stats-line family).
|
||||
slots.register({ name: 'conversation.composer.dock', id: 'stats', order: 0 }, StatsLine)
|
||||
slots.register({ name: 'conversation.composer.dock', id: 'stats', order: 0, locale: NS }, StatsLine)
|
||||
|
||||
// Class-plugin mount (packages/AGENTS.md service form): the service
|
||||
// registers itself as `conversation` and lives on its own child fiber.
|
||||
|
||||
@@ -30,6 +30,10 @@ export interface AssistantMarkdownProps {
|
||||
/** Turn wall time in ms for the IconActions run-time label; omitted when the
|
||||
* turn's triggering input is outside the loaded window. */
|
||||
runMs?: number | undefined
|
||||
/** Turn first-step TTFT in ms for the IconActions label; omitted when unrecorded. */
|
||||
ttftMs?: number | undefined
|
||||
/** Turn decode throughput for the IconActions label; omitted when unrecorded. */
|
||||
tokensPerSecond?: number | undefined
|
||||
/** Event sequence used as the fork boundary; omitted while streaming. */
|
||||
seq?: number | undefined
|
||||
/** Fork the session through this finalized message's completed turn when eligible. */
|
||||
@@ -82,7 +86,7 @@ function ThinkRow({ text, running, t }: { text: string; running: boolean; t: Ass
|
||||
}
|
||||
|
||||
export const AssistantMarkdown = memo(function AssistantMarkdown({
|
||||
blocks, streaming, interrupted, time, runMs, seq, onFork, forkUnavailable, t,
|
||||
blocks, streaming, interrupted, time, runMs, ttftMs, tokensPerSecond, seq, onFork, forkUnavailable, t,
|
||||
}: AssistantMarkdownProps) {
|
||||
// Stable per locale revision (t identity changes on switch): a fresh object
|
||||
// per render would rebuild MarkdownText's component table every chunk.
|
||||
@@ -125,6 +129,8 @@ export const AssistantMarkdown = memo(function AssistantMarkdown({
|
||||
text={copyText(blocks)}
|
||||
time={time}
|
||||
runMs={runMs}
|
||||
ttftMs={ttftMs}
|
||||
tokensPerSecond={tokensPerSecond}
|
||||
clock="end"
|
||||
onBranch={onFork === undefined || seq === undefined ? undefined : () => { onFork(seq) }}
|
||||
branchUnavailable={forkUnavailable}
|
||||
|
||||
@@ -36,6 +36,7 @@ import { GenericCommandCard } from './GenericCommandCard.tsx'
|
||||
import { GenericToolCard } from './GenericToolCard.tsx'
|
||||
import { MessageItem, PendingSteeringBubble } from './MessageItem.tsx'
|
||||
import { formatRunDuration } from './message-chrome.ts'
|
||||
import { deriveTurnMetrics } from './turn-metrics.ts'
|
||||
import css from './ChatView.module.css'
|
||||
|
||||
const FOLLOW_THRESHOLD = 24
|
||||
@@ -362,6 +363,7 @@ export function ChatView({
|
||||
const actionSeqs = useMemo(() => assistantActionsSeqs(nodes), [nodes])
|
||||
const branchSeqs = useMemo(() => messageBranchSeqs(nodes, turnEnds), [nodes, turnEnds])
|
||||
const runningTurnStart = useMemo(() => runningTurnStartTime(turnTimings), [turnTimings])
|
||||
const turnMetrics = useMemo(() => deriveTurnMetrics(nodes), [nodes])
|
||||
|
||||
const listRef = useRef<HTMLDivElement | null>(null)
|
||||
const columnRef = useRef<HTMLDivElement | null>(null)
|
||||
@@ -599,6 +601,9 @@ export function ChatView({
|
||||
const node: ConversationNode = item.node
|
||||
if (node.kind === 'assistant') {
|
||||
const timing = actionSeqs.has(node.seq) ? turnTimings.get(node.turn) : undefined
|
||||
// Metrics gate on the settled in-window timing: turn/start loaded means
|
||||
// every step of the turn is loaded, so first-step TTFT is genuine.
|
||||
const metrics = timing?.endTime === undefined ? undefined : turnMetrics.get(node.turn)
|
||||
return (
|
||||
<AssistantMarkdown
|
||||
blocks={node.blocks}
|
||||
@@ -608,6 +613,8 @@ export function ChatView({
|
||||
runMs={timing?.endTime === undefined
|
||||
? undefined
|
||||
: Math.max(0, timing.endTime - timing.startTime)}
|
||||
ttftMs={metrics?.ttftMs}
|
||||
tokensPerSecond={metrics?.tokensPerSecond}
|
||||
seq={node.seq}
|
||||
onFork={forkAt}
|
||||
forkUnavailable={!branchSeqs.has(node.seq)}
|
||||
|
||||
@@ -6,7 +6,7 @@ import {
|
||||
IconBranchOutline16, IconCheckOutline16, IconCopyOutline16, Tooltip, writeClipboard,
|
||||
} from '@deepseek-ai/dsh-client-ui-primitives'
|
||||
import type { ChatViewSlotProps } from '../contract/slots.ts'
|
||||
import { formatMessageClock, formatRunDuration } from './message-chrome.ts'
|
||||
import { formatLatencySeconds, formatMessageClock, formatRunDuration, formatTokensPerSecond } from './message-chrome.ts'
|
||||
import { useCalendarDay } from './use-calendar-day.ts'
|
||||
import css from './MessageIconActions.module.css'
|
||||
|
||||
@@ -17,6 +17,10 @@ export interface MessageIconActionsProps {
|
||||
time?: number | undefined
|
||||
/** Turn wall time in ms, appended to the clock as `· Ran for 15s`; omitted when the turn's start is unknown. */
|
||||
runMs?: number | undefined
|
||||
/** Turn first-step TTFT in ms, appended as `· TTFT 1.2s`; omitted when unrecorded. */
|
||||
ttftMs?: number | undefined
|
||||
/** Turn decode throughput, appended as `· 34 tok/s`; omitted when unrecorded. */
|
||||
tokensPerSecond?: number | undefined
|
||||
/** Clock before icons (user) or after (assistant). */
|
||||
clock: 'start' | 'end'
|
||||
/** Fork the session at this message; omission hides the branch action. */
|
||||
@@ -37,7 +41,7 @@ export interface MessageIconActionsProps {
|
||||
* @returns The actions row element.
|
||||
*/
|
||||
export function MessageIconActions({
|
||||
text, time, runMs, clock, onBranch, branchUnavailable = false, showBranch = true, className, t,
|
||||
text, time, runMs, ttftMs, tokensPerSecond, clock, onBranch, branchUnavailable = false, showBranch = true, className, t,
|
||||
}: MessageIconActionsProps) {
|
||||
const day = useCalendarDay()
|
||||
const reasonId = useId()
|
||||
@@ -67,15 +71,36 @@ export function MessageIconActions({
|
||||
}, 1000)
|
||||
})
|
||||
}, [copied, text])
|
||||
// The dot is decorative and stays hidden, but its margins separate the
|
||||
// readings only on screen: without the flanking spaces a reader hears one
|
||||
// run-on string ("Ran for 13sTTFT 0.2s12 tok/s") instead of three facts.
|
||||
const clockEl = time === undefined ? null : (
|
||||
<span className={clock === 'start' ? css.timeStart : css.timeEnd}>
|
||||
{formatMessageClock(time, t, day)}
|
||||
{runMs !== undefined && (
|
||||
<>
|
||||
{' '}
|
||||
<span className={css.runTimeDot} aria-hidden>·</span>
|
||||
{' '}
|
||||
{t('message.ranFor', { duration: formatRunDuration(runMs, t) })}
|
||||
</>
|
||||
)}
|
||||
{ttftMs !== undefined && (
|
||||
<>
|
||||
{' '}
|
||||
<span className={css.runTimeDot} aria-hidden>·</span>
|
||||
{' '}
|
||||
{t('message.ttft', { seconds: formatLatencySeconds(ttftMs) })}
|
||||
</>
|
||||
)}
|
||||
{tokensPerSecond !== undefined && (
|
||||
<>
|
||||
{' '}
|
||||
<span className={css.runTimeDot} aria-hidden>·</span>
|
||||
{' '}
|
||||
{t('message.tokensPerSecond', { tps: formatTokensPerSecond(tokensPerSecond) })}
|
||||
</>
|
||||
)}
|
||||
</span>
|
||||
)
|
||||
return (
|
||||
|
||||
@@ -2,10 +2,14 @@
|
||||
// Mounted on 'conversation.composer.dock' so it sticks with the composer in the
|
||||
// active conversation scrollport (see ConversationRoot data-conversation-scroll).
|
||||
|
||||
import { Fragment, memo, useMemo } from 'react'
|
||||
import { Fragment, memo, useLayoutEffect, useMemo, useRef, useState } from 'react'
|
||||
import { Tooltip } from '@deepseek-ai/dsh-client-ui-primitives'
|
||||
import type { ConversationSnapshot, UseProjection } from '@deepseek-ai/dsh-client-runtime/client'
|
||||
import type { SnapshotSelectorHook } from '@deepseek-ai/dsh-client-ui-slots'
|
||||
import type { ContextPressureProjection, TokenUsageProjection } from '@deepseek-ai/dsh-token-meter/client'
|
||||
import type { ComposerBarProps } from '../contract/slots.ts'
|
||||
import { formatTokensPerSecond } from './message-chrome.ts'
|
||||
import { assistantStepReading } from './turn-metrics.ts'
|
||||
import css from './StatsLine.module.css'
|
||||
|
||||
interface WindowStats {
|
||||
@@ -15,6 +19,14 @@ interface WindowStats {
|
||||
llmMs: number
|
||||
/** Summed tool wall time (tool/call → tool/result); 0 when no pair is in-window. */
|
||||
toolMs: number
|
||||
/** Summed first-token latency over `ttftSteps`; 0 when no step records it. */
|
||||
ttftMs: number
|
||||
/** Steps carrying a recorded TTFT. */
|
||||
ttftSteps: number
|
||||
/** Summed decode wall time over steps that also report output tokens. */
|
||||
decodeMs: number
|
||||
/** Summed output tokens over the same decode-timed steps. */
|
||||
decodeTokens: number
|
||||
}
|
||||
|
||||
/**
|
||||
@@ -32,6 +44,10 @@ export function deriveStats(nodes: ConversationSnapshot['nodes']): WindowStats {
|
||||
let steps = 0
|
||||
let llmMs = 0
|
||||
let toolMs = 0
|
||||
let ttftMs = 0
|
||||
let ttftSteps = 0
|
||||
let decodeMs = 0
|
||||
let decodeTokens = 0
|
||||
for (const node of nodes) {
|
||||
if (node.kind === 'tool-result') {
|
||||
if (node.callTime !== null) toolMs += Math.max(0, node.time - node.callTime)
|
||||
@@ -43,8 +59,17 @@ export function deriveStats(nodes: ConversationSnapshot['nodes']): WindowStats {
|
||||
if (node.timing !== undefined && node.timing.stepStartTime !== null) {
|
||||
llmMs += Math.max(0, node.timing.completedTime - node.timing.stepStartTime)
|
||||
}
|
||||
const reading = assistantStepReading(node)
|
||||
if (reading.ttftMs !== null) {
|
||||
ttftMs += reading.ttftMs
|
||||
ttftSteps += 1
|
||||
}
|
||||
if (reading.decodeMs !== null && reading.outputTokens !== null) {
|
||||
decodeMs += reading.decodeMs
|
||||
decodeTokens += reading.outputTokens
|
||||
}
|
||||
}
|
||||
return { turns: turns.size, steps, llmMs, toolMs }
|
||||
return { turns: turns.size, steps, llmMs, toolMs, ttftMs, ttftSteps, decodeMs, decodeTokens }
|
||||
}
|
||||
|
||||
/**
|
||||
@@ -84,30 +109,41 @@ export function cacheHitPercent(usage: TokenUsageProjection): number | null {
|
||||
: Math.round(usage.cacheReadTokens / denominator * 100)
|
||||
}
|
||||
|
||||
/** Sum the three disjoint prompt-side billing buckets. */
|
||||
function billedInputTokens(usage: TokenUsageProjection): number {
|
||||
/**
|
||||
* Sum the three disjoint prompt-side billing buckets.
|
||||
* @param usage - the session's token-usage projection value.
|
||||
* @returns billed input tokens.
|
||||
*/
|
||||
export function billedInputTokens(usage: TokenUsageProjection): number {
|
||||
return usage.uncachedInputTokens + usage.cacheReadTokens + usage.cacheWriteTokens
|
||||
}
|
||||
|
||||
interface ContextOccupancy {
|
||||
percent: number
|
||||
usedTokens: number
|
||||
contextWindow: number
|
||||
}
|
||||
|
||||
/**
|
||||
* Approximate context occupancy, using the TUI's integer rounding and upper
|
||||
* clamp. The numerator and capacity are independent last-wins projection
|
||||
* fields, so this is a reference figure rather than an exact measurement of one
|
||||
* request (see the token-meter README).
|
||||
* clamp. The numerator is `projectedTokens` — the provider sample carried
|
||||
* forward over the surface's movement since — so compaction shows immediately
|
||||
* instead of waiting for the next request to report usage; it falls back to the
|
||||
* bare sample only for a log whose projection predates that field. Numerator
|
||||
* and capacity remain independent last-wins projection fields, so this is a
|
||||
* reference figure rather than an exact measurement of one request (see the
|
||||
* token-meter README).
|
||||
* @param pressure - the session's context-pressure projection value.
|
||||
* @returns occupancy and its denominator, or null until both values are known.
|
||||
* @returns occupancy with its numerator and denominator, or null until both values are known.
|
||||
*/
|
||||
export function contextOccupancy(
|
||||
pressure: ContextPressureProjection | undefined,
|
||||
): ContextOccupancy | null {
|
||||
if (pressure?.pressureTokens === undefined || pressure.contextWindow === undefined) return null
|
||||
const usedTokens = pressure?.projectedTokens ?? pressure?.pressureTokens
|
||||
if (usedTokens === undefined || pressure?.contextWindow === undefined) return null
|
||||
return {
|
||||
percent: Math.min(100, Math.round(pressure.pressureTokens / pressure.contextWindow * 100)),
|
||||
percent: Math.min(100, Math.round(usedTokens / pressure.contextWindow * 100)),
|
||||
usedTokens,
|
||||
contextWindow: pressure.contextWindow,
|
||||
}
|
||||
}
|
||||
@@ -116,46 +152,73 @@ export function contextOccupancy(
|
||||
export interface StatsLineProps {
|
||||
useSession: SnapshotSelectorHook<ConversationSnapshot>
|
||||
useProjection: UseProjection
|
||||
/** The owning dock's locale seat. */
|
||||
t: ComposerBarProps['t']
|
||||
}
|
||||
|
||||
export const StatsLine = memo(function StatsLine({ useSession, useProjection }: StatsLineProps) {
|
||||
export const StatsLine = memo(function StatsLine({ useSession, useProjection, t }: StatsLineProps) {
|
||||
const nodes = useSession(s => s.nodes)
|
||||
const usage = useProjection('tokenUsage')
|
||||
const pressure = useProjection('contextPressure')
|
||||
const stats = useMemo(() => deriveStats(nodes), [nodes])
|
||||
// Pipe-separated groups (figma stats strip); a group with no data drops out whole.
|
||||
const groups: string[] = []
|
||||
if (stats.steps > 0) {
|
||||
groups.push(`${stats.turns} turns · ${stats.steps} steps`)
|
||||
groups.push(t('stats.counts', { turns: stats.turns, steps: stats.steps }))
|
||||
const durations: string[] = []
|
||||
if (stats.llmMs > 0) durations.push(`LLM ${formatDuration(stats.llmMs)}`)
|
||||
if (stats.toolMs > 0) durations.push(`Tool call ${formatDuration(stats.toolMs)}`)
|
||||
if (stats.llmMs > 0) durations.push(t('stats.llm', { duration: formatDuration(stats.llmMs) }))
|
||||
if (stats.toolMs > 0) durations.push(t('stats.toolCall', { duration: formatDuration(stats.toolMs) }))
|
||||
if (durations.length > 0) groups.push(durations.join(' · '))
|
||||
// Window-scoped like the wall times above: averages describe loaded steps.
|
||||
const speeds: string[] = []
|
||||
if (stats.ttftSteps > 0) {
|
||||
speeds.push(t('stats.ttftAverage', { duration: formatDuration(stats.ttftMs / stats.ttftSteps) }))
|
||||
}
|
||||
if (stats.decodeMs > 0) {
|
||||
speeds.push(t('stats.tokensPerSecond', {
|
||||
throughput: formatTokensPerSecond(stats.decodeTokens / (stats.decodeMs / 1_000)),
|
||||
}))
|
||||
}
|
||||
if (speeds.length > 0) groups.push(speeds.join(' · '))
|
||||
}
|
||||
const context = contextOccupancy(pressure)
|
||||
if (context !== null) {
|
||||
groups.push(`Context ${context.percent}% of ${formatTokens(context.contextWindow)}`)
|
||||
}
|
||||
// Context occupancy deliberately lives on the composer's ContextMeter ring,
|
||||
// not here — one home per fact.
|
||||
// Billing rides the durable projection, so these survive paging and
|
||||
// compaction. Suppress the empty projection on a brand-new session.
|
||||
if (usage !== undefined
|
||||
&& (stats.steps > 0 || billedInputTokens(usage) > 0 || usage.outputTokens > 0)) {
|
||||
const cacheHit = cacheHitPercent(usage)
|
||||
if (cacheHit !== null) groups.push(`Cache hit ${cacheHit}%`)
|
||||
groups.push(
|
||||
`Input ${formatTokens(billedInputTokens(usage))} tok`
|
||||
+ ` · Output ${formatTokens(usage.outputTokens)} tok`,
|
||||
)
|
||||
if (cacheHit !== null) groups.push(t('stats.cacheHit', { percent: cacheHit }))
|
||||
groups.push(t('stats.tokens', {
|
||||
input: formatTokens(billedInputTokens(usage)),
|
||||
output: formatTokens(usage.outputTokens),
|
||||
}))
|
||||
}
|
||||
const line = groups.join(' | ')
|
||||
// The row elides with ellipsis when overlong; a delayed hover tooltip carries
|
||||
// the full line, enabled only while content is actually clipped.
|
||||
const rootRef = useRef<HTMLDivElement | null>(null)
|
||||
const [truncated, setTruncated] = useState(false)
|
||||
useLayoutEffect(() => {
|
||||
const el = rootRef.current
|
||||
if (el === null) return
|
||||
const measure = () => { setTruncated(el.scrollWidth > el.clientWidth) }
|
||||
measure()
|
||||
if (typeof ResizeObserver === 'undefined') return
|
||||
const observer = new ResizeObserver(measure)
|
||||
observer.observe(el)
|
||||
return () => { observer.disconnect() }
|
||||
}, [line])
|
||||
if (groups.length === 0) return null
|
||||
return (
|
||||
<div className={css.root}>
|
||||
{groups.map((group, i) => (
|
||||
<Fragment key={group}>
|
||||
{i > 0 && <><span className={css.sep} aria-hidden>|</span>{' '}</>}
|
||||
<span>{group}</span>
|
||||
</Fragment>
|
||||
))}
|
||||
</div>
|
||||
<Tooltip label={line} side="top" delayMs={500} disabled={!truncated}>
|
||||
<div ref={rootRef} className={css.root}>
|
||||
{groups.map((group, i) => (
|
||||
<Fragment key={group}>
|
||||
{i > 0 && <><span className={css.sep} aria-hidden>|</span>{' '}</>}
|
||||
<span>{group}</span>
|
||||
</Fragment>
|
||||
))}
|
||||
</div>
|
||||
</Tooltip>
|
||||
)
|
||||
})
|
||||
|
||||
@@ -48,6 +48,27 @@ export function formatRunDuration(ms: number, t: RunDurationTranslate): string {
|
||||
: t('duration.seconds', { seconds })
|
||||
}
|
||||
|
||||
/**
|
||||
* Sub-turn latency figure: one decimal under ten seconds, whole seconds
|
||||
* beyond. Unit-less so the locale template owns the second suffix.
|
||||
* @param ms - Latency in milliseconds (negatives clamp to zero).
|
||||
* @returns Display number in seconds without unit.
|
||||
*/
|
||||
export function formatLatencySeconds(ms: number): string {
|
||||
const s = Math.max(0, ms) / 1000
|
||||
return s < 10 ? String(Math.round(s * 10) / 10) : String(Math.round(s))
|
||||
}
|
||||
|
||||
/**
|
||||
* Decode-throughput figure: whole tokens from ten up, one decimal below.
|
||||
* @param tps - Tokens per second.
|
||||
* @returns Display number without unit.
|
||||
*/
|
||||
export function formatTokensPerSecond(tps: number): string {
|
||||
const clamped = Math.max(0, tps)
|
||||
return clamped >= 10 ? String(Math.round(clamped)) : String(Math.round(clamped * 10) / 10)
|
||||
}
|
||||
|
||||
/**
|
||||
* Compact local timestamp for message IconActions. Same calendar day →
|
||||
* `HH:mm`; earlier this year → the `clock.md` date template + clock; other
|
||||
|
||||
@@ -0,0 +1,97 @@
|
||||
// Latency/throughput folds shared by the settled turn footer and StatsLine.
|
||||
|
||||
import type { ConversationSnapshot } from '@deepseek-ai/dsh-client-runtime/client'
|
||||
|
||||
/** Latency and decode-throughput readings for one turn's footer. */
|
||||
export interface TurnMetrics {
|
||||
/** First-step TTFT in ms; absent when that step carries no recorded timing. */
|
||||
ttftMs?: number
|
||||
/** Decode throughput over steps carrying both timing and provider usage. */
|
||||
tokensPerSecond?: number
|
||||
}
|
||||
|
||||
/** One assistant step's derivable latency facts; null marks an unrecorded part. */
|
||||
export interface StepReading {
|
||||
/** step/start → first token delta, in ms. */
|
||||
ttftMs: number | null
|
||||
/** First token delta → final message, in ms. */
|
||||
decodeMs: number | null
|
||||
/** Provider-reported completion tokens. */
|
||||
outputTokens: number | null
|
||||
}
|
||||
|
||||
interface UsageLike {
|
||||
outputTokens?: number
|
||||
}
|
||||
|
||||
type AssistantNode = Extract<ConversationSnapshot['nodes'][number], { kind: 'assistant' }>
|
||||
|
||||
function usageOutputTokens(usage: unknown): number | null {
|
||||
if (typeof usage !== 'object' || usage === null) return null
|
||||
const value = (usage as UsageLike).outputTokens
|
||||
return typeof value === 'number' && Number.isFinite(value) && value >= 0 ? value : null
|
||||
}
|
||||
|
||||
/**
|
||||
* Read one assistant node's TTFT, decode wall time, and output tokens.
|
||||
* @param node - A settled assistant node.
|
||||
* @returns Per-part readings with `null` for unrecorded values.
|
||||
*/
|
||||
export function assistantStepReading(node: AssistantNode): StepReading {
|
||||
const timing = node.timing
|
||||
const ttftMs = timing !== undefined && timing.stepStartTime !== null && timing.firstTokenTime !== null
|
||||
? Math.max(0, timing.firstTokenTime - timing.stepStartTime)
|
||||
: null
|
||||
const decodeMs = timing !== undefined && timing.firstTokenTime !== null
|
||||
? Math.max(0, timing.completedTime - timing.firstTokenTime)
|
||||
: null
|
||||
return { ttftMs, decodeMs, outputTokens: usageOutputTokens(node.usage) }
|
||||
}
|
||||
|
||||
interface TurnFold {
|
||||
firstStep: number
|
||||
firstStepTtftMs: number | null
|
||||
decodeMs: number
|
||||
outputTokens: number
|
||||
sampled: boolean
|
||||
}
|
||||
|
||||
/**
|
||||
* Fold assistant nodes into per-turn footer metrics.
|
||||
*
|
||||
* TTFT is the turn's lowest-step request-dispatch-to-first-token reading, so
|
||||
* it is only meaningful when the turn's start is inside
|
||||
* the loaded window (the caller gates on `turnTimings`, which shares that
|
||||
* window). Throughput divides summed output tokens by summed decode wall time,
|
||||
* counting only steps that carry both.
|
||||
* @param nodes - Snapshot nodes of the loaded window.
|
||||
* @returns Turn number → available metrics; turns with none are absent.
|
||||
*/
|
||||
export function deriveTurnMetrics(nodes: ConversationSnapshot['nodes']): Map<number, TurnMetrics> {
|
||||
const folds = new Map<number, TurnFold>()
|
||||
for (const node of nodes) {
|
||||
if (node.kind !== 'assistant') continue
|
||||
const reading = assistantStepReading(node)
|
||||
let fold = folds.get(node.turn)
|
||||
if (fold === undefined) {
|
||||
fold = { firstStep: node.step, firstStepTtftMs: reading.ttftMs, decodeMs: 0, outputTokens: 0, sampled: false }
|
||||
folds.set(node.turn, fold)
|
||||
} else if (node.step < fold.firstStep) {
|
||||
fold.firstStep = node.step
|
||||
fold.firstStepTtftMs = reading.ttftMs
|
||||
}
|
||||
if (reading.decodeMs !== null && reading.outputTokens !== null) {
|
||||
fold.decodeMs += reading.decodeMs
|
||||
fold.outputTokens += reading.outputTokens
|
||||
fold.sampled = true
|
||||
}
|
||||
}
|
||||
const metrics = new Map<number, TurnMetrics>()
|
||||
for (const [turn, fold] of folds) {
|
||||
const entry: TurnMetrics = {}
|
||||
if (fold.firstStepTtftMs !== null) entry.ttftMs = fold.firstStepTtftMs
|
||||
if (fold.sampled && fold.decodeMs > 0) entry.tokensPerSecond = fold.outputTokens / (fold.decodeMs / 1000)
|
||||
if (entry.ttftMs !== undefined || entry.tokensPerSecond !== undefined) metrics.set(turn, entry)
|
||||
}
|
||||
return metrics
|
||||
}
|
||||
@@ -23,6 +23,18 @@ export const zh = {
|
||||
'input.stop': '停止生成',
|
||||
'input.send': '发送消息',
|
||||
'input.accessMode': '访问模式，当前：{name}',
|
||||
'context.aria': '上下文已用 {percent}',
|
||||
'context.used': '上下文已用',
|
||||
'context.system': '系统提示词',
|
||||
'context.tools': '工具',
|
||||
'context.messages': '对话消息',
|
||||
'stats.counts': '{turns} 轮 · {steps} 步',
|
||||
'stats.llm': 'LLM {duration}',
|
||||
'stats.toolCall': '工具调用 {duration}',
|
||||
'stats.ttftAverage': '首 token 平均 {duration}',
|
||||
'stats.tokensPerSecond': '{throughput} tok/s',
|
||||
'stats.cacheHit': '缓存命中 {percent}%',
|
||||
'stats.tokens': '输入 {input} tok · 输出 {output} tok',
|
||||
'settings.enter.title': '繁忙时 Enter 键行为',
|
||||
'settings.enter.description': '仅在智能体运行时生效；Cmd/Ctrl+Enter 使用另一行为',
|
||||
'settings.enter.queue': '排队发送',
|
||||
@@ -83,6 +95,8 @@ export const zh = {
|
||||
'message.retry.failure': '失败原因：',
|
||||
'message.turnError': '本轮运行失败',
|
||||
'message.ranFor': '用时 {duration}',
|
||||
'message.ttft': '首 token {seconds}秒',
|
||||
'message.tokensPerSecond': '{tps} tok/s',
|
||||
'duration.seconds': '{seconds}秒',
|
||||
'duration.minutes': '{minutes}分{seconds}秒',
|
||||
'command.running': '执行中…',
|
||||
@@ -148,6 +162,18 @@ export const en = {
|
||||
'input.stop': 'Stop generating',
|
||||
'input.send': 'Send message',
|
||||
'input.accessMode': 'Access mode, current: {name}',
|
||||
'context.aria': '{percent} of context used',
|
||||
'context.used': 'of context used',
|
||||
'context.system': 'System prompt',
|
||||
'context.tools': 'Tools',
|
||||
'context.messages': 'Messages',
|
||||
'stats.counts': '{turns} turns · {steps} steps',
|
||||
'stats.llm': 'LLM {duration}',
|
||||
'stats.toolCall': 'Tool call {duration}',
|
||||
'stats.ttftAverage': 'TTFT avg {duration}',
|
||||
'stats.tokensPerSecond': '{throughput} tok/s',
|
||||
'stats.cacheHit': 'Cache hit {percent}%',
|
||||
'stats.tokens': 'Input {input} tok · Output {output} tok',
|
||||
'settings.enter.title': 'Enter behavior while busy',
|
||||
'settings.enter.description': 'Busy only; Cmd/Ctrl+Enter uses the other behavior',
|
||||
'settings.enter.queue': 'Queue',
|
||||
@@ -208,6 +234,8 @@ export const en = {
|
||||
'message.retry.failure': 'Failure reason: ',
|
||||
'message.turnError': 'This turn failed',
|
||||
'message.ranFor': 'Ran for {duration}',
|
||||
'message.ttft': 'TTFT {seconds}s',
|
||||
'message.tokensPerSecond': '{tps} tok/s',
|
||||
'duration.seconds': '{seconds}s',
|
||||
'duration.minutes': '{minutes}m {seconds}s',
|
||||
'command.running': 'Running…',
|
||||
|
||||
@@ -0,0 +1,147 @@
|
||||
/* Context-occupancy ring beside the send button plus its click-open breakdown
|
||||
panel (menu surface: r12, inverted hairline, shadow-lv3). */
|
||||
|
||||
.root {
|
||||
position: relative;
|
||||
display: inline-flex;
|
||||
}
|
||||
|
||||
/* Same 28px circular hit target family as the composer's attach button. */
|
||||
.trigger {
|
||||
display: grid;
|
||||
place-items: center;
|
||||
flex: none;
|
||||
width: 28px;
|
||||
height: 28px;
|
||||
border: none;
|
||||
border-radius: 999px;
|
||||
background: transparent;
|
||||
color: var(--dsw-alias-label-secondary);
|
||||
cursor: pointer;
|
||||
}
|
||||
|
||||
.trigger:hover {
|
||||
background: var(--dsw-alias-interactive-bg-hover);
|
||||
}
|
||||
|
||||
.track {
|
||||
fill: none;
|
||||
stroke: var(--dsw-alias-border-l3);
|
||||
stroke-width: 2;
|
||||
}
|
||||
|
||||
.fill {
|
||||
fill: none;
|
||||
stroke: var(--dsw-alias-label-tertiary);
|
||||
stroke-width: 2;
|
||||
stroke-linecap: round;
|
||||
}
|
||||
|
||||
.panel {
|
||||
position: absolute;
|
||||
bottom: calc(100% + 8px);
|
||||
right: 0;
|
||||
z-index: 100;
|
||||
box-sizing: border-box;
|
||||
width: 264px;
|
||||
padding: 12px;
|
||||
border: 1px solid var(--dsw-alias-border-inverted);
|
||||
border-radius: 12px;
|
||||
background: var(--dsw-specific-menu);
|
||||
box-shadow: var(--dsw-shadow-lv3);
|
||||
font-size: 12px;
|
||||
line-height: 20px;
|
||||
color: var(--dsw-alias-label-secondary);
|
||||
cursor: default;
|
||||
}
|
||||
|
||||
.header {
|
||||
display: flex;
|
||||
align-items: center;
|
||||
gap: 6px;
|
||||
}
|
||||
|
||||
.figures {
|
||||
margin-left: auto;
|
||||
font-weight: 500;
|
||||
font-variant-numeric: tabular-nums;
|
||||
color: var(--dsw-alias-label-primary);
|
||||
}
|
||||
|
||||
.percent {
|
||||
font-weight: 500;
|
||||
color: var(--dsw-alias-label-primary);
|
||||
}
|
||||
|
||||
.headline {
|
||||
color: var(--dsw-alias-label-tertiary);
|
||||
}
|
||||
|
||||
/* The headline brackets the reading, so the side a locale leaves empty must
|
||||
drop out of the flex row rather than spend a gap. */
|
||||
.headline:empty {
|
||||
display: none;
|
||||
}
|
||||
|
||||
.bar {
|
||||
display: flex;
|
||||
gap: 1px;
|
||||
margin: 10px 0 12px;
|
||||
height: 4px;
|
||||
border-radius: 999px;
|
||||
background: var(--dsw-alias-interactive-bg-hover);
|
||||
overflow: hidden;
|
||||
}
|
||||
|
||||
.segment {
|
||||
flex: none;
|
||||
min-width: 2px;
|
||||
height: 100%;
|
||||
border-radius: 1px;
|
||||
background: var(--meter-tint, var(--dsw-alias-label-tertiary));
|
||||
}
|
||||
|
||||
.swatch {
|
||||
display: inline-block;
|
||||
margin-right: 6px;
|
||||
width: 8px;
|
||||
height: 8px;
|
||||
border-radius: 2px;
|
||||
background: var(--meter-tint);
|
||||
vertical-align: baseline;
|
||||
}
|
||||
|
||||
.colorSystem {
|
||||
--meter-tint: var(--dsw-static-neutral-bluish-400);
|
||||
}
|
||||
|
||||
.colorTools {
|
||||
/* The design platform ships no purple static token; violet-400 literal. */
|
||||
--meter-tint: rgb(167, 139, 250);
|
||||
}
|
||||
|
||||
.colorMessages {
|
||||
--meter-tint: var(--dsw-static-blue-450);
|
||||
}
|
||||
|
||||
.rows {
|
||||
margin: 6px 0 0;
|
||||
}
|
||||
|
||||
.row {
|
||||
display: flex;
|
||||
align-items: center;
|
||||
justify-content: space-between;
|
||||
gap: 12px;
|
||||
padding: 2px 0;
|
||||
}
|
||||
|
||||
.row dt {
|
||||
color: var(--dsw-alias-label-secondary);
|
||||
}
|
||||
|
||||
.row dd {
|
||||
margin: 0;
|
||||
font-variant-numeric: tabular-nums;
|
||||
color: var(--dsw-alias-label-primary);
|
||||
}
|
||||
@@ -0,0 +1,153 @@
|
||||
/** Composer context-occupancy meter: a ring beside the send button fed by the
|
||||
* `contextPressure` projection, with a click-open panel of the heuristic
|
||||
* `contextBreakdown` composition (system prompt, tools, conversation).
|
||||
* Renders nothing until a provider reports both pressure and a route capacity
|
||||
* (same gate as the stats row used). */
|
||||
|
||||
import { useEffect, useRef, useState } from 'react'
|
||||
import type { UseProjection } from '@deepseek-ai/dsh-client-runtime/client'
|
||||
// Type-only: the `contextPressure` / `contextBreakdown` projection key merges.
|
||||
import type {} from '@deepseek-ai/dsh-token-meter/client'
|
||||
import { Tooltip } from '@deepseek-ai/dsh-client-ui-primitives'
|
||||
import type { ComposerBarProps } from '../contract/slots.ts'
|
||||
import { contextOccupancy, formatTokens } from '../chat/StatsLine.tsx'
|
||||
import css from './ContextMeter.module.css'
|
||||
|
||||
/** Ring geometry: 14px viewBox, 2px stroke. */
|
||||
const RADIUS = 5.5
|
||||
const CIRCUMFERENCE = 2 * Math.PI * RADIUS
|
||||
|
||||
/**
|
||||
* Marker the localized occupancy sentence is split on, so the panel headline
|
||||
* keeps the reading in its own tone while each locale still owns the word
|
||||
* order (`45% of context used` / `上下文已用 45%`).
|
||||
*/
|
||||
const READING_SLOT = '\u0000'
|
||||
|
||||
/** Panel legend rows, in bar-segment order; each color class carries the shared swatch/segment tint. */
|
||||
const ROWS = [
|
||||
{ key: 'systemTokens', label: 'context.system', color: css.colorSystem },
|
||||
{ key: 'toolsTokens', label: 'context.tools', color: css.colorTools },
|
||||
{ key: 'messageTokens', label: 'context.messages', color: css.colorMessages },
|
||||
] as const
|
||||
|
||||
export interface ContextMeterProps {
|
||||
useProjection: UseProjection
|
||||
/** The owning bar's locale seat, passed down as a plain prop. */
|
||||
t: ComposerBarProps['t']
|
||||
}
|
||||
|
||||
export function ContextMeter({ useProjection, t }: ContextMeterProps) {
|
||||
const pressure = useProjection('contextPressure')
|
||||
const breakdown = useProjection('contextBreakdown')
|
||||
const [open, setOpen] = useState(false)
|
||||
const rootRef = useRef<HTMLSpanElement | null>(null)
|
||||
const context = contextOccupancy(pressure)
|
||||
const available = context !== null
|
||||
|
||||
// A model switch can temporarily remove capacity while this component stays
|
||||
// mounted. Close the now-unavailable panel instead of preserving stale UI.
|
||||
useEffect(() => {
|
||||
if (!available && open) setOpen(false)
|
||||
}, [available, open])
|
||||
|
||||
// Outside click / Escape close, one document listener while open (Menu's pattern).
|
||||
useEffect(() => {
|
||||
if (!open || !available) return
|
||||
const onPointerDown = (e: PointerEvent): void => {
|
||||
if (e.target instanceof Node && rootRef.current?.contains(e.target) === true) return
|
||||
setOpen(false)
|
||||
}
|
||||
const onKeyDown = (e: KeyboardEvent): void => {
|
||||
if (e.key === 'Escape') setOpen(false)
|
||||
}
|
||||
document.addEventListener('pointerdown', onPointerDown)
|
||||
document.addEventListener('keydown', onKeyDown)
|
||||
return () => {
|
||||
document.removeEventListener('pointerdown', onPointerDown)
|
||||
document.removeEventListener('keydown', onKeyDown)
|
||||
}
|
||||
}, [available, open])
|
||||
|
||||
if (context === null) return null
|
||||
const percent = context.percent
|
||||
const reading = `${percent}%`
|
||||
const [headBefore = '', headAfter = ''] = t('context.aria', { percent: READING_SLOT })
|
||||
.split(READING_SLOT)
|
||||
.map(part => part.trim())
|
||||
|
||||
// The bar's overall length stays the provider-exact percent; the heuristic
|
||||
// breakdown only proportions its colored parts. A zero-width part is dropped
|
||||
// instead of rendered: `.segment`'s min-width keeps a hairline part visible,
|
||||
// which at 0% occupancy would draw a filled bar over an empty context.
|
||||
const breakdownTotal = breakdown === undefined
|
||||
? 0
|
||||
: breakdown.systemTokens + breakdown.toolsTokens + breakdown.messageTokens
|
||||
const parts = breakdown === undefined || breakdownTotal === 0
|
||||
? [{ key: 'total', color: undefined, width: percent }]
|
||||
: ROWS.map(row => ({ key: row.key, color: row.color, width: percent * breakdown[row.key] / breakdownTotal }))
|
||||
const segments = parts.filter(part => part.width > 0)
|
||||
|
||||
return (
|
||||
<span ref={rootRef} className={css.root}>
|
||||
<Tooltip label={t('context.aria', { percent: reading })} side="top" delayMs={200} disabled={open}>
|
||||
<button
|
||||
type="button"
|
||||
className={css.trigger}
|
||||
aria-label={t('context.aria', { percent: reading })}
|
||||
aria-haspopup="dialog"
|
||||
aria-expanded={open}
|
||||
onClick={() => { setOpen(!open) }}
|
||||
>
|
||||
<svg viewBox="0 0 14 14" width="14" height="14" aria-hidden>
|
||||
<circle className={css.track} cx="7" cy="7" r={RADIUS} />
|
||||
<circle
|
||||
className={css.fill}
|
||||
cx="7"
|
||||
cy="7"
|
||||
r={RADIUS}
|
||||
strokeDasharray={`${CIRCUMFERENCE * percent / 100} ${CIRCUMFERENCE}`}
|
||||
transform="rotate(-90 7 7)"
|
||||
/>
|
||||
</svg>
|
||||
</button>
|
||||
</Tooltip>
|
||||
{open && (
|
||||
<div className={css.panel} role="dialog" aria-label={t('context.used')}>
|
||||
<div className={css.header}>
|
||||
{/* Empty sides collapse through `.headline:empty` so the locale that
|
||||
needs no leading (or trailing) text spends no header gap. */}
|
||||
<span className={css.headline}>{headBefore}</span>
|
||||
<span className={css.percent}>{reading}</span>
|
||||
<span className={css.headline}>{headAfter}</span>
|
||||
<span className={css.figures}>
|
||||
{`~${formatTokens(context.usedTokens)} / ${formatTokens(context.contextWindow)}`}
|
||||
</span>
|
||||
</div>
|
||||
<div className={css.bar}>
|
||||
{segments.map(segment => (
|
||||
<div
|
||||
key={segment.key}
|
||||
className={segment.color === undefined ? css.segment : `${css.segment} ${segment.color}`}
|
||||
style={{ width: `${segment.width}%` }}
|
||||
/>
|
||||
))}
|
||||
</div>
|
||||
{breakdown !== undefined && (
|
||||
<dl className={css.rows}>
|
||||
{ROWS.map(row => (
|
||||
<div key={row.key} className={css.row}>
|
||||
<dt>
|
||||
<span className={`${css.swatch} ${row.color}`} aria-hidden />
|
||||
{t(row.label)}
|
||||
</dt>
|
||||
<dd>{`~${formatTokens(breakdown[row.key])}`}</dd>
|
||||
</div>
|
||||
))}
|
||||
</dl>
|
||||
)}
|
||||
</div>
|
||||
)}
|
||||
</span>
|
||||
)
|
||||
}
|
||||
@@ -19,6 +19,7 @@ import type { Translate } from '@deepseek-ai/dsh-client-ui-slots'
|
||||
import type { ComposerBarProps } from '../contract/slots.ts'
|
||||
import { deriveDecorations } from '../input/decorations.ts'
|
||||
import type { DraftDecorations } from '../input/decorations.ts'
|
||||
import { ContextMeter } from './ContextMeter.tsx'
|
||||
import { PermissionSelect } from './PermissionSelect.tsx'
|
||||
import css from './InputBar.module.css'
|
||||
|
||||
@@ -512,6 +513,7 @@ export function InputBar({
|
||||
<div className={css.trailing}>
|
||||
{rightItems}
|
||||
{renderSlot('conversation.input.model', { locked })}
|
||||
<ContextMeter useProjection={useProjection} t={t} />
|
||||
{/* {machineBusy && <span className={css.pending} data-input-pending aria-label="处理中" />} */}
|
||||
<Tooltip label={primaryLabel} side="top" delayMs={500}>
|
||||
<button
|
||||
|
||||
@@ -18,9 +18,18 @@ import { AssistantMarkdown } from '../src/client/chat/AssistantMarkdown.tsx'
|
||||
import { StatsLine, type StatsLineProps } from '../src/client/chat/StatsLine.tsx'
|
||||
import { zh } from '../src/client/locales.ts'
|
||||
|
||||
/** jsdom has no ResizeObserver; StatsLine watches its row for ellipsis truncation through one. */
|
||||
class ResizeObserverStub {
|
||||
observe(): void {}
|
||||
unobserve(): void {}
|
||||
disconnect(): void {}
|
||||
}
|
||||
|
||||
beforeEach(() => { vi.stubGlobal('ResizeObserver', ResizeObserverStub) })
|
||||
afterEach(() => {
|
||||
cleanup()
|
||||
vi.useRealTimers()
|
||||
vi.unstubAllGlobals()
|
||||
})
|
||||
|
||||
// Mirrors the real lookup chain (conversation namespace, then common).
|
||||
@@ -1002,12 +1011,13 @@ describe('small branch tails', () => {
|
||||
const source = { getSnapshot: () => snap, subscribe: () => () => {} }
|
||||
const view = render(
|
||||
<StatsLine
|
||||
t={t}
|
||||
useSession={bindSnapshotSelector(source) as unknown as StatsLineProps['useSession']}
|
||||
useProjection={(key: string) => key === 'tokenUsage'
|
||||
? { uncachedInputTokens: 0, outputTokens: 10, cacheReadTokens: 0, cacheWriteTokens: 0 }
|
||||
: undefined}
|
||||
/>,
|
||||
)
|
||||
expect(view.container.textContent).toBe('1 turns · 1 steps| Input 0 tok · Output 10 tok')
|
||||
expect(view.container.textContent).toBe('1 轮 · 1 步| 输入 0 tok · 输出 10 tok')
|
||||
})
|
||||
})
|
||||
|
||||
@@ -3,25 +3,40 @@
|
||||
// hard acceptance — zero renders during streaming. Bash sample row: ToolRow
|
||||
// chrome (Bash · description) without a row click target.
|
||||
|
||||
import { afterEach, describe, expect, it, vi } from 'vitest'
|
||||
import { act, cleanup, render } from '@testing-library/react'
|
||||
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
|
||||
import { act, cleanup, fireEvent, render } from '@testing-library/react'
|
||||
import type {
|
||||
AssistantMessageNode, ConversationSnapshot, SessionId, SessionListState, ToolResultNode,
|
||||
} from '@deepseek-ai/dsh-client-runtime/client'
|
||||
import { createSnapshotStore } from '@deepseek-ai/dsh-client-runtime/client'
|
||||
import { bindSnapshotSelector } from '@deepseek-ai/dsh-client-web-react'
|
||||
import { makeTranslate } from '@deepseek-ai/dsh-client-test-runtime'
|
||||
import { en as commonEn } from '@deepseek-ai/dsh-client-locale/src/locales/en.ts'
|
||||
import { zh as commonZh } from '@deepseek-ai/dsh-client-locale/src/locales/zh.ts'
|
||||
import { StatsLine, deriveStats, formatDuration, formatTokens, type StatsLineProps } from '../src/client/chat/StatsLine.tsx'
|
||||
import { StatsLine, contextOccupancy, deriveStats, formatDuration, formatTokens, type StatsLineProps } from '../src/client/chat/StatsLine.tsx'
|
||||
import { BashRow } from '../src/client/toolviews/bash-sample.tsx'
|
||||
import { zh } from '../src/client/locales.ts'
|
||||
import { en, zh } from '../src/client/locales.ts'
|
||||
|
||||
type BashRowProps = Parameters<typeof BashRow>[0]
|
||||
|
||||
// Mirrors the real lookup chain (conversation namespace, then common).
|
||||
const t: BashRowProps['t'] = makeTranslate(zh, commonZh)
|
||||
const tEn: StatsLineProps['t'] = makeTranslate(en, commonEn)
|
||||
|
||||
afterEach(cleanup)
|
||||
/** jsdom has no ResizeObserver; StatsLine watches its row for ellipsis truncation through one. */
|
||||
class ResizeObserverStub {
|
||||
observe(): void {}
|
||||
unobserve(): void {}
|
||||
disconnect(): void {}
|
||||
}
|
||||
|
||||
beforeEach(() => { vi.stubGlobal('ResizeObserver', ResizeObserverStub) })
|
||||
afterEach(() => {
|
||||
cleanup()
|
||||
vi.unstubAllGlobals()
|
||||
vi.restoreAllMocks()
|
||||
vi.useRealTimers()
|
||||
})
|
||||
|
||||
const SID = 's1' as SessionId
|
||||
|
||||
@@ -66,8 +81,11 @@ describe('deriveStats', () => {
|
||||
expect(stats.turns).toBe(2)
|
||||
expect(stats.steps).toBe(3)
|
||||
// Window-scoped by design: the paged window is not an accounting source, so
|
||||
// the fold exposes no token fields at all (billing rides the projection).
|
||||
expect(Object.keys(stats).sort()).toEqual(['llmMs', 'steps', 'toolMs', 'turns'])
|
||||
// the fold exposes no billing fields (billing rides the projection);
|
||||
// decodeTokens is a throughput input, not a billed total.
|
||||
expect(Object.keys(stats).sort()).toEqual(
|
||||
['decodeMs', 'decodeTokens', 'llmMs', 'steps', 'toolMs', 'ttftMs', 'ttftSteps', 'turns'],
|
||||
)
|
||||
})
|
||||
|
||||
it('ignores tool results with no call time', () => {
|
||||
@@ -97,6 +115,23 @@ describe('deriveStats', () => {
|
||||
expect(stats.llmMs).toBe(2_500)
|
||||
expect(stats.toolMs).toBe(3_000)
|
||||
})
|
||||
|
||||
it('sums ttft per recorded step and decode throughput inputs per usage-carrying step', () => {
|
||||
const sampled: AssistantMessageNode = {
|
||||
...assistant(1, 1, { outputTokens: 40 }),
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_800, completedTime: 4_800 },
|
||||
}
|
||||
const ttftOnly: AssistantMessageNode = {
|
||||
...assistant(2, 1),
|
||||
timing: { stepStartTime: 5_000, firstTokenTime: 5_400, completedTime: 7_400 },
|
||||
}
|
||||
const stats = deriveStats([sampled, ttftOnly, assistant(3, 2)])
|
||||
expect(stats.ttftMs).toBe(1_200)
|
||||
expect(stats.ttftSteps).toBe(2)
|
||||
// The usage-less step contributes no decode share, keeping the ratio honest.
|
||||
expect(stats.decodeMs).toBe(3_000)
|
||||
expect(stats.decodeTokens).toBe(40)
|
||||
})
|
||||
})
|
||||
|
||||
describe('formatters', () => {
|
||||
@@ -125,7 +160,7 @@ describe('StatsLine', () => {
|
||||
source: { getSnapshot(): ConversationSnapshot; subscribe(fn: () => void): () => void },
|
||||
values: Record<string, unknown> = { tokenUsage: USAGE },
|
||||
): StatsLineProps {
|
||||
return { useSession: bindSnapshotSelector(source), useProjection: projections(values) }
|
||||
return { useSession: bindSnapshotSelector(source), useProjection: projections(values), t: tEn }
|
||||
}
|
||||
|
||||
it('renders the grouped stats row and hides a brand-new empty session', () => {
|
||||
@@ -142,47 +177,84 @@ describe('StatsLine', () => {
|
||||
expect(emptyView.container.textContent).toBe('')
|
||||
})
|
||||
|
||||
it('keeps durable token and context groups after the visible step window is empty', () => {
|
||||
it('reveals the full line in a delayed hover tooltip only while the row is clipped', () => {
|
||||
vi.useFakeTimers()
|
||||
// jsdom lays nothing out; fake a row narrower than its content.
|
||||
vi.spyOn(Element.prototype, 'scrollWidth', 'get').mockReturnValue(800)
|
||||
vi.spyOn(Element.prototype, 'clientWidth', 'get').mockReturnValue(400)
|
||||
const { source } = makeSource({ nodes: [assistant(1, 1)] })
|
||||
const view = render(<StatsLine {...props(source)} />)
|
||||
fireEvent.mouseEnter(view.container.firstElementChild!)
|
||||
act(() => { vi.advanceTimersByTime(499) })
|
||||
expect(view.container.querySelector('[role="tooltip"]')).toBeNull()
|
||||
act(() => { vi.advanceTimersByTime(1) })
|
||||
expect(view.container.querySelector('[role="tooltip"]')?.textContent)
|
||||
.toBe('1 turns · 1 steps | Cache hit 90% | Input 100 tok · Output 5 tok')
|
||||
})
|
||||
|
||||
it('suppresses the tooltip while the row fits without truncation', () => {
|
||||
vi.useFakeTimers()
|
||||
const { source } = makeSource({ nodes: [assistant(1, 1)] })
|
||||
const view = render(<StatsLine {...props(source)} />)
|
||||
fireEvent.mouseEnter(view.container.firstElementChild!)
|
||||
act(() => { vi.advanceTimersByTime(500) })
|
||||
expect(view.container.querySelector('[role="tooltip"]')).toBeNull()
|
||||
})
|
||||
|
||||
it('renders window latency and throughput beside the wall-time group', () => {
|
||||
const timed: AssistantMessageNode = {
|
||||
...assistant(1, 1, { outputTokens: 60 }),
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_800, completedTime: 4_800 },
|
||||
}
|
||||
const { source } = makeSource({ nodes: [timed] })
|
||||
const view = render(<StatsLine {...props(source)} />)
|
||||
expect(view.container.textContent).toContain('LLM 3.8s| TTFT avg 0.8s · 20 tok/s')
|
||||
})
|
||||
|
||||
it('takes every stats label from the active locale', () => {
|
||||
const timed: AssistantMessageNode = {
|
||||
...assistant(1, 1, { outputTokens: 60 }),
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_800, completedTime: 4_800 },
|
||||
}
|
||||
const { source } = makeSource({ nodes: [timed] })
|
||||
const view = render(<StatsLine {...props(source)} t={t} />)
|
||||
expect(view.container.textContent)
|
||||
.toBe('1 轮 · 1 步| LLM 3.8s| 首 token 平均 0.8s · 20 tok/s| 缓存命中 90%| 输入 100 tok · 输出 5 tok')
|
||||
})
|
||||
|
||||
it('renders without ResizeObserver support', () => {
|
||||
vi.unstubAllGlobals()
|
||||
const { source } = makeSource({ nodes: [assistant(1, 1)] })
|
||||
expect(() => render(<StatsLine {...props(source)} />)).not.toThrow()
|
||||
})
|
||||
|
||||
it('keeps durable token groups after the visible step window is empty', () => {
|
||||
const { source } = makeSource()
|
||||
const view = render(<StatsLine {...props(source, {
|
||||
tokenUsage: USAGE,
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
})} />)
|
||||
// Context occupancy lives on the composer's ContextMeter ring, not here.
|
||||
expect(view.container.textContent)
|
||||
.toBe('Context 25% of 128K| Cache hit 90%| Input 100 tok · Output 5 tok')
|
||||
.toBe('Cache hit 90%| Input 100 tok · Output 5 tok')
|
||||
})
|
||||
|
||||
it('renders context occupancy only when the projection knows a capacity', () => {
|
||||
const { source } = makeSource({ nodes: [assistant(1, 1)] })
|
||||
const withCapacity = render(<StatsLine {...props(source, {
|
||||
tokenUsage: USAGE,
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
})} />)
|
||||
expect(withCapacity.container.textContent).toContain('Context 25% of 128K')
|
||||
// Pressure without capacity has no denominator: the group drops out.
|
||||
const noCapacity = render(<StatsLine {...props(source, {
|
||||
tokenUsage: USAGE,
|
||||
contextPressure: { pressureTokens: 32_000 },
|
||||
})} />)
|
||||
expect(noCapacity.container.textContent).not.toContain('Context')
|
||||
// Capacity arrives before usage in the log; no provider sample means there
|
||||
// is no numerator yet, rather than a synthetic 0%.
|
||||
const noPressure = render(<StatsLine {...props(source, {
|
||||
tokenUsage: USAGE,
|
||||
contextPressure: { contextWindow: 128_000 },
|
||||
})} />)
|
||||
expect(noPressure.container.textContent).not.toContain('Context')
|
||||
})
|
||||
|
||||
it('clamps occupancy at 100% when pressure exceeds the recorded capacity', () => {
|
||||
// Capacity and pressure are independent last-wins fields, so a model switch
|
||||
// can pair a smaller new window with the previous route's larger prompt.
|
||||
const { source } = makeSource({ nodes: [assistant(1, 1)] })
|
||||
const view = render(<StatsLine {...props(source, {
|
||||
tokenUsage: USAGE,
|
||||
contextPressure: { pressureTokens: 300_000, contextWindow: 128_000 },
|
||||
})} />)
|
||||
expect(view.container.textContent).toContain('Context 100% of 128K')
|
||||
it('computes context occupancy only when both a numerator and capacity are known', () => {
|
||||
// The projected figure wins: it is the provider sample carried forward over
|
||||
// the surface's movement, so a compaction shows without waiting a request.
|
||||
expect(contextOccupancy({ pressureTokens: 32_000, projectedTokens: 6_000, contextWindow: 128_000 }))
|
||||
.toEqual({ percent: 5, usedTokens: 6_000, contextWindow: 128_000 })
|
||||
// A log whose projection predates the field still reads its bare sample.
|
||||
expect(contextOccupancy({ pressureTokens: 32_000, contextWindow: 128_000 }))
|
||||
.toEqual({ percent: 25, usedTokens: 32_000, contextWindow: 128_000 })
|
||||
// A numerator without capacity has no denominator; capacity without a
|
||||
// provider sample has no numerator yet, rather than a synthetic 0%.
|
||||
expect(contextOccupancy({ pressureTokens: 32_000 })).toBeNull()
|
||||
expect(contextOccupancy({ contextWindow: 128_000 })).toBeNull()
|
||||
expect(contextOccupancy(undefined)).toBeNull()
|
||||
// Capacity and the sample are independent last-wins fields, so a model
|
||||
// switch can pair a smaller new window with the previous route's prompt.
|
||||
expect(contextOccupancy({ pressureTokens: 300_000, contextWindow: 128_000 })?.percent).toBe(100)
|
||||
})
|
||||
|
||||
it('drops every token group when no projection is composed', () => {
|
||||
|
||||
@@ -541,6 +541,45 @@ describe('ChatView', () => {
|
||||
expect(view.getAllByText(/用时 19秒/)).toHaveLength(1)
|
||||
})
|
||||
|
||||
it('the settled footer appends first-step ttft and turn decode throughput', () => {
|
||||
const first: AssistantMessageNode = {
|
||||
kind: 'assistant', seq: 2, time: 2_000, turn: 1, step: 1, blocks: [{ kind: 'text', text: 'mid' }],
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 2_200, completedTime: 5_200 },
|
||||
usage: { outputTokens: 40 },
|
||||
}
|
||||
const second: AssistantMessageNode = {
|
||||
kind: 'assistant', seq: 16, time: 16_000, turn: 1, step: 2, blocks: [{ kind: 'text', text: 'final' }],
|
||||
timing: { stepStartTime: 10_000, firstTokenTime: 10_200, completedTime: 12_200 },
|
||||
usage: { outputTokens: 60 },
|
||||
}
|
||||
const h = makeHarness({
|
||||
nodes: [user(1, 'hi'), first, second],
|
||||
turnTimings: new Map([[1, { startTime: 1_000, endTime: 20_000 }]]),
|
||||
turnEnds: new Map([[1, 20]]),
|
||||
})
|
||||
const view = render(<h.ChatView {...h.props} />)
|
||||
// First-step ttft (1.2s) plus 100 tokens over 5s of decode.
|
||||
expect(view.getAllByText(/用时 19秒/)).toHaveLength(1)
|
||||
expect(view.getAllByText(/首 token 1\.2秒/)).toHaveLength(1)
|
||||
expect(view.getAllByText(/20 tok\/s/)).toHaveLength(1)
|
||||
})
|
||||
|
||||
it('withholds ttft and throughput while the turn is still running', () => {
|
||||
const settled: AssistantMessageNode = {
|
||||
kind: 'assistant', seq: 2, time: 2_000, turn: 1, step: 1, blocks: [{ kind: 'text', text: 'answer' }],
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_500, completedTime: 2_000 },
|
||||
usage: { outputTokens: 10 },
|
||||
}
|
||||
const h = makeHarness({
|
||||
nodes: [user(1, 'hi'), settled],
|
||||
turnTimings: new Map([[1, { startTime: 1_000 }]]),
|
||||
turnEnds: new Map(),
|
||||
running: true,
|
||||
})
|
||||
const view = render(<h.ChatView {...h.props} />)
|
||||
expect(view.queryByText(/首 token|tok\/s/)).toBeNull()
|
||||
})
|
||||
|
||||
it('user and assistant message containers scope the hover-revealed time chrome', () => {
|
||||
const h = makeHarness({
|
||||
nodes: [user(1, 'hi'), assistant(2, 'answer')],
|
||||
|
||||
158
packages/client/ui-conversation/tests/context-meter.spec.tsx
Normal file
158
packages/client/ui-conversation/tests/context-meter.spec.tsx
Normal file
@@ -0,0 +1,158 @@
|
||||
// @vitest-environment jsdom
|
||||
// ContextMeter (composer trailing control): occupancy ring gating, the
|
||||
// click-open breakdown panel, and its close gestures.
|
||||
|
||||
import { afterEach, describe, expect, it } from 'vitest'
|
||||
import { cleanup, fireEvent, render } from '@testing-library/react'
|
||||
import { makeTranslate } from '@deepseek-ai/dsh-client-test-runtime'
|
||||
import { en as commonEn, zh as commonZh } from '@deepseek-ai/dsh-client-locale/src/locales/index.ts'
|
||||
import { ContextMeter, type ContextMeterProps } from '../src/client/skeleton/ContextMeter.tsx'
|
||||
import css from '../src/client/skeleton/ContextMeter.module.css'
|
||||
import { en, zh } from '../src/client/locales.ts'
|
||||
|
||||
afterEach(cleanup)
|
||||
|
||||
// Mirrors the real lookup chain (conversation namespace, then common).
|
||||
const t = makeTranslate(zh, commonZh) as ContextMeterProps['t']
|
||||
const tEn = makeTranslate(en, commonEn) as ContextMeterProps['t']
|
||||
|
||||
const BREAKDOWN = { systemTokens: 120, toolsTokens: 21_500, messageTokens: 477_000 }
|
||||
|
||||
const segmentClass = css.segment
|
||||
if (segmentClass === undefined) throw new Error('segment class missing from ContextMeter.module.css')
|
||||
|
||||
/** Stub the projection seat: a key-addressed table of whole values. */
|
||||
function projections(values: Record<string, unknown>): ContextMeterProps['useProjection'] {
|
||||
return (key: string) => values[key]
|
||||
}
|
||||
|
||||
function meter(values: Record<string, unknown>, translate: ContextMeterProps['t'] = t) {
|
||||
return render(<ContextMeter useProjection={projections(values)} t={translate} />)
|
||||
}
|
||||
|
||||
describe('ContextMeter', () => {
|
||||
it('renders nothing until both pressure and capacity are known', () => {
|
||||
expect(meter({}).container.textContent).toBe('')
|
||||
expect(meter({ contextPressure: { pressureTokens: 32_000 } }).container.textContent).toBe('')
|
||||
expect(meter({ contextPressure: { contextWindow: 128_000 } }).container.textContent).toBe('')
|
||||
})
|
||||
|
||||
it('shows the occupancy ring and opens the breakdown panel on click', () => {
|
||||
const view = meter({
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
})
|
||||
const trigger = view.getByRole('button', { name: '上下文已用 25%' })
|
||||
expect(view.container.querySelector('[role="dialog"]')).toBeNull()
|
||||
fireEvent.click(trigger)
|
||||
const panel = view.container.querySelector('[role="dialog"]')!
|
||||
expect(panel.textContent).toContain('~32K / 128K')
|
||||
expect(panel.textContent).toContain('25%')
|
||||
expect(panel.textContent).toContain('上下文已用')
|
||||
expect(panel.textContent).toContain('系统提示词~120')
|
||||
expect(panel.textContent).toContain('工具~21.5K')
|
||||
expect(panel.textContent).toContain('对话消息~477K')
|
||||
// The occupancy bar splits into one colored segment per composition row.
|
||||
expect(panel.getElementsByClassName(segmentClass)).toHaveLength(3)
|
||||
// Clicking the trigger again toggles the panel shut.
|
||||
fireEvent.click(trigger)
|
||||
expect(view.container.querySelector('[role="dialog"]')).toBeNull()
|
||||
})
|
||||
|
||||
it('lets each locale own the headline word order around the reading', () => {
|
||||
const values = {
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
}
|
||||
const zhView = meter(values)
|
||||
fireEvent.click(zhView.getByRole('button', { name: '上下文已用 25%' }))
|
||||
// The reading follows the label in Chinese and leads it in English; both
|
||||
// headers read as one sentence rather than a concatenated fragment.
|
||||
expect(zhView.container.querySelector('[role="dialog"]')!.textContent)
|
||||
.toMatch(/^上下文已用25%/)
|
||||
const enView = meter(values, tEn)
|
||||
fireEvent.click(enView.getByRole('button', { name: '25% of context used' }))
|
||||
expect(enView.container.querySelector('[role="dialog"]')!.textContent)
|
||||
.toMatch(/^25%of context used/)
|
||||
})
|
||||
|
||||
it('draws no bar segment at zero occupancy', () => {
|
||||
const view = meter({
|
||||
contextPressure: { pressureTokens: 0, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
})
|
||||
fireEvent.click(view.getByRole('button', { name: '上下文已用 0%' }))
|
||||
const panel = view.container.querySelector('[role="dialog"]')!
|
||||
// `.segment` carries a min-width, so a zero-width part would still paint a
|
||||
// filled sliver over an empty context.
|
||||
expect(panel.getElementsByClassName(segmentClass)).toHaveLength(0)
|
||||
expect(panel.textContent).toContain('~0 / 128K')
|
||||
})
|
||||
|
||||
it('reads the ring from the projected figure so a compaction shows at once', () => {
|
||||
// Same provider sample, a surface a compaction just shrank: the ring must
|
||||
// follow the projection rather than the sample it is anchored to.
|
||||
const view = meter({
|
||||
contextPressure: { pressureTokens: 32_000, projectedTokens: 3_000, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
})
|
||||
const trigger = view.getByRole('button', { name: '上下文已用 2%' })
|
||||
fireEvent.click(trigger)
|
||||
expect(view.container.querySelector('[role="dialog"]')!.textContent).toContain('~3K / 128K')
|
||||
})
|
||||
|
||||
it('omits the composition rows while the contextBreakdown projection is absent', () => {
|
||||
const view = meter({ contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 } })
|
||||
fireEvent.click(view.getByRole('button', { name: '上下文已用 25%' }))
|
||||
const panel = view.container.querySelector('[role="dialog"]')!
|
||||
expect(panel.textContent).toContain('~32K / 128K')
|
||||
expect(panel.textContent).not.toContain('系统提示词')
|
||||
expect(panel.textContent).not.toContain('对话消息')
|
||||
// Without composition shares, the bar falls back to one plain segment.
|
||||
expect(panel.getElementsByClassName(segmentClass)).toHaveLength(1)
|
||||
})
|
||||
|
||||
it('closes when capacity disappears and stays closed when it returns', () => {
|
||||
let values: Record<string, unknown> = {
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
}
|
||||
const view = render(<ContextMeter useProjection={(key: string) => values[key]} t={t} />)
|
||||
fireEvent.click(view.getByRole('button', { name: '上下文已用 25%' }))
|
||||
expect(view.container.querySelector('[role="dialog"]')).not.toBeNull()
|
||||
|
||||
values = { contextPressure: { pressureTokens: 32_000 }, contextBreakdown: BREAKDOWN }
|
||||
view.rerender(<ContextMeter useProjection={(key: string) => values[key]} t={t} />)
|
||||
expect(view.container.textContent).toBe('')
|
||||
|
||||
values = {
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
}
|
||||
view.rerender(<ContextMeter useProjection={(key: string) => values[key]} t={t} />)
|
||||
expect(view.getByRole('button', { name: '上下文已用 25%' }).getAttribute('aria-expanded')).toBe('false')
|
||||
expect(view.container.querySelector('[role="dialog"]')).toBeNull()
|
||||
})
|
||||
|
||||
it('closes on outside pointerdown and Escape — but not inside clicks', () => {
|
||||
const view = meter({
|
||||
contextPressure: { pressureTokens: 32_000, contextWindow: 128_000 },
|
||||
contextBreakdown: BREAKDOWN,
|
||||
})
|
||||
const trigger = view.getByRole('button', { name: '上下文已用 25%' })
|
||||
const openPanel = () => {
|
||||
fireEvent.click(trigger)
|
||||
return view.container.querySelector('[role="dialog"]')!
|
||||
}
|
||||
// A pointerdown inside the panel keeps it open; outside closes it.
|
||||
const again = openPanel()
|
||||
fireEvent.pointerDown(again)
|
||||
expect(view.container.querySelector('[role="dialog"]')).not.toBeNull()
|
||||
fireEvent.pointerDown(document.body)
|
||||
expect(view.container.querySelector('[role="dialog"]')).toBeNull()
|
||||
// Escape.
|
||||
openPanel()
|
||||
fireEvent.keyDown(document, { key: 'Escape' })
|
||||
expect(view.container.querySelector('[role="dialog"]')).toBeNull()
|
||||
})
|
||||
})
|
||||
@@ -1,6 +1,6 @@
|
||||
// @vitest-environment jsdom
|
||||
|
||||
import { afterEach, describe, expect, it, vi } from 'vitest'
|
||||
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
|
||||
import { cleanup, render } from '@testing-library/react'
|
||||
import { bindSnapshotSelector } from '@deepseek-ai/dsh-client-web-react'
|
||||
import { createSnapshotStore } from '@deepseek-ai/dsh-client-runtime/client'
|
||||
@@ -18,7 +18,18 @@ import { zh } from '../src/client/locales.ts'
|
||||
// Mirrors the real lookup chain (conversation namespace, then common).
|
||||
const t: AssistantMarkdownProps['t'] = makeTranslate(zh, commonZh)
|
||||
|
||||
afterEach(cleanup)
|
||||
/** jsdom has no ResizeObserver; StatsLine watches its row for ellipsis truncation through one. */
|
||||
class ResizeObserverStub {
|
||||
observe(): void {}
|
||||
unobserve(): void {}
|
||||
disconnect(): void {}
|
||||
}
|
||||
|
||||
beforeEach(() => { vi.stubGlobal('ResizeObserver', ResizeObserverStub) })
|
||||
afterEach(() => {
|
||||
cleanup()
|
||||
vi.unstubAllGlobals()
|
||||
})
|
||||
|
||||
const SID = 's1' as SessionId
|
||||
|
||||
@@ -56,11 +67,12 @@ describe('render branch tails', () => {
|
||||
const source = { getSnapshot: () => snap, subscribe: () => () => {} }
|
||||
const view = render(
|
||||
<StatsLine
|
||||
t={t}
|
||||
useSession={bindSnapshotSelector(source) as unknown as UseSession<ConversationSnapshot>}
|
||||
useProjection={() => undefined}
|
||||
/>,
|
||||
)
|
||||
expect(view.container.textContent).toBe('2 turns · 3 steps')
|
||||
expect(view.container.textContent).toBe('2 轮 · 3 步')
|
||||
})
|
||||
|
||||
it('AssistantMarkdown reasoning as the streaming tail renders the running ring', () => {
|
||||
|
||||
154
packages/client/ui-conversation/tests/turn-metrics.spec.ts
Normal file
154
packages/client/ui-conversation/tests/turn-metrics.spec.ts
Normal file
@@ -0,0 +1,154 @@
|
||||
// Per-turn latency/throughput fold and the footer figure formatters.
|
||||
|
||||
import { describe, expect, it } from 'vitest'
|
||||
import type { AssistantMessageNode, ConversationNode, UserMessageNode } from '@deepseek-ai/dsh-client-runtime/client'
|
||||
import { assistantStepReading, deriveTurnMetrics } from '../src/client/chat/turn-metrics.ts'
|
||||
import { formatLatencySeconds, formatTokensPerSecond } from '../src/client/chat/message-chrome.ts'
|
||||
|
||||
interface StepSpec {
|
||||
seq: number
|
||||
turn: number
|
||||
step: number
|
||||
timing?: AssistantMessageNode['timing']
|
||||
usage?: unknown
|
||||
}
|
||||
|
||||
const assistant = ({ seq, turn, step, timing, usage }: StepSpec): AssistantMessageNode => ({
|
||||
kind: 'assistant', seq, time: seq * 1_000, turn, step, blocks: [{ kind: 'text', text: `t${seq}` }],
|
||||
...(timing === undefined ? {} : { timing }),
|
||||
...(usage === undefined ? {} : { usage }),
|
||||
})
|
||||
|
||||
const user = (seq: number): UserMessageNode => ({
|
||||
kind: 'user', seq, time: seq * 1_000, content: [{ type: 'text', text: 'hi' }] as never, source: null,
|
||||
})
|
||||
|
||||
describe('assistantStepReading', () => {
|
||||
it('derives ttft, decode time, and output tokens from a fully recorded step', () => {
|
||||
const reading = assistantStepReading(assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_800, completedTime: 6_800 },
|
||||
usage: { outputTokens: 200 },
|
||||
}))
|
||||
expect(reading).toEqual({ ttftMs: 800, decodeMs: 5_000, outputTokens: 200 })
|
||||
})
|
||||
|
||||
it('returns nulls when timing is absent', () => {
|
||||
const reading = assistantStepReading(assistant({ seq: 2, turn: 1, step: 1, usage: { outputTokens: 5 } }))
|
||||
expect(reading).toEqual({ ttftMs: null, decodeMs: null, outputTokens: 5 })
|
||||
})
|
||||
|
||||
it('needs both boundaries for ttft and clamps negative spans to zero', () => {
|
||||
expect(assistantStepReading(assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: null, firstTokenTime: 1_800, completedTime: 6_800 },
|
||||
}))).toEqual({ ttftMs: null, decodeMs: 5_000, outputTokens: null })
|
||||
expect(assistantStepReading(assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: null, completedTime: 6_800 },
|
||||
}))).toEqual({ ttftMs: null, decodeMs: null, outputTokens: null })
|
||||
expect(assistantStepReading(assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: 2_000, firstTokenTime: 1_500, completedTime: 1_200 },
|
||||
}))).toEqual({ ttftMs: 0, decodeMs: 0, outputTokens: null })
|
||||
})
|
||||
|
||||
it('rejects non-object, missing, and non-finite usage token counts', () => {
|
||||
const timing = { stepStartTime: 1_000, firstTokenTime: 1_500, completedTime: 2_000 }
|
||||
expect(assistantStepReading(assistant({ seq: 2, turn: 1, step: 1, timing, usage: 'weird' })).outputTokens).toBeNull()
|
||||
expect(assistantStepReading(assistant({ seq: 2, turn: 1, step: 1, timing, usage: {} })).outputTokens).toBeNull()
|
||||
const nan = assistant({ seq: 2, turn: 1, step: 1, timing, usage: { outputTokens: Number.NaN } })
|
||||
expect(assistantStepReading(nan).outputTokens).toBeNull()
|
||||
expect(assistantStepReading(assistant({ seq: 2, turn: 1, step: 1, timing, usage: { outputTokens: -3 } })).outputTokens).toBeNull()
|
||||
})
|
||||
})
|
||||
|
||||
describe('deriveTurnMetrics', () => {
|
||||
it('takes ttft from the lowest step and throughput over all sampled steps', () => {
|
||||
const nodes: ConversationNode[] = [
|
||||
user(1),
|
||||
// Out of step order on purpose: the lowest step owns the ttft slot.
|
||||
assistant({
|
||||
seq: 4, turn: 1, step: 2,
|
||||
timing: { stepStartTime: 10_000, firstTokenTime: 10_200, completedTime: 12_200 },
|
||||
usage: { outputTokens: 60 },
|
||||
}),
|
||||
assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 2_200, completedTime: 5_200 },
|
||||
usage: { outputTokens: 40 },
|
||||
}),
|
||||
]
|
||||
// 100 tokens over 5s of decode.
|
||||
expect(deriveTurnMetrics(nodes).get(1)).toEqual({ ttftMs: 1_200, tokensPerSecond: 20 })
|
||||
})
|
||||
|
||||
it('emits ttft without throughput when no step carries usage', () => {
|
||||
const nodes = [assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_900, completedTime: 3_000 },
|
||||
})]
|
||||
expect(deriveTurnMetrics(nodes).get(1)).toEqual({ ttftMs: 900 })
|
||||
})
|
||||
|
||||
it('emits throughput without ttft when only a later step is recorded', () => {
|
||||
const nodes = [
|
||||
assistant({ seq: 2, turn: 1, step: 1 }),
|
||||
assistant({
|
||||
seq: 4, turn: 1, step: 2,
|
||||
timing: { stepStartTime: 10_000, firstTokenTime: 10_500, completedTime: 12_500 },
|
||||
usage: { outputTokens: 30 },
|
||||
}),
|
||||
]
|
||||
expect(deriveTurnMetrics(nodes).get(1)).toEqual({ tokensPerSecond: 15 })
|
||||
})
|
||||
|
||||
it('omits turns with no readings and zero-decode throughput', () => {
|
||||
const nodes = [
|
||||
assistant({ seq: 2, turn: 1, step: 1 }),
|
||||
assistant({
|
||||
seq: 4, turn: 2, step: 1,
|
||||
timing: { stepStartTime: null, firstTokenTime: 5_000, completedTime: 5_000 },
|
||||
usage: { outputTokens: 10 },
|
||||
}),
|
||||
]
|
||||
expect(deriveTurnMetrics(nodes).size).toBe(0)
|
||||
})
|
||||
|
||||
it('keeps turns independent and ignores non-assistant nodes', () => {
|
||||
const nodes: ConversationNode[] = [
|
||||
user(1),
|
||||
assistant({
|
||||
seq: 2, turn: 1, step: 1,
|
||||
timing: { stepStartTime: 1_000, firstTokenTime: 1_400, completedTime: 2_400 },
|
||||
usage: { outputTokens: 10 },
|
||||
}),
|
||||
user(3),
|
||||
assistant({
|
||||
seq: 4, turn: 2, step: 1,
|
||||
timing: { stepStartTime: 4_000, firstTokenTime: 4_100, completedTime: 6_100 },
|
||||
usage: { outputTokens: 100 },
|
||||
}),
|
||||
]
|
||||
const metrics = deriveTurnMetrics(nodes)
|
||||
expect(metrics.get(1)).toEqual({ ttftMs: 400, tokensPerSecond: 10 })
|
||||
expect(metrics.get(2)).toEqual({ ttftMs: 100, tokensPerSecond: 50 })
|
||||
})
|
||||
})
|
||||
|
||||
describe('footer figure formatters', () => {
|
||||
it('formats latency with one decimal under ten seconds and whole seconds beyond', () => {
|
||||
expect(formatLatencySeconds(840)).toBe('0.8')
|
||||
expect(formatLatencySeconds(1_000)).toBe('1')
|
||||
expect(formatLatencySeconds(9_949)).toBe('9.9')
|
||||
expect(formatLatencySeconds(12_400)).toBe('12')
|
||||
expect(formatLatencySeconds(-5)).toBe('0')
|
||||
})
|
||||
|
||||
it('formats throughput with whole tokens from ten up and one decimal below', () => {
|
||||
expect(formatTokensPerSecond(34.4)).toBe('34')
|
||||
expect(formatTokensPerSecond(9.96)).toBe('10')
|
||||
expect(formatTokensPerSecond(3.14)).toBe('3.1')
|
||||
expect(formatTokensPerSecond(-1)).toBe('0')
|
||||
})
|
||||
})
|
||||
@@ -25,9 +25,11 @@
|
||||
],
|
||||
"license": "BSD-3-Clause",
|
||||
"peerDependencies": {
|
||||
"@deepseek-ai/dsh-compact": "^0.0.1",
|
||||
"@deepseek-ai/dsh-invariants": "^0.0.1",
|
||||
"@deepseek-ai/dsh-llm": "^0.0.1",
|
||||
"@deepseek-ai/dsh-session": "^0.0.1",
|
||||
"@deepseek-ai/dsh-token-meter": "^0.0.1",
|
||||
"cordis": "^4.0.0-rc.7"
|
||||
},
|
||||
"dependencies": {
|
||||
@@ -36,9 +38,11 @@
|
||||
"devDependencies": {
|
||||
"@cordisjs/plugin-include": "workspace:^",
|
||||
"@cordisjs/plugin-loader": "workspace:^",
|
||||
"@deepseek-ai/dsh-compact": "workspace:^",
|
||||
"@deepseek-ai/dsh-invariants": "workspace:^",
|
||||
"@deepseek-ai/dsh-llm": "workspace:^",
|
||||
"@deepseek-ai/dsh-session": "workspace:^",
|
||||
"@deepseek-ai/dsh-token-meter": "workspace:^",
|
||||
"cordis": "^4.0.0-rc.7"
|
||||
}
|
||||
}
|
||||
|
||||
@@ -9,6 +9,10 @@ import z from 'schemastery'
|
||||
import { freezeMessage } from '@deepseek-ai/dsh-llm'
|
||||
import type { ContentBlock } from '@deepseek-ai/dsh-llm'
|
||||
import type { Session, SessionEvent, ToolResultMessage } from '@deepseek-ai/dsh-session'
|
||||
// Type-only: the `compact/*` SessionEventMap merges (the shadow-price event).
|
||||
import type {} from '@deepseek-ai/dsh-compact'
|
||||
// Type-only: the `ctx.tokenMeter` Context merge for the declared injection.
|
||||
import type {} from '@deepseek-ai/dsh-token-meter'
|
||||
import { codePointLength, DEFAULTS, PRUNE_MARKER, resolveConfig } from './config.ts'
|
||||
import type {
|
||||
PrunedEntry,
|
||||
@@ -38,6 +42,10 @@ interface SnapshotCandidate {
|
||||
|
||||
/** Deterministic head/middle/tail pruning for current tool-result surface nodes. */
|
||||
export class ToolResultPruneService extends Service {
|
||||
// The token meter prices each shadowed node for its logged shadow-price
|
||||
// event, so pruning genuinely requires the pricing capability.
|
||||
static inject = ['tokenMeter']
|
||||
|
||||
static Config: z<ToolResultPruneConfig> = z.object({
|
||||
thresholdChars: z.number().step(1).min(1).default(DEFAULTS.thresholdChars),
|
||||
headChars: z.number().step(1).min(0).default(DEFAULTS.headChars),
|
||||
@@ -116,7 +124,10 @@ export class ToolResultPruneService extends Service {
|
||||
/**
|
||||
* Prune every over-budget tool result from one stable current-surface snapshot.
|
||||
* Each replacement preserves the complete event data except for `content`,
|
||||
* and points at the shadowed node for durable provenance and replay.
|
||||
* points at the shadowed node for durable provenance and replay, and is
|
||||
* immediately preceded by a `compact/prune` shadow-price event pricing the
|
||||
* shadowed node through the injected token meter, so pure consumers can
|
||||
* subtract it without per-node state.
|
||||
* @param session - session whose current surface is rewritten.
|
||||
* @returns landed replacements and aggregate Unicode-code-point savings.
|
||||
* @throws when the session rejects a replacement; replacements committed
|
||||
@@ -145,6 +156,14 @@ export class ToolResultPruneService extends Service {
|
||||
content,
|
||||
}] as [typeof result],
|
||||
})
|
||||
// Shadow-price protocol: the metering event and its replacement are
|
||||
// appended synchronously adjacent, so pure consumers subtract the
|
||||
// shadowed node's heuristic price without retaining per-node state.
|
||||
session.append('compact/prune', {
|
||||
shadowedRange: { start: seq, end: seq },
|
||||
shadowedSeqs: [seq],
|
||||
shadowedTokenCount: this.ctx.tokenMeter.estimateMessage(event.data.message),
|
||||
})
|
||||
const replacement = session.append('tool/result', {
|
||||
...event.data,
|
||||
message,
|
||||
|
||||
@@ -6,6 +6,7 @@ import { afterEach, describe, expect, it } from 'vitest'
|
||||
import { Context } from 'cordis'
|
||||
import Loader from '@cordisjs/plugin-loader'
|
||||
import Include from '@cordisjs/plugin-include'
|
||||
import TokenMeterService from '@deepseek-ai/dsh-token-meter'
|
||||
import ToolResultPruneService from '@deepseek-ai/dsh-compact-tool-result-prune'
|
||||
|
||||
let root: string | undefined
|
||||
@@ -23,6 +24,7 @@ describe('compact-tool-result-prune real Loader composition', () => {
|
||||
root = await mkdtemp(join(tmpdir(), 'dsh-compact-tool-result-prune-loader-'))
|
||||
const configPath = join(root, 'cordis.yml')
|
||||
await writeFile(configPath, [
|
||||
"- name: '@deepseek-ai/dsh-token-meter'",
|
||||
"- name: '@deepseek-ai/dsh-compact-tool-result-prune'",
|
||||
' config:',
|
||||
' thresholdChars: 100',
|
||||
@@ -38,10 +40,9 @@ describe('compact-tool-result-prune real Loader composition', () => {
|
||||
context.loader.internal = {
|
||||
version: 'v2',
|
||||
async import(specifier: string) {
|
||||
if (specifier !== '@deepseek-ai/dsh-compact-tool-result-prune') {
|
||||
throw new Error(`unexpected Loader import: ${specifier}`)
|
||||
}
|
||||
return ToolResultPruneService
|
||||
if (specifier === '@deepseek-ai/dsh-token-meter') return TokenMeterService
|
||||
if (specifier === '@deepseek-ai/dsh-compact-tool-result-prune') return ToolResultPruneService
|
||||
throw new Error(`unexpected Loader import: ${specifier}`)
|
||||
},
|
||||
} as unknown as NonNullable<typeof context.loader.internal>
|
||||
await context.loader.create({
|
||||
@@ -60,6 +61,9 @@ describe('compact-tool-result-prune real Loader composition', () => {
|
||||
|
||||
it('rejects stale config after plugin schema normalization', async () => {
|
||||
context = new Context()
|
||||
// Satisfy the declared injection first: config normalization runs in the
|
||||
// service constructor, which a pending fiber never reaches.
|
||||
await context.plugin(TokenMeterService)
|
||||
await expect(context.plugin(ToolResultPruneService, {
|
||||
maxChars: 100,
|
||||
} as never)).rejects.toThrow(/unknown key "maxChars"/)
|
||||
|
||||
@@ -9,6 +9,7 @@ import SessionStore, {
|
||||
import type { SurfaceEvent } from '@deepseek-ai/dsh-session'
|
||||
import * as SessionInvariant from '@deepseek-ai/dsh-session/invariant'
|
||||
import InvariantService from '@deepseek-ai/dsh-invariants'
|
||||
import TokenMeterService from '@deepseek-ai/dsh-token-meter'
|
||||
import ToolResultPruneService, {
|
||||
codePointLength,
|
||||
DEFAULTS,
|
||||
@@ -25,9 +26,16 @@ const SMALL: ToolResultPruneConfig = {
|
||||
}
|
||||
|
||||
function service(config: ToolResultPruneConfig = SMALL): ToolResultPruneService {
|
||||
return new ToolResultPruneService(new Context(), config)
|
||||
const ctx = new Context()
|
||||
// Service constructors self-register, so `ctx.tokenMeter` resolves for the
|
||||
// shadow-price pricing without a full plugin boot.
|
||||
void new TokenMeterService(ctx)
|
||||
return new ToolResultPruneService(ctx, config)
|
||||
}
|
||||
|
||||
/** Pricing oracle mirroring the service's estimator for expectations. */
|
||||
const METER = new TokenMeterService(new Context())
|
||||
|
||||
function appendToolStep(
|
||||
session: Session,
|
||||
turn: number,
|
||||
@@ -203,6 +211,18 @@ describe('ToolResultPruneService session transaction', () => {
|
||||
sourceEventSeqs: [originalSeq],
|
||||
})
|
||||
expect(session.surface.nodes).not.toContain(originalSeq)
|
||||
|
||||
// Shadow-price protocol: the metering event sits directly before the
|
||||
// replacement and prices the shadowed node with the shared estimator.
|
||||
if (original.type !== 'tool/result') throw new Error('original is not a tool/result')
|
||||
expect(session.events[entry.replacementSeq - 1]).toMatchObject({
|
||||
type: 'compact/prune',
|
||||
data: {
|
||||
shadowedRange: { start: originalSeq, end: originalSeq },
|
||||
shadowedSeqs: [originalSeq],
|
||||
shadowedTokenCount: METER.estimateMessage(original.data.message),
|
||||
},
|
||||
})
|
||||
})
|
||||
|
||||
it('prunes multiple results, skips short ones, and converges in one pass', () => {
|
||||
@@ -240,6 +260,7 @@ describe('ToolResultPruneService session transaction', () => {
|
||||
await ctx.plugin(SessionStore)
|
||||
await ctx.plugin(InvariantService)
|
||||
await ctx.plugin(SessionInvariant)
|
||||
await ctx.plugin(TokenMeterService)
|
||||
const prune = new ToolResultPruneService(ctx, SMALL)
|
||||
const session = ctx.sessions.create(SessionId('invariants'))
|
||||
appendToolStep(session, 1, 'a', [{ type: 'text', text: 'A'.repeat(100) }])
|
||||
|
||||
@@ -10,7 +10,9 @@
|
||||
{ "path": "../../../vendor/cordis" },
|
||||
{ "path": "../../../vendor/schemastery" },
|
||||
{ "path": "../../llm/llm" },
|
||||
{ "path": "../../llm/token-meter" },
|
||||
{ "path": "../../core/session" },
|
||||
{ "path": "../compact" },
|
||||
{ "path": "../../support/invariants" }
|
||||
]
|
||||
}
|
||||
|
||||
@@ -20,8 +20,11 @@ declare module '@deepseek-ai/dsh-session' {
|
||||
/**
|
||||
* Provenance record of a completed summarization — log-only, no surfaceOp.
|
||||
* The summary content is in `data.summary`; the actual surface replacement
|
||||
* is performed by a subsequent `user/message` event that shadows the
|
||||
* compacted range.
|
||||
* is performed by the immediately following `user/message` event that
|
||||
* shadows the compacted range. That adjacency is contractual — the
|
||||
* shadowed pricing fields are the replacement's shadow price, so a
|
||||
* consumer may pair a replacement with the metering event directly
|
||||
* before it (`compact/prune` documents the shared protocol).
|
||||
*/
|
||||
'compact/summary': {
|
||||
summary: ContentBlock[]
|
||||
@@ -49,6 +52,23 @@ declare module '@deepseek-ai/dsh-session' {
|
||||
* matches `compact/start`; `error` records an unsuccessful attempt.
|
||||
*/
|
||||
'compact/end': { turn: number | null; error?: string }
|
||||
/**
|
||||
* Shadow price of one model-free prune replacement — log-only, no
|
||||
* surfaceOp. The shared shadow-price protocol: a surface `replace` event
|
||||
* is priced by the metering event immediately before it (`compact/summary`
|
||||
* for a summarizing compaction, this event for a prune), which states the
|
||||
* heuristic token price of the exact replaced range so a pure consumer
|
||||
* can subtract it without retaining per-node prices. The replacement MUST
|
||||
* be appended synchronously right after this event.
|
||||
*/
|
||||
'compact/prune': {
|
||||
/** The replaced range's first and last surface-node seqs (a surface-position span, like {@link CompactionResult.shadowedRange}). */
|
||||
shadowedRange: { start: number; end: number }
|
||||
/** The seqs of all shadowed surface nodes, in surface order. */
|
||||
shadowedSeqs: number[]
|
||||
/** Heuristic price of the shadowed content under the token-meter's fixed estimator. */
|
||||
shadowedTokenCount: number
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -1044,7 +1044,7 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
|
||||
},
|
||||
{
|
||||
signature: 'estimateMessage(message: Message): number',
|
||||
jsDoc: '/**\n * Heuristically price one model-visible message.\n * @param message - message to price without mutation.\n * @returns content and role-framing tokens under the fixed service heuristic.\n */',
|
||||
jsDoc: '/**\n * Heuristically price one model-visible message (instance face of the pure\n * `estimateMessage` export from `estimate.ts`).\n * @param message - message to price without mutation.\n * @returns content and role-framing tokens under the fixed service heuristic.\n */',
|
||||
},
|
||||
],
|
||||
},
|
||||
@@ -1062,7 +1062,7 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
|
||||
},
|
||||
{
|
||||
signature: 'pruneSession(session: Session): PruneResult',
|
||||
jsDoc: '/**\n * Prune every over-budget tool result from one stable current-surface snapshot.\n * Each replacement preserves the complete event data except for `content`,\n * and points at the shadowed node for durable provenance and replay.\n * @param session - session whose current surface is rewritten.\n * @returns landed replacements and aggregate Unicode-code-point savings.\n * @throws when the session rejects a replacement; replacements committed\n * earlier in the pass remain durable.\n */',
|
||||
jsDoc: '/**\n * Prune every over-budget tool result from one stable current-surface snapshot.\n * Each replacement preserves the complete event data except for `content`,\n * points at the shadowed node for durable provenance and replay, and is\n * immediately preceded by a `compact/prune` shadow-price event pricing the\n * shadowed node through the injected token meter, so pure consumers can\n * subtract it without per-node state.\n * @param session - session whose current surface is rewritten.\n * @returns landed replacements and aggregate Unicode-code-point savings.\n * @throws when the session rejects a replacement; replacements committed\n * earlier in the pass remain durable.\n */',
|
||||
},
|
||||
],
|
||||
},
|
||||
|
||||
@@ -15,7 +15,7 @@ import type { Message } from '@deepseek-ai/dsh-llm'
|
||||
import { SESSION_FORMAT_VERSION, SessionId } from './types.ts'
|
||||
import type { CreateSessionOptions, EpochHeader, PrepareSessionOptions, RequestContext, SessionEvent, SessionEventMap, SessionEventType, SessionHeader, SurfaceIntent, SurfaceEventType } from './types.ts'
|
||||
import { snapshotJsonValue } from './json.ts'
|
||||
import { SurfaceManager } from './surface.ts'
|
||||
import { deriveEventMessage, SurfaceManager } from './surface.ts'
|
||||
import type { SessionSurface } from './surface.ts'
|
||||
import { foldRequestHeader } from './request-header.ts'
|
||||
|
||||
@@ -29,7 +29,7 @@ export { interruptedTurnClosers, lastActivityTime, TOOL_NOT_STARTED, TOOL_OUTCOM
|
||||
export { decodeStorageRecord, packChunkRuns } from './chunk-rows.ts'
|
||||
export type { ChunkRow, StorageRecord } from './chunk-rows.ts'
|
||||
export type { SessionSurface, SurfaceFoldReplacement, SurfaceFoldResult } from './surface.ts'
|
||||
export { foldSurface, isAppendSurfaceEvent, isReplacementSurfaceEvent, isSurfaceEvent, isSurfaceEligibleType } from './surface.ts'
|
||||
export { deriveEventMessage, foldSurface, isAppendSurfaceEvent, isReplacementSurfaceEvent, isSurfaceEvent, isSurfaceEligibleType } from './surface.ts'
|
||||
export { canonicalHeader, foldRequestHeader, headerEquals } from './request-header.ts'
|
||||
|
||||
/**
|
||||
@@ -755,50 +755,13 @@ export class Session {
|
||||
}
|
||||
|
||||
/**
|
||||
* Project a single event into the LLM message it derives to, or null when
|
||||
* it produces none — a non-surface event (chunk, boundary, log-only record)
|
||||
* or an empty-content assistant/message (which exists only to host usage).
|
||||
* The per-node pure function {@link deriveMessages} folds over the surface;
|
||||
* an external reconstructor (or the dev invariant) folds the same function
|
||||
* over a log prefix's surface to rebuild the exact messages any request was
|
||||
* built from (the reconstructability Agent Note). The returned message is
|
||||
* the already frozen message nested in the event wrapper and shared by
|
||||
* delivery, durable history, and model requests.
|
||||
* Instance face of the pure per-node `deriveEventMessage` export from
|
||||
* `surface.ts`.
|
||||
* @param event - the event to project.
|
||||
* @returns the derived message, or null when the event produces none.
|
||||
*/
|
||||
deriveEventMessage(event: SessionEvent): Message | null {
|
||||
// Intentionally non-exhaustive: only message-producing events derive
|
||||
// history; turn/step boundaries, chunks, usage, and errors are
|
||||
// trace/replay data.
|
||||
|
||||
switch (event.type) {
|
||||
// Ordinary prompts and injected context project in user role: the
|
||||
// event's model-facing content stays verbatim. Do NOT
|
||||
// re-add per-type framing (e.g. `<context>`) here: framing is
|
||||
// caller-owned — a producer bakes it into `content`, as workspace-context
|
||||
// does with `<system-reminder>` — or, if reintroduced, must be driven by
|
||||
// the event `meta` map and a dedicated renderer, keeping this projection a
|
||||
// verbatim pass-through. See the deferred design note in
|
||||
// ../../../../.agents/notes/implemented/simplification/2026-07-20-unwrap-injected-content-envelopes.md
|
||||
case 'user/message': {
|
||||
return event.data
|
||||
}
|
||||
case 'assistant/message': {
|
||||
// Skip an empty-content assistant/message: it exists only to host a
|
||||
// max-tokens step's usage and must not inject a content-less assistant
|
||||
// turn into the provider transcript.
|
||||
if (event.data.message.content.length === 0) return null
|
||||
return event.data.message
|
||||
}
|
||||
case 'tool/result': {
|
||||
return event.data.message
|
||||
}
|
||||
default:
|
||||
// A non-surface event (boundary, chunk, log-only record) projects to
|
||||
// no message. Merge-extensible union: no assertNever here.
|
||||
return null
|
||||
}
|
||||
return deriveEventMessage(event)
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -8,6 +8,7 @@
|
||||
* @module @deepseek-ai/dsh-session/surface
|
||||
*/
|
||||
|
||||
import type { Message } from '@deepseek-ai/dsh-llm'
|
||||
import type { SessionEvent, SurfaceEvent, SurfaceEventType, SurfaceOp } from './types.ts'
|
||||
|
||||
/** Runtime counterpart of the message-producing event union. */
|
||||
@@ -66,6 +67,52 @@ export function isReplacementSurfaceEvent(
|
||||
return isSurfaceEvent(event) && event.surfaceOp !== 'append'
|
||||
}
|
||||
|
||||
/**
|
||||
* Project a single event into the LLM message it derives to, or null when it
|
||||
* produces none — a non-surface event (chunk, boundary, log-only record) or an
|
||||
* empty-content assistant/message (which exists only to host usage). This is
|
||||
* THE per-node projection rule: `Session.deriveMessages` folds it over the
|
||||
* live surface, external reconstructors and pure projections fold the same
|
||||
* function over a log prefix's surface to rebuild the exact messages any
|
||||
* request was built from. The returned message is the already frozen message
|
||||
* nested in the event wrapper and shared by delivery, durable history, and
|
||||
* model requests.
|
||||
* @param event - the event to project.
|
||||
* @returns the derived message, or null when the event produces none.
|
||||
*/
|
||||
export function deriveEventMessage(event: SessionEvent): Message | null {
|
||||
// Intentionally non-exhaustive: only message-producing events derive
|
||||
// history; turn/step boundaries, chunks, usage, and errors are trace/replay
|
||||
// data.
|
||||
switch (event.type) {
|
||||
// Ordinary prompts and injected context project in user role: the event's
|
||||
// model-facing content stays verbatim. Do NOT re-add per-type framing
|
||||
// (e.g. `<context>`) here: framing is caller-owned — a producer bakes it
|
||||
// into `content`, as workspace-context does with `<system-reminder>` — or,
|
||||
// if reintroduced, must be driven by the event `meta` map and a dedicated
|
||||
// renderer, keeping this projection a verbatim pass-through. See the
|
||||
// deferred design note in
|
||||
// ../../../../.agents/notes/implemented/simplification/2026-07-20-unwrap-injected-content-envelopes.md
|
||||
case 'user/message': {
|
||||
return event.data
|
||||
}
|
||||
case 'assistant/message': {
|
||||
// Skip an empty-content assistant/message: it exists only to host a
|
||||
// max-tokens step's usage and must not inject a content-less assistant
|
||||
// turn into the provider transcript.
|
||||
if (event.data.message.content.length === 0) return null
|
||||
return event.data.message
|
||||
}
|
||||
case 'tool/result': {
|
||||
return event.data.message
|
||||
}
|
||||
default:
|
||||
// A non-surface event (boundary, chunk, log-only record) projects to
|
||||
// no message. Merge-extensible union: no assertNever here.
|
||||
return null
|
||||
}
|
||||
}
|
||||
|
||||
/** One replacement operation observed while folding a session surface. */
|
||||
export interface SurfaceFoldReplacement {
|
||||
/** Seq of the event that replaced the prior surface range. */
|
||||
|
||||
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write packages/llm/token-meter/README.md
|
||||
README.md: 0935a48a5f5773fbb280bc45e07faaa05c0a4f6e
|
||||
README.zh.md: 83282ab47e6d406cfca30bbd8af40e0c94050504
|
||||
README.md: 8f868f25f3c4caf1fdab5b50965aab41efecf5af
|
||||
README.zh.md: 3621105ff35606b62b0587063038116b4772c6cf
|
||||
|
||||
@@ -23,17 +23,21 @@ Usage accounting sums disjoint input, cache-read, cache-write, and output bucket
|
||||
|
||||
## Session projections
|
||||
|
||||
When the composition provides `ctx.sessionProjections`, token-meter registers two units through an optional child fiber.
|
||||
When the composition provides `ctx.sessionProjections`, token-meter registers three units through an optional child fiber.
|
||||
|
||||
`tokenUsage` carries the complete durable log's `uncachedInputTokens`, `outputTokens`, `cacheReadTokens`, and `cacheWriteTokens`. Usage chunks are counted even when a request later fails; a final assistant-message usage for the same `(turn, step)` replaces that sample instead of double-counting it. Reasoning remains an output subdivision. The single last-sample slot relies on a session-log ordering property: once a later step reports usage, a legal log never reports usage for an earlier step again.
|
||||
|
||||
`contextPressure` carries optional `pressureTokens` — the newest provider-reported prompt size, summing uncached input plus cache reads and writes — and optional `contextWindow` from the newest `request/context` record. Pressure stays absent until a provider reports usage; capacity stays absent for a route whose adapter advertises none. Output is excluded, so the numerator holds still while a turn streams and steps forward when the next request reports its usage.
|
||||
`contextPressure` carries optional `pressureTokens` — the newest provider-reported prompt size, summing uncached input plus cache reads and writes — optional `projectedTokens`, and optional `contextWindow` from the newest `request/context` record. Both figures stay absent until a provider reports usage; capacity stays absent for a route whose adapter advertises none. Output is excluded, so `pressureTokens` holds still while a turn streams and steps forward when the next request reports its usage.
|
||||
|
||||
Both units use the standard projection baseline, live frame, higher-seq-wins store, and JSON checkpoint paths. Unloading token-meter removes both keys. A composition without the projection seam keeps the measurement service's existing behavior.
|
||||
`projectedTokens` is what the NEXT request's prompt would cost: the sample plus the heuristic repricing of everything the surface gained or lost since it was taken, clamped at zero and folded through the same `surface-fold.ts` the measurement service replays. Only the delta is estimated, so the figure stays anchored to the provider while reacting the moment content lands — or a compaction shadows a span. That last case is why the field exists: compaction summarizes through a direct `ctx.llm.stream()` call and appends no usage of its own, so `pressureTokens` alone reports the pre-compaction prompt until an entire further turn completes. Occupancy displays read `projectedTokens`.
|
||||
|
||||
`contextBreakdown` carries heuristic `systemTokens`, `toolsTokens`, and `messageTokens` — the context's composition rather than its provider-billed size. The envelope figures reprice last-wins on every `request/header`; the message figure replays `surface-fold.ts` — the same positional fold `measure()` runs — so it equals `measure().surfaceTokens` at every event boundary and compaction shrinks it the way it shrinks the next request. All three figures use the measurement service's fixed heuristic and are estimates: they will not sum to `projectedTokens`, whose provider anchor carries exactly the error — CJK text and JSON schemas underprice badly at four characters per token — that the composition rows still contain. Present them as an approximate composition, never as a total.
|
||||
|
||||
All three units use the standard projection baseline, live frame, higher-seq-wins store, and JSON checkpoint paths. Unloading token-meter removes all three keys. A composition without the projection seam keeps the measurement service's existing behavior.
|
||||
|
||||
### Context occupancy is an approximation, by design
|
||||
|
||||
`pressureTokens` and `contextWindow` are independent last-wins fields and are **not** one atomic observation of a single request. Switching models pairs the fresh capacity with the previous route's pressure until the next request reports usage, and `pressureTokens` describes the last request rather than the surface as it stands right now.
|
||||
The occupancy fields are independent last-wins records and are **not** one atomic observation of a single request. Switching models pairs the fresh capacity with the previous route's sample until the next request reports usage, and `pressureTokens` describes the last request rather than the surface as it stands right now — `projectedTokens` carries that sample forward over the surface's movement, but its anchor is still the older request.
|
||||
|
||||
This is deliberate. An occupancy percentage is a user-facing reference figure, not a billing record or a gating input — nothing in the harness makes decisions from it, and compaction reads `measure()` instead. A UI computes occupancy by dividing measured pressure by the separately resolved capacity for the selected model.
|
||||
|
||||
|
||||
@@ -23,17 +23,21 @@ fold 跟踪完整请求标头快照、步骤边界、表层追加与替换、成
|
||||
|
||||
## 会话投影
|
||||
|
||||
当组合提供 `ctx.sessionProjections` 时，token-meter 会通过一个可选子 fiber 注册两个单元。
|
||||
当组合提供 `ctx.sessionProjections` 时，token-meter 会通过一个可选子 fiber 注册三个单元。
|
||||
|
||||
`tokenUsage` 携带完整持久日志中的 `uncachedInputTokens`、`outputTokens`、`cacheReadTokens` 和 `cacheWriteTokens`。即使请求随后失败，用量分片仍会计入；同一 `(turn, step)` 的最终 assistant 消息用量会替换该样本，而不是重复计数。推理仍是输出的一个细分项。只保留单个最新样本，依赖的是会话日志的一条顺序性质：一旦某个更晚的步骤报告了用量，合法日志就绝不会再为更早的步骤报告用量。
|
||||
|
||||
`contextPressure` 携带可选的 `pressureTokens`（提供方报告的最新提示词规模，为未缓存输入加缓存读取与写入之和），以及来自最新一条 `request/context` 记录的可选 `contextWindow`。提供方报告用量前压力保持缺失；路由适配器未公布容量时容量也保持缺失。输出不计入其中，因此轮次流式输出期间分子保持不动，等到下一个请求报告用量时才前进。
|
||||
`contextPressure` 携带可选的 `pressureTokens`（提供方报告的最新提示词规模，为未缓存输入加缓存读取与写入之和）、可选的 `projectedTokens`，以及来自最新一条 `request/context` 记录的可选 `contextWindow`。提供方报告用量前两个数字都保持缺失；路由适配器未公布容量时容量也保持缺失。输出不计入其中，因此轮次流式输出期间 `pressureTokens` 保持不动，等到下一个请求报告用量时才前进。
|
||||
|
||||
两个单元都使用标准的投影基线、实时帧、seq 高者胜值仓和 JSON 检查点路径。卸载 token-meter 会移除这两个键。不带投影 seam 的组合会保留测量服务的既有行为。
|
||||
`projectedTokens` 是「下一个请求的提示词要花多少」：在该样本之上，加上自取样以来表层增减部分的启发式重新计价，下界钳制为零，折叠走的是测量服务重放的同一份 `surface-fold.ts`。只有增量部分是估算的，因此这个数字既锚定在提供方读数上，又能在内容落地——或压缩遮蔽一段区间——的瞬间做出反应。最后这种情况正是该字段存在的理由：压缩通过直连的 `ctx.llm.stream()` 调用生成摘要，自身不追加任何用量，所以仅凭 `pressureTokens` 会一直报告压缩前的提示词规模，直到又跑完一整轮为止。占用率展示读取 `projectedTokens`。
|
||||
|
||||
`contextBreakdown` 携带启发式的 `systemTokens`、`toolsTokens` 与 `messageTokens`，描述上下文的组成而非提供方计费规模。envelope 数字在每条 `request/header` 上按后者胜重新计价；消息数字重放 `surface-fold.ts`——与 `measure()` 运行的位置折叠是同一份——因此它在每个事件边界上都等于 `measure().surfaceTokens`，压缩会像缩小下一个请求那样缩小它。三个数字都使用测量服务的固定启发式规则，属于估算值：它们加起来不等于 `projectedTokens`——后者的提供方锚点恰好把这些明细行仍然带着的误差排除在外（按「4 字符 ≈ 1 token」计价，CJK 文本与 JSON schema 会被严重低估）。请把它们当作近似的**组成**呈现，而不是总量。
|
||||
|
||||
三个单元都使用标准的投影基线、实时帧、seq 高者胜值仓和 JSON 检查点路径。卸载 token-meter 会移除这三个键。不带投影 seam 的组合会保留测量服务的既有行为。
|
||||
|
||||
### 上下文占用率是刻意为之的近似值
|
||||
|
||||
`pressureTokens` 与 `contextWindow` 是两个各自后者胜的独立字段，**不是**对单个请求的一次原子观测。切换模型时，新容量会与上一路由的压力配对，直到下一个请求报告用量为止；而 `pressureTokens` 描述的是最后一个请求，不是此刻的表层。
|
||||
这些占用率字段各自后者胜、彼此独立，**不是**对单个请求的一次原子观测。切换模型时，新容量会与上一路由的样本配对，直到下一个请求报告用量为止；而 `pressureTokens` 描述的是最后一个请求，不是此刻的表层——`projectedTokens` 把该样本沿表层的增减推进到当下，但它的锚点仍然是那个较早的请求。
|
||||
|
||||
这是刻意的选择。占用率百分比是面向用户的参考数字，既不是计费记录，也不是门控输入：harness 中没有任何环节依据它做决策，压缩改为直接读取 `measure()`。UI 用测得的压力除以为所选模型单独解析出的容量来计算占用率。
|
||||
|
||||
|
||||
@@ -30,6 +30,7 @@
|
||||
],
|
||||
"license": "BSD-3-Clause",
|
||||
"peerDependencies": {
|
||||
"@deepseek-ai/dsh-compact": "^0.0.1",
|
||||
"@deepseek-ai/dsh-invariants": "^0.0.1",
|
||||
"@deepseek-ai/dsh-llm": "^0.0.1",
|
||||
"@deepseek-ai/dsh-session": "^0.0.1",
|
||||
@@ -41,6 +42,7 @@
|
||||
"zod": "^4.4.3"
|
||||
},
|
||||
"devDependencies": {
|
||||
"@deepseek-ai/dsh-compact": "workspace:^",
|
||||
"@deepseek-ai/dsh-invariants": "workspace:^",
|
||||
"@deepseek-ai/dsh-llm": "workspace:^",
|
||||
"@deepseek-ai/dsh-session": "workspace:^",
|
||||
|
||||
69
packages/llm/token-meter/src/breakdown-projection.ts
Normal file
69
packages/llm/token-meter/src/breakdown-projection.ts
Normal file
@@ -0,0 +1,69 @@
|
||||
/**
|
||||
* Pure fold for the heuristic context-composition projection: system prompt
|
||||
* and tool schemas from the newest request envelope, conversation from the
|
||||
* live surface. Prices with the same shared estimator as the meter service,
|
||||
* so the three figures match `measure()`'s heuristic vocabulary exactly.
|
||||
*/
|
||||
|
||||
import { z } from 'zod'
|
||||
import { canonicalHeader } from '@deepseek-ai/dsh-session'
|
||||
import type { ProjectionDefinition } from '@deepseek-ai/dsh-session-projection'
|
||||
import { estimateSystemTokens, estimateToolsTokens } from './estimate.ts'
|
||||
import { foldSurfaceProjection } from './surface-projection.ts'
|
||||
import type { ShadowPriceClaim } from './surface-projection.ts'
|
||||
// Import for the `contextBreakdown` SessionProjectionMap key merge.
|
||||
import type {} from './projection.ts'
|
||||
|
||||
interface ContextBreakdownState {
|
||||
systemTokens: number
|
||||
toolsTokens: number
|
||||
messageTokens: number
|
||||
/** Shadow price armed by the immediately preceding metering event. */
|
||||
claim?: ShadowPriceClaim
|
||||
}
|
||||
|
||||
const breakdownSchema = z.object({
|
||||
systemTokens: z.number().int().nonnegative(),
|
||||
toolsTokens: z.number().int().nonnegative(),
|
||||
messageTokens: z.number().int().nonnegative(),
|
||||
}).strict()
|
||||
|
||||
/**
|
||||
* Token-meter's context-composition projection unit.
|
||||
*
|
||||
* Envelope figures are last-wins per `request/header`; the message figure
|
||||
* rides {@link foldSurfaceProjection} — the same O(1) fold the occupancy
|
||||
* projection uses — so it equals `measure().surfaceTokens` at every event
|
||||
* boundary and compaction shrinks it by its logged shadow price, the way it
|
||||
* shrinks the next request. The state is a fixed handful of numbers, so the
|
||||
* persisted checkpoint stays O(1) over the session's life.
|
||||
*/
|
||||
export const contextBreakdownProjectionDefinition:
|
||||
ProjectionDefinition<'contextBreakdown', ContextBreakdownState> = {
|
||||
key: 'contextBreakdown',
|
||||
schema: breakdownSchema,
|
||||
init: () => ({ systemTokens: 0, toolsTokens: 0, messageTokens: 0 }),
|
||||
apply: (state, event) => {
|
||||
const fold = foldSurfaceProjection(state.claim, event)
|
||||
let systemTokens = state.systemTokens
|
||||
let toolsTokens = state.toolsTokens
|
||||
if (event.type === 'request/header') {
|
||||
const header = canonicalHeader(event.data.header)
|
||||
systemTokens = estimateSystemTokens(header)
|
||||
toolsTokens = estimateToolsTokens(header)
|
||||
}
|
||||
if (systemTokens === state.systemTokens
|
||||
&& toolsTokens === state.toolsTokens
|
||||
&& fold.deltaTokens === 0
|
||||
&& fold.claim === undefined
|
||||
&& state.claim === undefined) return state
|
||||
return {
|
||||
systemTokens,
|
||||
toolsTokens,
|
||||
messageTokens: state.messageTokens + fold.deltaTokens,
|
||||
...fold.claim === undefined ? {} : { claim: fold.claim },
|
||||
}
|
||||
},
|
||||
view: ({ systemTokens, toolsTokens, messageTokens }) => ({ systemTokens, toolsTokens, messageTokens }),
|
||||
stateVersion: 2,
|
||||
}
|
||||
87
packages/llm/token-meter/src/estimate.ts
Normal file
87
packages/llm/token-meter/src/estimate.ts
Normal file
@@ -0,0 +1,87 @@
|
||||
/**
|
||||
* Fixed-density heuristic token pricing shared by the meter service and the
|
||||
* pure context-breakdown projection, so both surfaces price identical content
|
||||
* to identical numbers.
|
||||
*
|
||||
* @module @deepseek-ai/dsh-token-meter/estimate
|
||||
*/
|
||||
|
||||
import type { ContentBlock, Message } from '@deepseek-ai/dsh-llm'
|
||||
import type { EpochHeader } from '@deepseek-ai/dsh-session'
|
||||
|
||||
/** Fixed text-density estimate used until exact tokenization is needed. */
|
||||
const CHARS_PER_TOKEN = 4
|
||||
|
||||
/** Per-block structural overhead for JSON framing and type tags. */
|
||||
const BLOCK_OVERHEAD = 4
|
||||
|
||||
/** Role-field framing overhead added to every priced message. */
|
||||
export const ROLE_OVERHEAD = 4
|
||||
|
||||
/**
|
||||
* Price content blocks recursively under the fixed density heuristic.
|
||||
* @param blocks - content blocks to price without mutation.
|
||||
* @returns heuristic tokens including per-block structural overhead.
|
||||
*/
|
||||
export function estimateContent(blocks: readonly ContentBlock[]): number {
|
||||
let tokens = 0
|
||||
for (const block of blocks) {
|
||||
switch (block.type) {
|
||||
case 'text':
|
||||
case 'reasoning':
|
||||
tokens += Math.ceil(block.text.length / CHARS_PER_TOKEN) + BLOCK_OVERHEAD
|
||||
break
|
||||
case 'tool-call':
|
||||
tokens += Math.ceil(block.name.length / CHARS_PER_TOKEN)
|
||||
+ Math.ceil(block.arguments.length / CHARS_PER_TOKEN)
|
||||
+ BLOCK_OVERHEAD
|
||||
break
|
||||
case 'tool-result':
|
||||
tokens += estimateContent(block.content) + BLOCK_OVERHEAD
|
||||
break
|
||||
default:
|
||||
// ContentBlockMap is merge-extensible; unknown blocks retain a
|
||||
// conservative structural JSON price under the fixed heuristic.
|
||||
tokens += BLOCK_OVERHEAD + Math.ceil(JSON.stringify(block).length / CHARS_PER_TOKEN)
|
||||
}
|
||||
}
|
||||
return tokens
|
||||
}
|
||||
|
||||
/**
|
||||
* Heuristically price one model-visible message.
|
||||
* @param message - message to price without mutation.
|
||||
* @returns content and role-framing tokens under the fixed heuristic.
|
||||
*/
|
||||
export function estimateMessage(message: Message): number {
|
||||
return estimateContent(message.content) + ROLE_OVERHEAD
|
||||
}
|
||||
|
||||
/**
|
||||
* Price the system-prompt part of a canonical request envelope.
|
||||
* @param header - canonical envelope, or undefined before any request.
|
||||
* @returns heuristic system-prompt tokens; 0 when absent.
|
||||
*/
|
||||
export function estimateSystemTokens(header: EpochHeader | undefined): number {
|
||||
if (header?.system === undefined) return 0
|
||||
return Math.ceil(header.system.length / CHARS_PER_TOKEN) + ROLE_OVERHEAD
|
||||
}
|
||||
|
||||
/**
|
||||
* Price the tool-schema part of a canonical request envelope.
|
||||
* @param header - canonical envelope, or undefined before any request.
|
||||
* @returns heuristic tool-schema tokens; 0 when absent or empty.
|
||||
*/
|
||||
export function estimateToolsTokens(header: EpochHeader | undefined): number {
|
||||
if (header?.tools === undefined || header.tools.length === 0) return 0
|
||||
return Math.ceil(JSON.stringify(header.tools).length / CHARS_PER_TOKEN) + BLOCK_OVERHEAD
|
||||
}
|
||||
|
||||
/**
|
||||
* Price the complete non-surface request envelope.
|
||||
* @param header - canonical envelope, or undefined before any request.
|
||||
* @returns heuristic system plus tool tokens.
|
||||
*/
|
||||
export function estimateHeader(header: EpochHeader | undefined): number {
|
||||
return estimateSystemTokens(header) + estimateToolsTokens(header)
|
||||
}
|
||||
@@ -7,8 +7,8 @@
|
||||
import { Context, Service } from 'cordis'
|
||||
import z from 'schemastery'
|
||||
import { BlockAssembler, deepFreeze } from '@deepseek-ai/dsh-llm'
|
||||
import type { ContentBlock, Message, TokenUsage } from '@deepseek-ai/dsh-llm'
|
||||
import type { EpochHeader, Session, SessionEvent, SurfaceEvent } from '@deepseek-ai/dsh-session'
|
||||
import type { Message, TokenUsage } from '@deepseek-ai/dsh-llm'
|
||||
import type { EpochHeader, Session, SessionEvent } from '@deepseek-ai/dsh-session'
|
||||
import { canonicalHeader, headerEquals, isSurfaceEvent } from '@deepseek-ai/dsh-session'
|
||||
// Type-only: resolves the optional projection registry Context seam.
|
||||
import type {} from '@deepseek-ai/dsh-session-projection'
|
||||
@@ -18,19 +18,13 @@ import type {
|
||||
TokenMeterConfig,
|
||||
TokenSurfaceNode,
|
||||
} from './types.ts'
|
||||
import { contextBreakdownProjectionDefinition } from './breakdown-projection.ts'
|
||||
import { contextPressureProjectionDefinition, tokenUsageProjectionDefinition } from './usage-projection.ts'
|
||||
import { estimateContent, estimateHeader, estimateMessage, ROLE_OVERHEAD } from './estimate.ts'
|
||||
import { foldSurfaceTokens } from './surface-fold.ts'
|
||||
|
||||
export type * from './types.ts'
|
||||
|
||||
/** Fixed text-density estimate used until exact tokenization is needed. */
|
||||
const CHARS_PER_TOKEN = 4
|
||||
|
||||
/** Per-block structural overhead for JSON framing and type tags. */
|
||||
const BLOCK_OVERHEAD = 4
|
||||
|
||||
/** Role-field framing overhead added to every priced message. */
|
||||
const ROLE_OVERHEAD = 4
|
||||
|
||||
interface MeasurementAnchor {
|
||||
readonly header: EpochHeader | undefined
|
||||
readonly surfaceTokens: number
|
||||
@@ -46,11 +40,6 @@ interface ReplayState {
|
||||
anchor: MeasurementAnchor | undefined
|
||||
}
|
||||
|
||||
interface PreparedSurfaceMutation {
|
||||
readonly tokens: number
|
||||
commit(state: ReplayState): void
|
||||
}
|
||||
|
||||
/** Sum disjoint provider usage buckets without double-counting reasoning output. */
|
||||
function usageTokens(usage: TokenUsage): number {
|
||||
return usage.inputTokens
|
||||
@@ -98,6 +87,7 @@ export class TokenMeterService extends Service {
|
||||
ctx.inject(['sessionProjections'], (projectionCtx) => {
|
||||
projectionCtx.sessionProjections.register(tokenUsageProjectionDefinition)
|
||||
projectionCtx.sessionProjections.register(contextPressureProjectionDefinition)
|
||||
projectionCtx.sessionProjections.register(contextBreakdownProjectionDefinition)
|
||||
})
|
||||
|
||||
// Readers catch up independently, while eager observation bounds ordinary
|
||||
@@ -141,7 +131,7 @@ export class TokenMeterService extends Service {
|
||||
} else {
|
||||
baseline = {
|
||||
kind: 'estimated',
|
||||
tokens: this._estimateHeader(header) + state.surfaceTokens,
|
||||
tokens: estimateHeader(header) + state.surfaceTokens,
|
||||
}
|
||||
surfaceDeltaTokens = 0
|
||||
}
|
||||
@@ -157,12 +147,13 @@ export class TokenMeterService extends Service {
|
||||
}
|
||||
|
||||
/**
|
||||
* Heuristically price one model-visible message.
|
||||
* Heuristically price one model-visible message (instance face of the pure
|
||||
* `estimateMessage` export from `estimate.ts`).
|
||||
* @param message - message to price without mutation.
|
||||
* @returns content and role-framing tokens under the fixed service heuristic.
|
||||
*/
|
||||
estimateMessage(message: Message): number {
|
||||
return this._estimateContent(message.content) + ROLE_OVERHEAD
|
||||
return estimateMessage(message)
|
||||
}
|
||||
|
||||
/** Catch one session's fold up to the current durable tail. */
|
||||
@@ -224,7 +215,7 @@ export class TokenMeterService extends Service {
|
||||
}
|
||||
|
||||
const surface = isSurfaceEvent(event)
|
||||
? this._prepareSurfaceMutation(session, state, event)
|
||||
? foldSurfaceTokens(state.surface, event)
|
||||
: undefined
|
||||
|
||||
if (event.type === 'assistant/message') {
|
||||
@@ -246,7 +237,7 @@ export class TokenMeterService extends Service {
|
||||
)
|
||||
const anchorSurfaceTokens = stepStart.surfaceTokens + providerAssistantTokens
|
||||
const providerTokens = usageTokens(event.data.usage)
|
||||
const estimatedAnchorTokens = this._estimateHeader(nextHeader) + anchorSurfaceTokens
|
||||
const estimatedAnchorTokens = estimateHeader(nextHeader) + anchorSurfaceTokens
|
||||
nextAnchor = {
|
||||
header: nextHeader,
|
||||
surfaceTokens: anchorSurfaceTokens,
|
||||
@@ -263,7 +254,7 @@ export class TokenMeterService extends Service {
|
||||
surfaceTokens: anchorSurfaceTokens,
|
||||
baseline: {
|
||||
kind: 'estimated',
|
||||
tokens: this._estimateHeader(nextHeader) + anchorSurfaceTokens,
|
||||
tokens: estimateHeader(nextHeader) + anchorSurfaceTokens,
|
||||
},
|
||||
}
|
||||
}
|
||||
@@ -271,53 +262,13 @@ export class TokenMeterService extends Service {
|
||||
|
||||
state.header = nextHeader
|
||||
state.stepStart = nextStepStart
|
||||
if (surface !== undefined) surface.commit(state)
|
||||
if (surface !== undefined) {
|
||||
state.surface = surface.nodes
|
||||
state.surfaceTokens += surface.deltaTokens
|
||||
}
|
||||
state.anchor = nextAnchor
|
||||
}
|
||||
|
||||
/** Validate one surface operation and return its allocation-light commit. */
|
||||
private _prepareSurfaceMutation(
|
||||
session: Session,
|
||||
state: ReplayState,
|
||||
event: SurfaceEvent,
|
||||
): PreparedSurfaceMutation {
|
||||
const tokens = this._estimateSurfaceEvent(session, event)
|
||||
const op = event.surfaceOp
|
||||
if (op === 'append') {
|
||||
return {
|
||||
tokens,
|
||||
commit(target) {
|
||||
target.surface.push({ seq: event.seq, tokens })
|
||||
target.surfaceTokens += tokens
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
const startIdx = state.surface.findIndex(node => node.seq === op.start)
|
||||
const endIdx = state.surface.findIndex(node => node.seq === op.end)
|
||||
if (startIdx === -1 || endIdx === -1 || startIdx > endIdx) {
|
||||
throw new Error(
|
||||
`token meter: replace at seq ${event.seq} has invalid current range ${op.start}-${op.end}`,
|
||||
)
|
||||
}
|
||||
const removedTokens = state.surface
|
||||
.slice(startIdx, endIdx + 1)
|
||||
.reduce((total, node) => total + node.tokens, 0)
|
||||
return {
|
||||
tokens,
|
||||
commit(target) {
|
||||
target.surface.splice(startIdx, endIdx - startIdx + 1, { seq: event.seq, tokens })
|
||||
target.surfaceTokens += tokens - removedTokens
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
/** Price one current surface event exactly as it projects to a request. */
|
||||
private _estimateSurfaceEvent(session: Session, event: SurfaceEvent): number {
|
||||
const message = session.deriveEventMessage(event)
|
||||
return message === null ? 0 : this.estimateMessage(message)
|
||||
}
|
||||
|
||||
/**
|
||||
* Reassemble provider output from exact chunk provenance for a usage anchor.
|
||||
* Missing legacy provenance conservatively treats the durable output as the
|
||||
@@ -355,46 +306,7 @@ export class TokenMeterService extends Service {
|
||||
assembler.push(sourceEvent.data.chunk)
|
||||
}
|
||||
const providerContent = assembler.blocks()
|
||||
return providerContent.length === 0 ? 0 : this._estimateContent(providerContent) + ROLE_OVERHEAD
|
||||
}
|
||||
|
||||
/** Price content blocks recursively under the fixed density heuristic. */
|
||||
private _estimateContent(blocks: readonly ContentBlock[]): number {
|
||||
let tokens = 0
|
||||
for (const block of blocks) {
|
||||
switch (block.type) {
|
||||
case 'text':
|
||||
case 'reasoning':
|
||||
tokens += Math.ceil(block.text.length / CHARS_PER_TOKEN) + BLOCK_OVERHEAD
|
||||
break
|
||||
case 'tool-call':
|
||||
tokens += Math.ceil(block.name.length / CHARS_PER_TOKEN)
|
||||
+ Math.ceil(block.arguments.length / CHARS_PER_TOKEN)
|
||||
+ BLOCK_OVERHEAD
|
||||
break
|
||||
case 'tool-result':
|
||||
tokens += this._estimateContent(block.content) + BLOCK_OVERHEAD
|
||||
break
|
||||
default:
|
||||
// ContentBlockMap is merge-extensible; unknown blocks retain a
|
||||
// conservative structural JSON price under the fixed heuristic.
|
||||
tokens += BLOCK_OVERHEAD + Math.ceil(JSON.stringify(block).length / CHARS_PER_TOKEN)
|
||||
}
|
||||
}
|
||||
return tokens
|
||||
}
|
||||
|
||||
/** Price the canonical non-surface request envelope. */
|
||||
private _estimateHeader(header: EpochHeader | undefined): number {
|
||||
if (header === undefined) return 0
|
||||
let tokens = 0
|
||||
if (header.system !== undefined) {
|
||||
tokens += Math.ceil(header.system.length / CHARS_PER_TOKEN) + ROLE_OVERHEAD
|
||||
}
|
||||
if (header.tools !== undefined && header.tools.length > 0) {
|
||||
tokens += Math.ceil(JSON.stringify(header.tools).length / CHARS_PER_TOKEN) + BLOCK_OVERHEAD
|
||||
}
|
||||
return tokens
|
||||
return providerContent.length === 0 ? 0 : estimateContent(providerContent) + ROLE_OVERHEAD
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -17,9 +17,14 @@ export const inject = ['invariants']
|
||||
/**
|
||||
* No runtime invariant: token estimates are per-call outputs and the private
|
||||
* session cache is invalidated at its event mutation boundary. The package's
|
||||
* projection does expose an observation stream, but its schema fixes the JSON
|
||||
* payload and its pure fold replaces same-step samples; totals need not be
|
||||
* monotone when a final usage sample corrects an earlier chunk.
|
||||
* three projections do expose observation streams, but their schemas fix the
|
||||
* JSON payloads; the usage folds replace same-step samples, so totals need not
|
||||
* be monotone when a final sample corrects an earlier chunk, and the
|
||||
* composition fold prices through the same `estimate.ts` heuristic as the
|
||||
* measurement service and subtracts producer-logged shadow prices derived
|
||||
* from that service's own nodes, which makes its message figure equal
|
||||
* `measure().surfaceTokens` by construction rather than by a relation worth
|
||||
* observing at runtime.
|
||||
*/
|
||||
const install: InvariantInstaller = () => {}
|
||||
|
||||
|
||||
@@ -20,13 +20,12 @@ export interface TokenUsageProjection {
|
||||
/**
|
||||
* Approximate context occupancy for a status display.
|
||||
*
|
||||
* The two fields, when present, are deliberately NOT one atomic request
|
||||
* observation: `pressureTokens` is the newest provider-reported prompt size,
|
||||
* `contextWindow` the newest recorded route capacity. Switching models can
|
||||
* therefore pair a fresh capacity with the previous route's pressure until the
|
||||
* next request reports usage. This is an intentional trade — the value is a
|
||||
* user-facing reference, not a billing or gating input. See the token-meter
|
||||
* README for the full rationale.
|
||||
* The fields, when present, are deliberately NOT one atomic request
|
||||
* observation: each is a last-wins record of a different moment. Switching
|
||||
* models can therefore pair a fresh capacity with the previous route's
|
||||
* pressure until the next request reports usage. This is an intentional trade
|
||||
* — the value is a user-facing reference, not a billing or gating input. See
|
||||
* the token-meter README for the full rationale.
|
||||
*/
|
||||
export interface ContextPressureProjection {
|
||||
/**
|
||||
@@ -35,15 +34,44 @@ export interface ContextPressureProjection {
|
||||
* grow as the current turn streams. Absent until a provider reports usage.
|
||||
*/
|
||||
pressureTokens?: number
|
||||
/**
|
||||
* What the NEXT request's prompt would cost: {@link pressureTokens} plus the
|
||||
* heuristic repricing of everything the surface gained or lost since that
|
||||
* sample. Only the delta is estimated, so the figure stays anchored to the
|
||||
* provider while still reacting the moment a compaction shadows a span —
|
||||
* which `pressureTokens` alone cannot do, since compaction reports no usage
|
||||
* of its own. Absent until a provider reports usage.
|
||||
*/
|
||||
projectedTokens?: number
|
||||
/** Newest recorded route capacity; absent when no adapter advertised one. */
|
||||
contextWindow?: number
|
||||
}
|
||||
|
||||
/**
|
||||
* Heuristic composition of the next request's context: what the prompt is
|
||||
* made of, not what it costs. All three figures use the meter's fixed
|
||||
* density estimate, so they will not sum to the provider-anchored
|
||||
* `projectedTokens`: the estimator systematically underprices CJK text and
|
||||
* JSON schemas, which is exactly the error the anchoring in
|
||||
* {@link ContextPressureProjection.projectedTokens} keeps out of the occupancy
|
||||
* figure. Present these as approximations of composition, never as a total.
|
||||
*/
|
||||
export interface ContextBreakdownProjection {
|
||||
/** Heuristic tokens of the newest request envelope's system prompt; 0 before any request. */
|
||||
systemTokens: number
|
||||
/** Heuristic tokens of the newest request envelope's tool schemas; 0 before any request. */
|
||||
toolsTokens: number
|
||||
/** Heuristic tokens of the current model-visible conversation surface. */
|
||||
messageTokens: number
|
||||
}
|
||||
|
||||
declare module '@deepseek-ai/dsh-session-projection/types' {
|
||||
interface SessionProjectionMap {
|
||||
/** Provider-reported usage accumulated across the complete durable log. */
|
||||
tokenUsage: TokenUsageProjection
|
||||
/** Newest request pressure paired with the newest known route capacity. */
|
||||
contextPressure: ContextPressureProjection
|
||||
/** Heuristic system/tools/message composition of the next request. */
|
||||
contextBreakdown: ContextBreakdownProjection
|
||||
}
|
||||
}
|
||||
|
||||
64
packages/llm/token-meter/src/surface-fold.ts
Normal file
64
packages/llm/token-meter/src/surface-fold.ts
Normal file
@@ -0,0 +1,64 @@
|
||||
/**
|
||||
* The measurement service's positional surface fold: the per-node priced
|
||||
* surface `measure()` serves and compaction plans against. The projection
|
||||
* units deliberately do NOT share this fold — their state must stay O(1)
|
||||
* for the persisted checkpoint, so they ride `surface-projection.ts`'s
|
||||
* shadow-price protocol instead. The two stay in agreement by construction:
|
||||
* both price through `estimate.ts`, and every logged shadow price is derived
|
||||
* from THIS fold's nodes by the replace producer.
|
||||
*
|
||||
* @module @deepseek-ai/dsh-token-meter/surface-fold
|
||||
*/
|
||||
|
||||
import { deriveEventMessage } from '@deepseek-ai/dsh-session'
|
||||
import type { SurfaceEvent } from '@deepseek-ai/dsh-session'
|
||||
import type { TokenSurfaceNode } from './types.ts'
|
||||
import { estimateMessage } from './estimate.ts'
|
||||
|
||||
/** One surface event's placement and cost against the surface preceding it. */
|
||||
export interface SurfaceTokenFold {
|
||||
/** Heuristic price of the event's own message; 0 when it derives none. */
|
||||
readonly tokens: number
|
||||
/** The surface after the event, detached from the input. */
|
||||
readonly nodes: TokenSurfaceNode[]
|
||||
/** Signed change in the surface total: `tokens` minus anything shadowed. */
|
||||
readonly deltaTokens: number
|
||||
}
|
||||
|
||||
/**
|
||||
* Fold one surface event onto a priced surface.
|
||||
*
|
||||
* Total and allocation-fresh: the caller assigns the result rather than
|
||||
* mutating in place, so a throw here leaves the caller's state untouched and
|
||||
* the same malformed event fails identically on every retry.
|
||||
* @param nodes - the priced surface preceding this event, in model-visible order.
|
||||
* @param event - the surface event to place.
|
||||
* @returns the event's price, the next surface, and the signed total delta.
|
||||
* @throws when a replacement names a range absent from `nodes` — committed
|
||||
* logs are surface-validated at append time, so an unresolvable range is log
|
||||
* corruption and must fail loud rather than skip the event.
|
||||
*/
|
||||
export function foldSurfaceTokens(
|
||||
nodes: readonly TokenSurfaceNode[],
|
||||
event: SurfaceEvent,
|
||||
): SurfaceTokenFold {
|
||||
const message = deriveEventMessage(event)
|
||||
const tokens = message === null ? 0 : estimateMessage(message)
|
||||
const op = event.surfaceOp
|
||||
if (op === 'append') {
|
||||
return { tokens, nodes: [...nodes, { seq: event.seq, tokens }], deltaTokens: tokens }
|
||||
}
|
||||
const startIdx = nodes.findIndex(node => node.seq === op.start)
|
||||
const endIdx = nodes.findIndex(node => node.seq === op.end)
|
||||
if (startIdx === -1 || endIdx === -1 || startIdx > endIdx) {
|
||||
throw new Error(
|
||||
`token surface: replace at seq ${event.seq} has invalid current range ${op.start}-${op.end}`,
|
||||
)
|
||||
}
|
||||
const removed = nodes
|
||||
.slice(startIdx, endIdx + 1)
|
||||
.reduce((total, node) => total + node.tokens, 0)
|
||||
const next = [...nodes]
|
||||
next.splice(startIdx, endIdx - startIdx + 1, { seq: event.seq, tokens })
|
||||
return { tokens, nodes: next, deltaTokens: tokens - removed }
|
||||
}
|
||||
84
packages/llm/token-meter/src/surface-projection.ts
Normal file
84
packages/llm/token-meter/src/surface-projection.ts
Normal file
@@ -0,0 +1,84 @@
|
||||
/**
|
||||
* The O(1) surface-token fold shared by the token-meter projection units.
|
||||
*
|
||||
* A projection state must stay bounded — the persisted projection cache
|
||||
* checkpoints every unit's whole state, so carrying the priced surface
|
||||
* (one node per model-visible message) would grow a checkpoint without
|
||||
* bound over the session's life. Instead, replacements ride the compact
|
||||
* seam's shadow-price protocol: the metering event immediately before a
|
||||
* surface `replace` (`compact/summary` or `compact/prune`) states the
|
||||
* heuristic price of the exact replaced range, so the fold keeps a running
|
||||
* total plus at most one pending claim and never retains per-node prices.
|
||||
* The counts are exact by construction: producers derive them from the same
|
||||
* fixed estimator this module prices appends with.
|
||||
*
|
||||
* @module @deepseek-ai/dsh-token-meter/surface-projection
|
||||
*/
|
||||
|
||||
import { deriveEventMessage, isSurfaceEvent } from '@deepseek-ai/dsh-session'
|
||||
import type { SessionEvent } from '@deepseek-ai/dsh-session'
|
||||
// Type-only: the `compact/*` SessionEventMap merges (shadow-price events).
|
||||
import type {} from '@deepseek-ai/dsh-compact'
|
||||
import { estimateMessage } from './estimate.ts'
|
||||
|
||||
/**
|
||||
* One armed shadow price: the heuristic tokens of the surface range the
|
||||
* IMMEDIATELY following event replaces. Plain JSON — it is part of the
|
||||
* persisted unit state while armed.
|
||||
*/
|
||||
export interface ShadowPriceClaim {
|
||||
/** Declared inclusive first surface-node seq of the priced range. */
|
||||
start: number
|
||||
/** Declared inclusive last surface-node seq of the priced range. */
|
||||
end: number
|
||||
/** Heuristic tokens of the priced range under the fixed estimator. */
|
||||
tokens: number
|
||||
}
|
||||
|
||||
/** One event's effect on a running surface-token total. */
|
||||
export interface SurfaceTokensFold {
|
||||
/** Signed change in the surface total; 0 for events off the surface. */
|
||||
readonly deltaTokens: number
|
||||
/** Claim to carry into the next event; undefined when none survives. */
|
||||
readonly claim: ShadowPriceClaim | undefined
|
||||
}
|
||||
|
||||
/**
|
||||
* Fold one committed event onto a running surface-token total.
|
||||
*
|
||||
* A shadow-price event arms a claim; any other event expires it, and a
|
||||
* surface `replace` must consume a claim naming its exact range — the
|
||||
* producers append the metering event and the replacement synchronously
|
||||
* adjacent, so a surviving claim always prices the very next event.
|
||||
* @param claim - the claim armed by the immediately preceding event, if any.
|
||||
* @param event - the next committed session event.
|
||||
* @returns the signed token delta and the claim state after this event.
|
||||
* @throws when a replacement arrives without a claim for its exact range —
|
||||
* every in-repo replace producer meters its replacement, so an unpriced
|
||||
* replacement is a shadow-price contract violation and must fail loud
|
||||
* rather than let the total drift.
|
||||
*/
|
||||
export function foldSurfaceProjection(
|
||||
claim: ShadowPriceClaim | undefined,
|
||||
event: SessionEvent,
|
||||
): SurfaceTokensFold {
|
||||
if (event.type === 'compact/summary' || event.type === 'compact/prune') {
|
||||
const { shadowedRange, shadowedTokenCount } = event.data
|
||||
return {
|
||||
deltaTokens: 0,
|
||||
claim: { start: shadowedRange.start, end: shadowedRange.end, tokens: shadowedTokenCount },
|
||||
}
|
||||
}
|
||||
if (!isSurfaceEvent(event)) return { deltaTokens: 0, claim: undefined }
|
||||
const message = deriveEventMessage(event)
|
||||
const tokens = message === null ? 0 : estimateMessage(message)
|
||||
const op = event.surfaceOp
|
||||
if (op === 'append') return { deltaTokens: tokens, claim: undefined }
|
||||
if (claim === undefined || claim.start !== op.start || claim.end !== op.end) {
|
||||
throw new Error(
|
||||
`token surface: replace at seq ${event.seq} over range ${op.start}-${op.end} has no adjacent shadow price`
|
||||
+ (claim === undefined ? '' : ` (armed claim covers ${claim.start}-${claim.end})`),
|
||||
)
|
||||
}
|
||||
return { deltaTokens: tokens - claim.tokens, claim: undefined }
|
||||
}
|
||||
@@ -6,7 +6,7 @@
|
||||
|
||||
import type { TokenUsage } from '@deepseek-ai/dsh-llm'
|
||||
|
||||
export type { ContextPressureProjection, TokenUsageProjection } from './projection.ts'
|
||||
export type { ContextBreakdownProjection, ContextPressureProjection, TokenUsageProjection } from './projection.ts'
|
||||
|
||||
/** Token-meter plugin configuration; the fixed estimator has no settings. */
|
||||
export type TokenMeterConfig = Record<string, never>
|
||||
|
||||
@@ -4,8 +4,11 @@
|
||||
|
||||
import { z } from 'zod'
|
||||
import type { TokenUsage } from '@deepseek-ai/dsh-llm'
|
||||
import type { SessionEvent } from '@deepseek-ai/dsh-session'
|
||||
import type { ProjectionDefinition } from '@deepseek-ai/dsh-session-projection'
|
||||
import type { ContextPressureProjection, TokenUsageProjection } from './projection.ts'
|
||||
import { foldSurfaceProjection } from './surface-projection.ts'
|
||||
import type { ShadowPriceClaim } from './surface-projection.ts'
|
||||
|
||||
interface UsageSample {
|
||||
turn: number
|
||||
@@ -60,6 +63,7 @@ const projectionSchema = z.object({
|
||||
// `number | undefined` where the interface declares absent-or-number fields.
|
||||
const pressureSchema = z.object({
|
||||
pressureTokens: z.number().int().nonnegative().optional(),
|
||||
projectedTokens: z.number().int().nonnegative().optional(),
|
||||
contextWindow: z.number().int().positive().optional(),
|
||||
}).strict() as unknown as z.ZodType<ContextPressureProjection>
|
||||
|
||||
@@ -67,6 +71,29 @@ const pressureSchema = z.object({
|
||||
const pressureFrom = (usage: TokenUsage): number =>
|
||||
usage.inputTokens + (usage.cacheReadTokens ?? 0) + (usage.cacheWriteTokens ?? 0)
|
||||
|
||||
/** The usage a chunk or finalized message reports for its step, if any. */
|
||||
const usageOf = (event: SessionEvent): TokenUsage | undefined =>
|
||||
event.type === 'assistant/chunk' && event.data.chunk.type === 'usage'
|
||||
? event.data.chunk.usage
|
||||
: event.type === 'assistant/message'
|
||||
? event.data.usage
|
||||
: undefined
|
||||
|
||||
/**
|
||||
* Context-occupancy state: the two independent last-wins records plus the
|
||||
* O(1) running surface total needed to carry the newest sample forward.
|
||||
*/
|
||||
interface ContextPressureState {
|
||||
contextWindow?: number
|
||||
pressureTokens?: number
|
||||
/** Running heuristic total over the current surface ({@link foldSurfaceProjection}). */
|
||||
surfaceTokens: number
|
||||
/** {@link surfaceTokens} at the newest usage sample; absent until one lands. */
|
||||
sampledSurfaceTokens?: number
|
||||
/** Shadow price armed by the immediately preceding metering event. */
|
||||
claim?: ShadowPriceClaim
|
||||
}
|
||||
|
||||
/**
|
||||
* Token-meter's session projection unit.
|
||||
*
|
||||
@@ -115,39 +142,64 @@ ProjectionDefinition<'tokenUsage', TokenUsageState> = {
|
||||
/**
|
||||
* Token-meter's context-occupancy projection unit.
|
||||
*
|
||||
* Two independent last-wins slots: the newest usage sample supplies the
|
||||
* Independent last-wins slots: the newest usage sample supplies the provider
|
||||
* numerator, the newest `request/context` record the denominator. Both are
|
||||
* whole values, so replay order alone decides the result and no cross-field
|
||||
* consistency is claimed — the pair is explicitly not one atomic request
|
||||
* observation (see {@link ContextPressureProjection}).
|
||||
*
|
||||
* The numerator is prompt-side only, so it holds still while a turn streams
|
||||
* and steps forward once the next request reports its usage.
|
||||
* `pressureTokens` is prompt-side only, so it holds still while a turn streams
|
||||
* and steps forward once the next request reports its usage. Because nothing
|
||||
* but a request reports usage, it also cannot see a compaction: the fold
|
||||
* therefore carries a running surface total alongside it and publishes
|
||||
* `projectedTokens` — the sample plus the surface's signed movement since it
|
||||
* was taken — so occupancy answers for the next request rather than the last
|
||||
* one. The total rides {@link foldSurfaceProjection}, so the state stays O(1)
|
||||
* and a replacement shrinks it by its logged shadow price. A usage sample is
|
||||
* stamped BEFORE the same event joins the surface, so an `assistant/message`
|
||||
* anchors against the surface its own request saw.
|
||||
*/
|
||||
export const contextPressureProjectionDefinition:
|
||||
ProjectionDefinition<'contextPressure', ContextPressureProjection> = {
|
||||
ProjectionDefinition<'contextPressure', ContextPressureState> = {
|
||||
key: 'contextPressure',
|
||||
schema: pressureSchema,
|
||||
init: () => ({}),
|
||||
init: () => ({ surfaceTokens: 0 }),
|
||||
apply: (state, event) => {
|
||||
const fold = foldSurfaceProjection(state.claim, event)
|
||||
let next = state
|
||||
if (event.type === 'request/context') {
|
||||
const contextWindow = event.data.contextWindow
|
||||
if (contextWindow === state.contextWindow) return state
|
||||
if (contextWindow !== undefined) return { ...state, contextWindow }
|
||||
const { contextWindow: _removed, ...withoutContextWindow } = state
|
||||
return withoutContextWindow
|
||||
if (contextWindow !== state.contextWindow) {
|
||||
if (contextWindow !== undefined) {
|
||||
next = { ...next, contextWindow }
|
||||
} else {
|
||||
const { contextWindow: _removed, ...withoutContextWindow } = next
|
||||
next = withoutContextWindow
|
||||
}
|
||||
}
|
||||
}
|
||||
const usage = event.type === 'assistant/chunk' && event.data.chunk.type === 'usage'
|
||||
? event.data.chunk.usage
|
||||
: event.type === 'assistant/message'
|
||||
? event.data.usage
|
||||
: undefined
|
||||
if (usage === undefined) return state
|
||||
const pressureTokens = pressureFrom(usage)
|
||||
return pressureTokens === state.pressureTokens
|
||||
? state
|
||||
: { ...state, pressureTokens }
|
||||
const usage = usageOf(event)
|
||||
if (usage !== undefined) {
|
||||
const pressureTokens = pressureFrom(usage)
|
||||
if (pressureTokens !== next.pressureTokens || next.sampledSurfaceTokens !== next.surfaceTokens) {
|
||||
next = { ...next, pressureTokens, sampledSurfaceTokens: next.surfaceTokens }
|
||||
}
|
||||
}
|
||||
if (fold.deltaTokens !== 0) {
|
||||
next = { ...next, surfaceTokens: next.surfaceTokens + fold.deltaTokens }
|
||||
}
|
||||
// A defined fold.claim is always freshly built, so presence decides claim
|
||||
// bookkeeping: no claim before or after this event leaves `next` as is.
|
||||
if (state.claim === undefined && fold.claim === undefined) return next
|
||||
const { claim: _expired, ...withoutClaim } = next
|
||||
return fold.claim === undefined ? withoutClaim : { ...withoutClaim, claim: fold.claim }
|
||||
},
|
||||
view: state => state,
|
||||
stateVersion: 2,
|
||||
view: ({ contextWindow, pressureTokens, surfaceTokens, sampledSurfaceTokens }) => ({
|
||||
...contextWindow === undefined ? {} : { contextWindow },
|
||||
...pressureTokens === undefined ? {} : { pressureTokens },
|
||||
...pressureTokens === undefined || sampledSurfaceTokens === undefined
|
||||
? {}
|
||||
: { projectedTokens: Math.max(0, pressureTokens + surfaceTokens - sampledSurfaceTokens) },
|
||||
}),
|
||||
stateVersion: 4,
|
||||
}
|
||||
|
||||
@@ -0,0 +1,307 @@
|
||||
// contextBreakdown projection: heuristic system/tools/message composition,
|
||||
// plus the shared estimator's pricing branches.
|
||||
|
||||
import { describe, expect, it } from 'vitest'
|
||||
import { Context } from 'cordis'
|
||||
import { createMessage, createUserMessage } from '@deepseek-ai/dsh-llm'
|
||||
import type { ContentBlock, ToolSchema } from '@deepseek-ai/dsh-llm'
|
||||
import SessionStore from '@deepseek-ai/dsh-session'
|
||||
import type { Session, SessionEvent } from '@deepseek-ai/dsh-session'
|
||||
import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection'
|
||||
import TokenMeterService from '@deepseek-ai/dsh-token-meter'
|
||||
import type { ContextBreakdownProjection } from '@deepseek-ai/dsh-token-meter/client'
|
||||
import { contextBreakdownProjectionDefinition } from '../src/breakdown-projection.ts'
|
||||
import {
|
||||
estimateContent,
|
||||
estimateHeader,
|
||||
estimateMessage,
|
||||
estimateSystemTokens,
|
||||
estimateToolsTokens,
|
||||
} from '../src/estimate.ts'
|
||||
|
||||
const CONFIG = { provider: 'test', model: 'test-model' }
|
||||
|
||||
const TOOLS: ToolSchema[] = [{
|
||||
name: 'bash',
|
||||
description: 'run a command',
|
||||
parameters: { type: 'object', properties: {} },
|
||||
}]
|
||||
|
||||
async function harness(): Promise<{ ctx: Context; session: Session }> {
|
||||
const ctx = new Context()
|
||||
await ctx.plugin(SessionStore)
|
||||
await ctx.plugin(SessionProjectionRegistry)
|
||||
await ctx.plugin(TokenMeterService)
|
||||
return { ctx, session: ctx.sessions.create() }
|
||||
}
|
||||
|
||||
const projected = (ctx: Context, session: Session): ContextBreakdownProjection => {
|
||||
const value = ctx.sessionProjections.snapshot(session).values.contextBreakdown
|
||||
if (value === undefined) throw new Error('contextBreakdown projection is not registered')
|
||||
return value
|
||||
}
|
||||
|
||||
function appendUser(session: Session, text: string): number {
|
||||
return session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text }],
|
||||
source: { kind: 'user' },
|
||||
}), { surfaceOp: 'append' }).seq
|
||||
}
|
||||
|
||||
/**
|
||||
* Meter one upcoming replacement the way compact-basic does: price the
|
||||
* replaced span from the measurement service's own nodes and log the
|
||||
* shadow-price event directly before the replace.
|
||||
*/
|
||||
function appendSummaryMeter(ctx: Context, session: Session, start: number, end: number): void {
|
||||
const nodes = ctx.tokenMeter.measure(session).nodes
|
||||
const startIdx = nodes.findIndex(node => node.seq === start)
|
||||
const endIdx = nodes.findIndex(node => node.seq === end)
|
||||
const shadowed = nodes.slice(startIdx, endIdx + 1)
|
||||
session.append('compact/summary', {
|
||||
summary: [{ type: 'text', text: 'summary' }],
|
||||
shadowedRange: { start, end },
|
||||
shadowedSeqs: shadowed.map(node => node.seq),
|
||||
shadowedTokenCount: shadowed.reduce((total, node) => total + node.tokens, 0),
|
||||
provider: 'mock',
|
||||
model: 'mock',
|
||||
})
|
||||
}
|
||||
|
||||
describe('contextBreakdown session projection', () => {
|
||||
it('serves zeros for an empty log', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
expect(projected(ctx, session)).toEqual({ systemTokens: 0, toolsTokens: 0, messageTokens: 0 })
|
||||
})
|
||||
|
||||
it('prices the newest envelope last-wins and pushes no change for a restated one', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
session.append('request/header', {
|
||||
header: { config: CONFIG, system: 'You are terse.', tools: TOOLS },
|
||||
reason: 'initial',
|
||||
})
|
||||
expect(projected(ctx, session)).toEqual({
|
||||
systemTokens: estimateSystemTokens({ config: CONFIG, system: 'You are terse.' }),
|
||||
toolsTokens: estimateToolsTokens({ config: CONFIG, tools: TOOLS }),
|
||||
messageTokens: 0,
|
||||
})
|
||||
|
||||
const changed: string[] = []
|
||||
ctx.sessionProjections.onChanged((_session, key) => { changed.push(key) })
|
||||
session.append('request/header', {
|
||||
header: { config: CONFIG, system: 'You are terse.', tools: TOOLS },
|
||||
reason: 'change',
|
||||
})
|
||||
session.append('todo/write', { todos: [] })
|
||||
expect(changed).not.toContain('contextBreakdown')
|
||||
|
||||
// A system-less, tool-less envelope prices back to zero.
|
||||
session.append('request/header', { header: { config: CONFIG }, reason: 'change' })
|
||||
expect(projected(ctx, session)).toEqual({ systemTokens: 0, toolsTokens: 0, messageTokens: 0 })
|
||||
})
|
||||
|
||||
it('sums surface appends and skips an empty-content assistant message', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
appendUser(session, 'abcd')
|
||||
session.append('step/start', { turn: 1, step: 1 })
|
||||
session.append('assistant/message', {
|
||||
turn: 1,
|
||||
step: 1,
|
||||
message: createMessage({
|
||||
role: 'assistant',
|
||||
content: [],
|
||||
source: { kind: 'model', provider: 'mock', model: 'mock' },
|
||||
}),
|
||||
usage: { inputTokens: 9, outputTokens: 0 },
|
||||
}, { surfaceOp: 'append', sourceEventSeqs: [] })
|
||||
session.append('step/end', { turn: 1, step: 1 })
|
||||
// 'abcd' prices to 9 (1 text + 4 block + 4 role); the usage-only assistant
|
||||
// message derives to no transcript entry and adds nothing.
|
||||
expect(projected(ctx, session).messageTokens).toBe(9)
|
||||
})
|
||||
|
||||
it('shrinks the message figure when a metered replacement compacts the surface', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
const first = appendUser(session, 'before compaction, a longer message')
|
||||
const second = appendUser(session, 'and a second entry')
|
||||
const summary = createUserMessage({
|
||||
content: [{ type: 'text', text: 'summary' }],
|
||||
source: { kind: 'plugin', plugin: 'test' },
|
||||
})
|
||||
appendSummaryMeter(ctx, session, first, second)
|
||||
session.append('user/message', summary, {
|
||||
surfaceOp: { op: 'replace', start: first, end: second },
|
||||
sourceEventSeqs: [first, second],
|
||||
})
|
||||
expect(projected(ctx, session).messageTokens).toBe(estimateMessage(summary))
|
||||
})
|
||||
|
||||
it('keeps the message figure equal to the service surface across appends and a compaction', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
// The panel's composition rows and `measure()` answer the same question in
|
||||
// the same vocabulary; one shared fold is what makes that true.
|
||||
const agree = (): number => {
|
||||
const messageTokens = projected(ctx, session).messageTokens
|
||||
expect(messageTokens).toBe(ctx.tokenMeter.measure(session).surfaceTokens)
|
||||
return messageTokens
|
||||
}
|
||||
session.append('request/header', {
|
||||
header: { config: CONFIG, system: 'You are terse.', tools: TOOLS },
|
||||
reason: 'initial',
|
||||
})
|
||||
expect(agree()).toBe(0)
|
||||
|
||||
const question = appendUser(session, 'a first question, long enough to price above zero')
|
||||
session.append('step/start', { turn: 1, step: 1 })
|
||||
const answer = session.append('assistant/message', {
|
||||
turn: 1,
|
||||
step: 1,
|
||||
message: createMessage({
|
||||
role: 'assistant',
|
||||
content: [{ type: 'text', text: 'a considered answer' }],
|
||||
source: { kind: 'model', provider: 'mock', model: 'mock' },
|
||||
}),
|
||||
usage: { inputTokens: 40, outputTokens: 7 },
|
||||
}, { surfaceOp: 'append', sourceEventSeqs: [] }).seq
|
||||
session.append('step/end', { turn: 1, step: 1 })
|
||||
const grown = agree()
|
||||
expect(grown).toBeGreaterThan(0)
|
||||
|
||||
appendSummaryMeter(ctx, session, question, answer)
|
||||
// The armed shadow price must not move the published figure by itself.
|
||||
expect(agree()).toBe(grown)
|
||||
session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text: 'summary' }],
|
||||
source: { kind: 'plugin', plugin: 'test' },
|
||||
}), {
|
||||
surfaceOp: { op: 'replace', start: question, end: answer },
|
||||
sourceEventSeqs: [question, answer],
|
||||
})
|
||||
expect(agree()).toBeLessThan(grown)
|
||||
})
|
||||
|
||||
it('fails loud on a replacement without an adjacent matching shadow price', () => {
|
||||
const definition = contextBreakdownProjectionDefinition
|
||||
const replace = (start: number, end: number): SessionEvent => ({
|
||||
type: 'user/message',
|
||||
seq: 9,
|
||||
time: 0,
|
||||
data: createUserMessage({ content: [{ type: 'text', text: 'x' }], source: { kind: 'user' } }),
|
||||
surfaceOp: { op: 'replace', start, end },
|
||||
sourceEventSeqs: [start, end],
|
||||
} as unknown as SessionEvent)
|
||||
const append = (seq: number): SessionEvent => ({
|
||||
type: 'user/message',
|
||||
seq,
|
||||
time: 0,
|
||||
data: createUserMessage({ content: [{ type: 'text', text: 'x' }], source: { kind: 'user' } }),
|
||||
surfaceOp: 'append',
|
||||
} as unknown as SessionEvent)
|
||||
const meter = (start: number, end: number, seq: number): SessionEvent => ({
|
||||
type: 'compact/prune',
|
||||
seq,
|
||||
time: 0,
|
||||
data: { shadowedRange: { start, end }, shadowedSeqs: [start, end], shadowedTokenCount: 5 },
|
||||
} as unknown as SessionEvent)
|
||||
let state = definition.init()
|
||||
state = definition.apply(state, append(1))
|
||||
state = definition.apply(state, append(3))
|
||||
// No metering event at all.
|
||||
expect(() => definition.apply(state, replace(1, 3))).toThrow('no adjacent shadow price')
|
||||
// A claim for a different range does not price this replacement.
|
||||
const mismatched = definition.apply(state, meter(1, 1, 8))
|
||||
expect(() => definition.apply(mismatched, replace(1, 3))).toThrow('no adjacent shadow price')
|
||||
// A claim expires after one intervening event instead of lingering.
|
||||
let expired = definition.apply(state, meter(1, 3, 8))
|
||||
expired = definition.apply(expired, { type: 'todo/write', seq: 9, time: 0, data: { todos: [] } } as unknown as SessionEvent)
|
||||
expect(() => definition.apply(expired, replace(1, 3))).toThrow('no adjacent shadow price')
|
||||
// The armed claim prices exactly the next event's matching replacement.
|
||||
const armed = definition.apply(state, meter(1, 3, 8))
|
||||
expect(definition.view(definition.apply(armed, replace(1, 3))).messageTokens)
|
||||
.toBe(definition.view(state).messageTokens - 5 + estimateMessage(
|
||||
createUserMessage({ content: [{ type: 'text', text: 'x' }], source: { kind: 'user' } }),
|
||||
))
|
||||
})
|
||||
|
||||
it('keeps the persisted checkpoint O(1) as the surface grows and compacts', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
const first = appendUser(session, 'the first of many messages')
|
||||
for (let index = 0; index < 24; index += 1) appendUser(session, `message number ${index} with some text`)
|
||||
const last = appendUser(session, 'the last message before compaction')
|
||||
const stateKeys = (): string[] => {
|
||||
const row = ctx.sessionProjections.checkpoint(session)['contextBreakdown']
|
||||
if (row === undefined) throw new Error('contextBreakdown checkpoint row is missing')
|
||||
return Object.keys(row.val as Record<string, unknown>).sort()
|
||||
}
|
||||
// Growth adds no per-node bookkeeping to the durable state.
|
||||
expect(stateKeys()).toEqual(['messageTokens', 'systemTokens', 'toolsTokens'])
|
||||
const shadowed = session.surface.nodes.slice(
|
||||
session.surface.nodes.indexOf(first),
|
||||
session.surface.nodes.indexOf(last) + 1,
|
||||
)
|
||||
appendSummaryMeter(ctx, session, first, last)
|
||||
session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text: 'summary' }],
|
||||
source: { kind: 'plugin', plugin: 'test' },
|
||||
}), {
|
||||
surfaceOp: { op: 'replace', start: first, end: last },
|
||||
sourceEventSeqs: [...shadowed],
|
||||
})
|
||||
expect(stateKeys()).toEqual(['messageTokens', 'systemTokens', 'toolsTokens'])
|
||||
expect(projected(ctx, session).messageTokens)
|
||||
.toBe(ctx.tokenMeter.measure(session).surfaceTokens)
|
||||
})
|
||||
|
||||
it('restores from a JSON checkpoint and unregisters with the token-meter fiber', async () => {
|
||||
const ctx = new Context()
|
||||
await ctx.plugin(SessionStore)
|
||||
await ctx.plugin(SessionProjectionRegistry)
|
||||
const meterFiber = await ctx.plugin(TokenMeterService)
|
||||
const session = ctx.sessions.create()
|
||||
session.append('request/header', {
|
||||
header: { config: CONFIG, system: 'You are terse.' },
|
||||
reason: 'initial',
|
||||
})
|
||||
appendUser(session, 'abcd')
|
||||
const checkpoint = JSON.parse(JSON.stringify(
|
||||
ctx.sessionProjections.checkpoint(session),
|
||||
)) as ReturnType<typeof ctx.sessionProjections.checkpoint>
|
||||
|
||||
await meterFiber.dispose()
|
||||
expect(ctx.sessionProjections.snapshot(session).values).not.toHaveProperty('contextBreakdown')
|
||||
|
||||
await ctx.plugin(TokenMeterService)
|
||||
expect(ctx.sessionProjections.viewCheckpoint(checkpoint).contextBreakdown).toEqual({
|
||||
systemTokens: estimateSystemTokens({ config: CONFIG, system: 'You are terse.' }),
|
||||
toolsTokens: 0,
|
||||
messageTokens: 9,
|
||||
})
|
||||
})
|
||||
})
|
||||
|
||||
describe('shared estimator', () => {
|
||||
it('prices every content-block shape under the fixed heuristic', () => {
|
||||
expect(estimateContent([{ type: 'text', text: 'abcd' }])).toBe(5)
|
||||
expect(estimateContent([{ type: 'reasoning', text: 'abcdefgh' }] as ContentBlock[])).toBe(6)
|
||||
expect(estimateContent([{ type: 'tool-call', id: 'c' as never, name: 'bash', arguments: '{"a":1}' }])).toBe(7)
|
||||
expect(estimateContent([{
|
||||
type: 'tool-result', toolCallId: 'c' as never,
|
||||
content: [{ type: 'text', text: 'abcd' }],
|
||||
}])).toBe(9)
|
||||
const unknown = { type: 'mystery', payload: 'abc' } as unknown as ContentBlock
|
||||
expect(estimateContent([unknown])).toBe(4 + Math.ceil(JSON.stringify(unknown).length / 4))
|
||||
})
|
||||
|
||||
it('prices envelope parts independently and absent parts to zero', () => {
|
||||
expect(estimateSystemTokens(undefined)).toBe(0)
|
||||
expect(estimateSystemTokens({ config: CONFIG })).toBe(0)
|
||||
expect(estimateSystemTokens({ config: CONFIG, system: 'abcdefgh' })).toBe(6)
|
||||
expect(estimateToolsTokens(undefined)).toBe(0)
|
||||
expect(estimateToolsTokens({ config: CONFIG, tools: [] })).toBe(0)
|
||||
expect(estimateToolsTokens({ config: CONFIG, tools: TOOLS }))
|
||||
.toBe(Math.ceil(JSON.stringify(TOOLS).length / 4) + 4)
|
||||
expect(estimateHeader(undefined)).toBe(0)
|
||||
expect(estimateHeader({ config: CONFIG, system: 'abcdefgh', tools: TOOLS }))
|
||||
.toBe(6 + Math.ceil(JSON.stringify(TOOLS).length / 4) + 4)
|
||||
})
|
||||
})
|
||||
@@ -70,6 +70,26 @@ const projected = (ctx: Context, session: Session): TokenUsageProjection => {
|
||||
return value
|
||||
}
|
||||
|
||||
/**
|
||||
* Meter one upcoming replacement the way compact-basic does: price the
|
||||
* replaced span from the measurement service's own nodes and log the
|
||||
* shadow-price event directly before the replace.
|
||||
*/
|
||||
function appendSummaryMeter(ctx: Context, session: Session, start: number, end: number): void {
|
||||
const nodes = ctx.tokenMeter.measure(session).nodes
|
||||
const startIdx = nodes.findIndex(node => node.seq === start)
|
||||
const endIdx = nodes.findIndex(node => node.seq === end)
|
||||
const shadowed = nodes.slice(startIdx, endIdx + 1)
|
||||
session.append('compact/summary', {
|
||||
summary: [{ type: 'text', text: 'summary' }],
|
||||
shadowedRange: { start, end },
|
||||
shadowedSeqs: shadowed.map(node => node.seq),
|
||||
shadowedTokenCount: shadowed.reduce((total, node) => total + node.tokens, 0),
|
||||
provider: 'mock',
|
||||
model: 'mock',
|
||||
})
|
||||
}
|
||||
|
||||
describe('tokenUsage session projection', () => {
|
||||
it('serves zero buckets for an empty log', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
@@ -184,6 +204,7 @@ describe('tokenUsage session projection', () => {
|
||||
content: [{ type: 'text', text: 'before compaction' }],
|
||||
source: { kind: 'user' },
|
||||
}), { surfaceOp: 'append' })
|
||||
appendSummaryMeter(ctx, session, before.seq, before.seq)
|
||||
session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text: 'compacted' }],
|
||||
source: { kind: 'plugin', plugin: 'test' },
|
||||
@@ -235,6 +256,34 @@ function recordContext(session: Session, model: string, contextWindow?: number):
|
||||
})
|
||||
}
|
||||
|
||||
/** Append one model-visible user turn and return its surface seq. */
|
||||
function appendUser(session: Session, text: string): number {
|
||||
return session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text }],
|
||||
source: { kind: 'user' },
|
||||
}), { surfaceOp: 'append' }).seq
|
||||
}
|
||||
|
||||
/** Append one finalized assistant turn carrying its provider usage. */
|
||||
function appendAssistant(
|
||||
session: Session,
|
||||
text: string,
|
||||
usage: TokenUsage,
|
||||
turn: number,
|
||||
step: number,
|
||||
): number {
|
||||
return session.append('assistant/message', {
|
||||
turn,
|
||||
step,
|
||||
message: createMessage({
|
||||
role: 'assistant',
|
||||
content: [{ type: 'text', text }],
|
||||
source: { kind: 'model', provider: 'mock', model: 'mock' },
|
||||
}),
|
||||
usage,
|
||||
}, { surfaceOp: 'append', sourceEventSeqs: [] }).seq
|
||||
}
|
||||
|
||||
describe('contextPressure session projection', () => {
|
||||
it('serves no pressure or capacity for an empty log', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
@@ -277,9 +326,13 @@ describe('contextPressure session projection', () => {
|
||||
startStep(session, 1, 1)
|
||||
recordContext(session, 'small', 64_000)
|
||||
usageChunk(session, { inputTokens: 100, outputTokens: 10 }, 1, 1)
|
||||
expect(pressure(ctx, session)).toEqual({ pressureTokens: 100, contextWindow: 64_000 })
|
||||
expect(pressure(ctx, session)).toEqual({
|
||||
pressureTokens: 100, projectedTokens: 100, contextWindow: 64_000,
|
||||
})
|
||||
recordContext(session, 'large', 256_000)
|
||||
expect(pressure(ctx, session)).toEqual({ pressureTokens: 100, contextWindow: 256_000 })
|
||||
expect(pressure(ctx, session)).toEqual({
|
||||
pressureTokens: 100, projectedTokens: 100, contextWindow: 256_000,
|
||||
})
|
||||
})
|
||||
|
||||
it('removes an older capacity when the newest route advertises none', async () => {
|
||||
@@ -288,7 +341,7 @@ describe('contextPressure session projection', () => {
|
||||
recordContext(session, 'small', 64_000)
|
||||
usageChunk(session, { inputTokens: 100, outputTokens: 10 }, 1, 1)
|
||||
recordContext(session, 'unknown')
|
||||
expect(pressure(ctx, session)).toEqual({ pressureTokens: 100 })
|
||||
expect(pressure(ctx, session)).toEqual({ pressureTokens: 100, projectedTokens: 100 })
|
||||
})
|
||||
|
||||
it('pushes no change for unrelated events or a restated capacity', async () => {
|
||||
@@ -319,7 +372,7 @@ describe('contextPressure session projection', () => {
|
||||
const checkpoint = JSON.parse(JSON.stringify(
|
||||
ctx.sessionProjections.checkpoint(session),
|
||||
)) as ReturnType<typeof ctx.sessionProjections.checkpoint>
|
||||
expect(checkpoint.contextPressure?.ver).toBe(2)
|
||||
expect(checkpoint.contextPressure?.ver).toBe(4)
|
||||
|
||||
await meterFiber.dispose()
|
||||
expect(ctx.sessionProjections.snapshot(session).values).not.toHaveProperty('contextPressure')
|
||||
@@ -327,7 +380,62 @@ describe('contextPressure session projection', () => {
|
||||
await ctx.plugin(TokenMeterService)
|
||||
expect(ctx.sessionProjections.viewCheckpoint(checkpoint).contextPressure).toEqual({
|
||||
pressureTokens: 42,
|
||||
projectedTokens: 42,
|
||||
contextWindow: 64_000,
|
||||
})
|
||||
})
|
||||
|
||||
it('carries the sample forward over surface growth and a compaction', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
recordContext(session, 'large', 128_000)
|
||||
const question = appendUser(session, 'a first question worth a few tokens')
|
||||
startStep(session, 1, 1)
|
||||
// The provider prices the prompt its request actually carried; the sample
|
||||
// must anchor against the surface as of that request, not after the
|
||||
// assistant message joins it.
|
||||
const answer = appendAssistant(session, 'an answer of some length', { inputTokens: 900, outputTokens: 20 }, 1, 1)
|
||||
session.append('step/end', { turn: 1, step: 1 })
|
||||
const afterTurn = pressure(ctx, session)
|
||||
expect(afterTurn.pressureTokens).toBe(900)
|
||||
// The assistant message landed after the sample, so it already shows.
|
||||
expect(afterTurn.projectedTokens).toBeGreaterThan(900)
|
||||
|
||||
const grown = appendUser(session, 'a follow-up question that grows the surface further')
|
||||
const beforeCompaction = pressure(ctx, session).projectedTokens
|
||||
expect(beforeCompaction).toBeGreaterThan(afterTurn.projectedTokens!)
|
||||
|
||||
// Compaction reports no usage of its own, so `pressureTokens` cannot move;
|
||||
// the projected figure must shrink anyway — the defect this field fixes.
|
||||
appendSummaryMeter(ctx, session, question, grown)
|
||||
session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text: 'summary' }],
|
||||
source: { kind: 'plugin', plugin: 'test' },
|
||||
}), {
|
||||
surfaceOp: { op: 'replace', start: question, end: grown },
|
||||
sourceEventSeqs: [question, answer, grown],
|
||||
})
|
||||
const compacted = pressure(ctx, session)
|
||||
expect(compacted.pressureTokens).toBe(900)
|
||||
expect(compacted.projectedTokens).toBeLessThan(beforeCompaction!)
|
||||
})
|
||||
|
||||
it('clamps a projection that heuristic error drove below zero', async () => {
|
||||
const { ctx, session } = await harness()
|
||||
recordContext(session, 'large', 128_000)
|
||||
const question = appendUser(session, 'a question long enough to outprice the sample'.repeat(4))
|
||||
startStep(session, 1, 1)
|
||||
// A provider sample far below the heuristic price of what it replaced:
|
||||
// shadowing that span subtracts more than the sample holds.
|
||||
appendAssistant(session, 'ok', { inputTokens: 3, outputTokens: 1 }, 1, 1)
|
||||
session.append('step/end', { turn: 1, step: 1 })
|
||||
appendSummaryMeter(ctx, session, question, question)
|
||||
session.append('user/message', createUserMessage({
|
||||
content: [{ type: 'text', text: '.' }],
|
||||
source: { kind: 'plugin', plugin: 'test' },
|
||||
}), {
|
||||
surfaceOp: { op: 'replace', start: question, end: question },
|
||||
sourceEventSeqs: [question],
|
||||
})
|
||||
expect(pressure(ctx, session).projectedTokens).toBe(0)
|
||||
})
|
||||
})
|
||||
|
||||
@@ -23,6 +23,9 @@
|
||||
{
|
||||
"path": "../../core/session"
|
||||
},
|
||||
{
|
||||
"path": "../../compact/compact"
|
||||
},
|
||||
{
|
||||
"path": "../../session-projection/session-projection"
|
||||
},
|
||||
|
||||
Reference in New Issue
Block a user