refactor: narrow compaction surface

This commit is contained in:
Tianyi Cui
2026-07-14 01:24:20 +08:00
parent f95a411b0a
commit d7de8a8d13
11 changed files with 95 additions and 148 deletions

View File

@@ -113,12 +113,12 @@ Implementations MUST honor:
```ts cordis-catalog ```ts cordis-catalog
abstract compactIfNeeded( agent: CompactAgentContext, fullSystemPrompt: string, sessionPrefix: readonly Message[], signal: AbortSignal, ): Promise<CompactionResult | null> abstract compactIfNeeded( agent: CompactAgentContext, fullSystemPrompt: string, sessionPrefix: readonly Message[], signal: AbortSignal, ): Promise<CompactionResult | null>
abstract compactRegion( session: Session, start: number, end: number, agent: CompactAgentContext, signal?: AbortSignal, ): Promise<CompactionResult> abstract compactRegion( start: number, end: number, agent: CompactAgentContext, signal?: AbortSignal, ): Promise<CompactionResult>
``` ```
Types: [Message](../core-data-structures/core.md) Types: [Message](../core-data-structures/core.md)
Source: [`packages/compact/compact/src/index.ts:65`](../../packages/compact/compact/src/index.ts) Source: [`packages/compact/compact/src/index.ts:66`](../../packages/compact/compact/src/index.ts)
## `ctx.fs` — `FileSystem` (abstract seam) ## `ctx.fs` — `FileSystem` (abstract seam)

View File

@@ -1,6 +1,6 @@
# Compaction # Compaction
The compaction seam — a [capability seam](../rfc/implemented/architecture/2026-06-13-capability-seams.md) split like bash: interface ([dsh-compact](../../packages/compact/compact), `ctx.compact`), implementation (a backend such as [dsh-compact-basic](../../packages/compact/compact-basic)), and consumer (a `/compact` tool, deferred). Compaction is **one optional capability**, not part of the agent-loop spine — so its vocabulary lives here, not in [core.md](core.md). A tokenizer- or template-based backend is a sibling package implementing the same interface. Unlike bash, the interface necessarily depends on `dsh-session` and `dsh-llm`: its verbs are defined over a `Session` and its output is the `ContentBlock` vocabulary (see the [compaction capability-seam RFC](../rfc/implemented/feature/2026-06-18-compaction-capability-seam.md)). The compaction seam — a [capability seam](../rfc/implemented/architecture/2026-06-13-capability-seams.md) split like bash: interface ([dsh-compact](../../packages/compact/compact), `ctx.compact`), implementation (a backend such as [dsh-compact-basic](../../packages/compact/compact-basic)), and consumer (a `/compact` tool, deferred). Compaction is **one optional capability**, not part of the agent-loop spine — so its vocabulary lives here, not in [core.md](core.md). A tokenizer- or template-based backend is a sibling package implementing the same interface. Unlike bash, the interface necessarily depends on `dsh-session` and `dsh-llm`: its verbs act on an agent-owned `Session`, and its durable summary event uses the `ContentBlock` vocabulary (see the [compaction capability-seam RFC](../rfc/implemented/feature/2026-06-18-compaction-capability-seam.md)).
Source: [`packages/compact/compact/src/types.ts`](../../packages/compact/compact/src/types.ts) Source: [`packages/compact/compact/src/types.ts`](../../packages/compact/compact/src/types.ts)
@@ -20,18 +20,10 @@ These variants are merged inside a `declare module '@deepseek-ai/dsh-session'` b
## `CompactionResult` ## `CompactionResult`
What a successful compaction returns to its caller: the seqs of the three appended `compact/*` events, the summary blocks, and the shadowed range/seqs plus the estimated token count. What a successful compaction returns to its caller: the shadowed range and seqs plus the estimated token count. The durable `compact/summary` event owns the raw summary and bookkeeping-event identity.
```ts type-equiv ```ts type-equiv
interface CompactionResult { interface CompactionResult {
/** The seq of the appended `compact/start` event. */
startSeq: number
/** The seq of the appended `compact/summary` event. */
summarySeq: number
/** The seq of the appended `compact/end` event. */
endSeq: number
/** The summary content blocks produced by the backend. */
summary: ContentBlock[]
/** /**
* The surface-boundary pair that was shadowed: the seqs of the first * The surface-boundary pair that was shadowed: the seqs of the first
* (`start`) and last (`end`) surface nodes of the replaced range. A * (`start`) and last (`end`) surface nodes of the replaced range. A
@@ -50,6 +42,6 @@ interface CompactionResult {
## The service ## The service
`CompactService` (`ctx.compact`, abstract — defined in [`packages/compact/compact/src/index.ts`](../../packages/compact/compact/src/index.ts)) declares two abstract methods: `compactIfNeeded(agent, fullSystemPrompt, sessionPrefix, signal)` checks token pressure and compacts an older range if the history is too large (returning `null` when nothing needs it), and `compactRegion(session, start, end, agent, signal?)` forcibly summarizes surface nodes `[start, end]` into a single replacement node. `compactIfNeeded`'s parameters are all required — the loop's `agent/pre-step` checkpoint supplies the agent, the assembled `fullSystemPrompt`, the instance's composed `sessionPrefix` (request-only messages the derived history omits, so the pressure estimate must count them), and the turn `signal`. A backend summarizing via `ctx.llm.stream()` must forward `signal` into the call's `GenerateOptions.signal`, so an abort or dispose tears down the in-flight summarization. The entire strategy — token estimation, retention policy, event sequencing, summarization — is a HOW decision owned by the implementation. `CompactService` (`ctx.compact`, abstract — defined in [`packages/compact/compact/src/index.ts`](../../packages/compact/compact/src/index.ts)) declares two abstract methods: `compactIfNeeded(agent, fullSystemPrompt, sessionPrefix, signal)` checks token pressure and compacts an older range if the history is too large (returning `null` when nothing needs it), and `compactRegion(start, end, agent, signal?)` forcibly summarizes surface nodes `[start, end]` from `agent.session` into a single replacement node. `compactIfNeeded`'s parameters are all required — the loop's `agent/pre-step` checkpoint supplies the agent, the assembled `fullSystemPrompt`, the instance's composed `sessionPrefix` (request-only messages the derived history omits, so the pressure estimate must count them), and the turn `signal`. A backend summarizing via `ctx.llm.stream()` must forward `signal` into the call's `GenerateOptions.signal`, so an abort or dispose tears down the in-flight summarization. The entire strategy — token estimation, retention policy, event sequencing, summarization — is a HOW decision owned by the implementation.
Auto-compaction runs on the serial `agent/pre-step` loop seam (fired once per step, after `turn/start` and BEFORE the step opens and its request history is derived), not the `agent/request` waterfall: compaction mutates the session surface in place — with its log-only `compact/*` records landing cleanly outside any step — and the loop derives the request from the already-compacted surface. Retention is turn-agnostic — the only structural guard is tool-pairing balance (a compacted region's edges are balanced cuts on the surface, so it never splits a step's tool-calls from their results), so a single runaway turn that alone exceeds the window compacts its own early closed steps rather than being retained verbatim. The backend that ships this (`dsh-compact-basic`) documents the retention walk, summary shrink validation, bounded re-compaction, and the crash/recoverable failure taxonomy. Auto-compaction runs on the serial `agent/pre-step` loop seam (fired once per step, after `turn/start` and BEFORE the step opens and its request history is derived), not the `agent/request` waterfall: compaction mutates the session surface in place — with its log-only `compact/*` records landing cleanly outside any step — and the loop derives the request from the already-compacted surface. Retention is turn-agnostic — the only structural guard is tool-pairing balance (a compacted region's edges are balanced cuts on the surface, so it never splits a step's tool-calls from their results), so a single runaway turn that alone exceeds the window compacts its own early closed steps rather than being retained verbatim. The backend that ships this (`dsh-compact-basic`) documents the retention walk, summary shrink validation, bounded re-compaction, and the crash/recoverable failure taxonomy.

View File

@@ -22,7 +22,7 @@ Per the [capability-seams RFC](../../implemented/architecture/2026-06-13-capabil
### The contract depends on `dsh-session` and `dsh-llm` — a deliberate deviation ### The contract depends on `dsh-session` and `dsh-llm` — a deliberate deviation
The capability-seams RFC states the interface package "depends only on cordis" (true of `dsh-bash`, whose vocabulary is self-contained). Compaction **cannot** honor that: its verbs are defined *over* a `Session` (`compactRegion(session, start, end)`) and its output *is* the content vocabulary (`CompactionResult.summary: ContentBlock[]`). There is no way to express the contract without naming `Session`/`SessionEvent` (from `dsh-session`) and `ContentBlock` (from `dsh-llm`). The capability-seams RFC states the interface package "depends only on cordis" (true of `dsh-bash`, whose vocabulary is self-contained). Compaction **cannot** honor that: its verbs act on an agent-owned `Session` (`compactRegion(start, end, agent)`) and the durable `compact/summary` event carries `ContentBlock[]`. There is no way to express the contract without naming `Session`/`SessionEvent` (from `dsh-session`) and `ContentBlock` (from `dsh-llm`).
This is not a coupling smell — it is the contract's domain. The "only cordis" guidance was always shorthand for "the interface depends only on what the contract genuinely names, and never on an implementation." `dsh-session` and `dsh-llm` are themselves interface/vocabulary packages, not implementations; `dsh-compact` still imports no backend. The seam's real invariant — *consumers and implementations evolve independently behind an abstract service* — holds intact. This is not a coupling smell — it is the contract's domain. The "only cordis" guidance was always shorthand for "the interface depends only on what the contract genuinely names, and never on an implementation." `dsh-session` and `dsh-llm` are themselves interface/vocabulary packages, not implementations; `dsh-compact` still imports no backend. The seam's real invariant — *consumers and implementations evolve independently behind an abstract service* — holds intact.
@@ -30,7 +30,7 @@ This is not a coupling smell — it is the contract's domain. The "only cordis"
An earlier draft put the full algorithm (the retention walk, token-summing, text extraction) as concrete methods on the interface, with only `estimateContentTokens()` and `summarize()` abstract. That recouples the contract to one strategy: a backend that wants a different retention policy or a different event-sequencing would have to fight inherited concrete code. Making both core methods abstract puts every *how* decision in the backend, where it belongs, and keeps the interface a pure statement of *what*. The backend remains internally factored — `estimateContentTokens()` and `summarize()` are `protected` hooks a sub-backend can override without reimplementing the walk — but that factoring is the backend's private concern, not the contract's. An earlier draft put the full algorithm (the retention walk, token-summing, text extraction) as concrete methods on the interface, with only `estimateContentTokens()` and `summarize()` abstract. That recouples the contract to one strategy: a backend that wants a different retention policy or a different event-sequencing would have to fight inherited concrete code. Making both core methods abstract puts every *how* decision in the backend, where it belongs, and keeps the interface a pure statement of *what*. The backend remains internally factored — `estimateContentTokens()` and `summarize()` are `protected` hooks a sub-backend can override without reimplementing the walk — but that factoring is the backend's private concern, not the contract's.
`compactIfNeeded(agent, turn, step, fullSystemPrompt, signal)` takes **required** parameters (not the original all-optional shape). The auto-compaction seam (below) always supplies the agent, lifecycle context, assembled system prompt (counted toward the estimate), and the turn's abort signal, so optionality would only invite a hidden default at the seam. The session being compacted comes from the agent context. `compactRegion(session, start, end, agent, turn, step, signal?)` keeps an optional signal (a manual caller may omit it). Passing lifecycle context rather than a concrete model keeps router agents honest: the backend's summarization request can run through `agent/request`, where model-routing plugins already choose the actual model. `compactIfNeeded(agent, fullSystemPrompt, sessionPrefix, signal)` takes required inputs from the auto-compaction seam: the agent, assembled system prompt, composed request prefix, and turn abort signal. `compactRegion(start, end, agent, signal?)` uses `agent.session` as its single session identity and keeps an optional signal for manual callers. The backend's summarization request is a direct `ctx.llm.stream()` call; the configured summarization model falls back to the agent's model, and adapters can still route through the LLM seam.
### Auto-compaction runs on `agent/pre-step`, a dedicated surface-mutation seam ### Auto-compaction runs on `agent/pre-step`, a dedicated surface-mutation seam

View File

@@ -17,7 +17,7 @@ The abstract contract states only WHAT compaction does; this backend owns every
- **Auto-compaction** — an `agent/pre-step` listener delegates to `compactIfNeeded()` before every step (not just a turn's first — a tool-heavy turn grows the surface mid-turn, so a runaway turn still compacts, and per-step firing is the only moment to rescue it before overflow). `agent/pre-step` is a serial (awaited, in-order) surface-mutation checkpoint that fires after `turn/start` and BEFORE the step opens (`step/start`) and its request history is derived, so compaction mutates the surface — with its log-only `compact/*` records landing cleanly outside any step — and the loop derives once from the result: no double-derive, and the listener cannot see (or need to rewrite) an already-assembled `messages` array. The listener owns no threshold logic of its own (the single token-pressure check lives in `compactIfNeeded()`); because Cordis `serial` bails early on non-void return values, the listener returns `void` and does not use the dispatcher's bail channel as a veto surface. - **Auto-compaction** — an `agent/pre-step` listener delegates to `compactIfNeeded()` before every step (not just a turn's first — a tool-heavy turn grows the surface mid-turn, so a runaway turn still compacts, and per-step firing is the only moment to rescue it before overflow). `agent/pre-step` is a serial (awaited, in-order) surface-mutation checkpoint that fires after `turn/start` and BEFORE the step opens (`step/start`) and its request history is derived, so compaction mutates the surface — with its log-only `compact/*` records landing cleanly outside any step — and the loop derives once from the result: no double-derive, and the listener cannot see (or need to rewrite) an already-assembled `messages` array. The listener owns no threshold logic of its own (the single token-pressure check lives in `compactIfNeeded()`); because Cordis `serial` bails early on non-void return values, the listener returns `void` and does not use the dispatcher's bail channel as a veto surface.
- **Failure handling** — the `compact/start … compact/end` bracket is a log-recorded lock: it makes a crash mid-summarization a detectable orphan (a `compact/start` with no `compact/end`), records provenance, and prevents a concurrent compaction. Two failure paths: a **crash** (the loop dies mid-summarization) leaves a dangling `compact/start` that is inert — `compact/*` events are log-only, the surface replacement never landed, so the full history derives fine and generic turn-repair closes the turn; a **recoverable** failure (summarization throws but the loop survives) appends `compact/end` with its `error` field set, leaving the surface untouched so the call proceeds with full history. Core session repair stays compaction-agnostic by design — it never learns about `compact/*`. - **Failure handling** — the `compact/start … compact/end` bracket is a log-recorded lock: it makes a crash mid-summarization a detectable orphan (a `compact/start` with no `compact/end`), records provenance, and prevents a concurrent compaction. Two failure paths: a **crash** (the loop dies mid-summarization) leaves a dangling `compact/start` that is inert — `compact/*` events are log-only, the surface replacement never landed, so the full history derives fine and generic turn-repair closes the turn; a **recoverable** failure (summarization throws but the loop survives) appends `compact/end` with its `error` field set, leaving the surface untouched so the call proceeds with full history. Core session repair stays compaction-agnostic by design — it never learns about `compact/*`.
`estimateContentTokens()` and `summarize()` are overridable hooks: a tokenizer-based or template-based backend can subclass `BasicCompactService` and override just those, reusing the retention walk and surface plumbing. `summarize()` returns the summary blocks together with the call envelope it actually used (`{ summary, model, maxTokens? }`) — the caller logs that envelope on the `compact/summary` provenance event, so an overriding backend reports its own envelope honestly. The protected `estimateContentTokens()` and `summarize()` methods are overridable hooks: a tokenizer-based or template-based backend can subclass `BasicCompactService` and override just those, reusing the private retention/pressure accounting and surface plumbing. `summarize()` returns the summary blocks together with the call envelope it actually used (`{ summary, model, maxTokens? }`) — the caller logs that envelope on the `compact/summary` provenance event, so an overriding backend reports its own envelope honestly.
## Config (`BasicCompactConfig`) ## Config (`BasicCompactConfig`)

View File

@@ -41,12 +41,11 @@ import type { BasicCompactConfig, ResolvedConfig } from './types.ts'
import { resolveConfig } from './types.ts' import { resolveConfig } from './types.ts'
export type { BasicCompactConfig, ResolvedConfig } from './types.ts' export type { BasicCompactConfig, ResolvedConfig } from './types.ts'
export { resolveConfig } from './types.ts'
/** Per-block structural overhead for JSON framing / type tag. */ /** Per-block structural overhead for JSON framing / type tag. */
const BLOCK_OVERHEAD = 4 const BLOCK_OVERHEAD = 4
/** Role-field framing overhead added per message in {@link BasicCompactService.estimateTokens}. */ /** Role-field framing overhead added per message in the request estimator. */
const ROLE_OVERHEAD = 4 const ROLE_OVERHEAD = 4
/** Tags wrapping the structured summary inside the landed checkpoint node. */ /** Tags wrapping the structured summary inside the landed checkpoint node. */
@@ -222,7 +221,7 @@ export class BasicCompactService extends CompactService {
* their JSON-stringified length. * their JSON-stringified length.
* @returns the estimated token count. * @returns the estimated token count.
*/ */
estimateContentTokens(blocks: readonly ContentBlock[]): number { protected estimateContentTokens(blocks: readonly ContentBlock[]): number {
const { charsPerToken } = this.config const { charsPerToken } = this.config
let tokens = 0 let tokens = 0
for (const block of blocks) { for (const block of blocks) {
@@ -257,7 +256,8 @@ export class BasicCompactService extends CompactService {
* @returns the estimated token count of the event's content, or 0 for a * @returns the estimated token count of the event's content, or 0 for a
* non-message event. * non-message event.
*/ */
estimateEventTokens(event: SessionEvent): number { private estimateEventTokens(event: SessionEvent): number {
/* v8 ignore next -- callers traverse surface nodes, whose event types are the five cases below */
switch (event.type) { switch (event.type) {
case 'user/message': case 'user/message':
case 'assistant/message': case 'assistant/message':
@@ -278,7 +278,7 @@ export class BasicCompactService extends CompactService {
* @param systemPrompt - counted at chars / `charsPerToken` when provided. * @param systemPrompt - counted at chars / `charsPerToken` when provided.
* @returns the estimated token footprint of the whole request. * @returns the estimated token footprint of the whole request.
*/ */
estimateTokens(messages: readonly Message[], systemPrompt?: string): number { private estimateTokens(messages: readonly Message[], systemPrompt?: string): number {
let total = 0 let total = 0
for (const msg of messages) { for (const msg of messages) {
total += this.estimateContentTokens(msg.content) total += this.estimateContentTokens(msg.content)
@@ -318,7 +318,7 @@ export class BasicCompactService extends CompactService {
* @returns the text-only summary blocks plus the call envelope used * @returns the text-only summary blocks plus the call envelope used
* (`model`, and `maxTokens` when the summarizer has a cap). * (`model`, and `maxTokens` when the summarizer has a cap).
*/ */
async summarize( protected async summarize(
text: string, agent: Agent, signal?: AbortSignal, text: string, agent: Agent, signal?: AbortSignal,
): Promise<{ summary: ContentBlock[]; model: string; maxTokens?: number }> { ): Promise<{ summary: ContentBlock[]; model: string; maxTokens?: number }> {
const assembler = new BlockAssembler() const assembler = new BlockAssembler()
@@ -417,7 +417,7 @@ export class BasicCompactService extends CompactService {
break break
} }
result = await this.compactRegion(session, range.start, range.end, agent, signal) result = await this.compactRegion(range.start, range.end, agent, signal)
} }
const totalTokens = this.estimatePressure(session, fullSystemPrompt, sessionPrefix) const totalTokens = this.estimatePressure(session, fullSystemPrompt, sessionPrefix)
@@ -439,17 +439,17 @@ export class BasicCompactService extends CompactService {
* @param sessionPrefix - the instance's composed session prefix (counts toward pressure). * @param sessionPrefix - the instance's composed session prefix (counts toward pressure).
* @returns the estimated token total the next request will carry. * @returns the estimated token total the next request will carry.
*/ */
estimatePressure(session: Session, fullSystemPrompt: string, sessionPrefix: readonly Message[]): number { private estimatePressure(session: Session, fullSystemPrompt: string, sessionPrefix: readonly Message[]): number {
return this.estimateTokens([...sessionPrefix, ...session.deriveMessages()], fullSystemPrompt) return this.estimateTokens([...sessionPrefix, ...session.deriveMessages()], fullSystemPrompt)
} }
override async compactRegion( override async compactRegion(
session: Session,
start: number, start: number,
end: number, end: number,
agent: Agent, agent: Agent,
signal?: AbortSignal, signal?: AbortSignal,
): Promise<CompactionResult> { ): Promise<CompactionResult> {
const session = agent.session
// Resolve the range by surface POSITION, not numeric seq interval. A prior // Resolve the range by surface POSITION, not numeric seq interval. A prior
// replace lands a fresh high-seq summary node AT the shadowed range's // replace lands a fresh high-seq summary node AT the shadowed range's
// position, so the surface order (head→tail) no longer tracks seq order — // position, so the surface order (head→tail) no longer tracks seq order —
@@ -556,13 +556,9 @@ export class BasicCompactService extends CompactService {
// compact/start and here leaves a detectable orphaned lock (a compact/start // compact/start and here leaves a detectable orphaned lock (a compact/start
// with no matching compact/end) rather than a compact/end that falsely // with no matching compact/end) rather than a compact/end that falsely
// claims compaction finished before the surface replacement landed. // claims compaction finished before the surface replacement landed.
const endEvent = session.append('compact/end', { turn: openTurn }) session.append('compact/end', { turn: openTurn })
return { return {
startSeq: startEvent.seq,
summarySeq: summaryEvent.seq,
endSeq: endEvent.seq,
summary,
shadowedRange: { start, end }, shadowedRange: { start, end },
shadowedSeqs, shadowedSeqs,
shadowedTokenCount, shadowedTokenCount,

View File

@@ -5,7 +5,7 @@ import type { BasicCompactConfig } from '@deepseek-ai/dsh-compact-basic'
import type { ContentBlock, GenerateOptions, Message, StreamChunk } from '@deepseek-ai/dsh-llm' import type { ContentBlock, GenerateOptions, Message, StreamChunk } from '@deepseek-ai/dsh-llm'
import { CallId, LlmAdapter, LlmService } from '@deepseek-ai/dsh-llm' import { CallId, LlmAdapter, LlmService } from '@deepseek-ai/dsh-llm'
import SessionStore, { Session, SessionId } from '@deepseek-ai/dsh-session' import SessionStore, { Session, SessionId } from '@deepseek-ai/dsh-session'
import type { SessionEvent, SurfaceEvent } from '@deepseek-ai/dsh-session' import type { SurfaceEvent } from '@deepseek-ai/dsh-session'
import * as Invariants from '@deepseek-ai/dsh-invariants' import * as Invariants from '@deepseek-ai/dsh-invariants'
import type { Agent } from '@deepseek-ai/dsh-agent' import type { Agent } from '@deepseek-ai/dsh-agent'
@@ -68,6 +68,20 @@ class TestCompactService extends BasicCompactService {
} }
} }
/** Expose the backend's protected extension hooks for their focused contract tests. */
class InspectableCompactService extends BasicCompactService {
estimateContent(blocks: readonly ContentBlock[]): number {
return this.estimateContentTokens(blocks)
}
summarizeForTest(
text: string,
agent: Agent,
): Promise<{ summary: ContentBlock[]; model: string; maxTokens?: number }> {
return this.summarize(text, agent)
}
}
function isFramedCheckpoint(blocks: readonly ContentBlock[]): boolean { function isFramedCheckpoint(blocks: readonly ContentBlock[]): boolean {
const first = blocks[0] const first = blocks[0]
const last = blocks[blocks.length - 1] const last = blocks[blocks.length - 1]
@@ -333,51 +347,6 @@ describe('BasicCompactService step-alignment (never split a tool-call/result pai
}) })
}) })
describe('BasicCompactService.estimateEventTokens', () => {
it('returns 0 for non-message events (boundary, chunk, step/end, tool/call)', () => {
const svc = createTestService()
expect(svc.estimateEventTokens({ type: 'turn/start', seq: 0, time: 1, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } })).toBe(0)
expect(svc.estimateEventTokens({ type: 'step/start', seq: 1, time: 2, data: { turn: 1, step: 1 } })).toBe(0)
expect(svc.estimateEventTokens({ type: 'assistant/chunk', seq: 2, time: 3, data: { turn: 1, step: 1, chunk: { type: 'text-delta', index: 0, text: 'h' } } })).toBe(0)
expect(svc.estimateEventTokens({ type: 'step/end', seq: 3, time: 4, data: { turn: 1, step: 1 } })).toBe(0)
expect(svc.estimateEventTokens({ type: 'tool/call', seq: 4, time: 5, data: { turn: 1, step: 1, callId: CallId('c1'), name: 'read', arguments: '{}' } })).toBe(0)
})
it('returns estimate for message-producing events', () => {
const svc = createTestService()
const userEvent: SessionEvent = { type: 'user/message', seq: 0, time: 1, data: { content: [{ type: 'text', text: 'hello' }], source: { kind: 'user' } } }
expect(svc.estimateEventTokens(userEvent)).toBe(10)
const asstEvent: SessionEvent = { type: 'assistant/message', seq: 1, time: 2, data: { turn: 1, step: 1, content: [{ type: 'text', text: 'a' }, { type: 'text', text: 'b' }] } }
expect(svc.estimateEventTokens(asstEvent)).toBe(20)
const toolEvent: SessionEvent = { type: 'tool/result', seq: 2, time: 3, data: { turn: 1, step: 1, callId: CallId('c1'), content: [{ type: 'text', text: 'output' }], isError: false } }
expect(svc.estimateEventTokens(toolEvent)).toBe(10)
})
})
describe('BasicCompactService.estimateTokens', () => {
it('sums token estimates across messages', () => {
const svc = createTestService()
const messages: Message[] = [
{ role: 'user', content: [{ type: 'text', text: 'hello' }] },
{ role: 'assistant', content: [{ type: 'text', text: 'hi' }, { type: 'text', text: 'there' }] },
]
// 1 block * 10 + 4 (role) + 2 blocks * 10 + 4 (role) = 10 + 4 + 20 + 4 = 38
expect(svc.estimateTokens(messages)).toBe(38)
})
it('includes system prompt in the estimate', () => {
const svc = createTestService()
const messages: Message[] = [
{ role: 'user', content: [{ type: 'text', text: 'hi' }] },
]
const systemPrompt = 'You are a helpful assistant.'
// 1 block * 10 + 4 (role) + ceil(28/4) = 10 + 4 + 7 = 21
expect(svc.estimateTokens(messages, systemPrompt)).toBe(21)
})
})
describe('BasicCompactService.compactRegion', () => { describe('BasicCompactService.compactRegion', () => {
it('shadows surface nodes and inserts a summary via user/message', async () => { it('shadows surface nodes and inserts a summary via user/message', async () => {
const svc = createTestService() const svc = createTestService()
@@ -393,7 +362,7 @@ describe('BasicCompactService.compactRegion', () => {
expect(result.shadowedSeqs).toEqual([firstSeq, secondSeq]) expect(result.shadowedSeqs).toEqual([firstSeq, secondSeq])
expect(result.shadowedRange.start).toBe(firstSeq) expect(result.shadowedRange.start).toBe(firstSeq)
expect(result.shadowedRange.end).toBe(secondSeq) expect(result.shadowedRange.end).toBe(secondSeq)
expect(result.summary).toEqual(svc.mockSummary) expect(result.shadowedTokenCount).toBe(20)
const events = session.events const events = session.events
const startEvent = events.findLast(e => e.type === 'compact/start') const startEvent = events.findLast(e => e.type === 'compact/start')
@@ -405,6 +374,7 @@ describe('BasicCompactService.compactRegion', () => {
// The provenance record carries the summarize call's envelope, so "which // The provenance record carries the summarize call's envelope, so "which
// model wrote this summary" is answerable from the log alone. // model wrote this summary" is answerable from the log alone.
expect(summaryEvent?.type === 'compact/summary' && summaryEvent.data.model).toBe('test-model') expect(summaryEvent?.type === 'compact/summary' && summaryEvent.data.model).toBe('test-model')
expect(summaryEvent?.type === 'compact/summary' && summaryEvent.data.summary).toEqual(svc.mockSummary)
// compact/* events are log-only — no surfaceOp (type system enforces this). // compact/* events are log-only — no surfaceOp (type system enforces this).
const startRaw = startEvent as unknown as { surfaceOp?: unknown } const startRaw = startEvent as unknown as { surfaceOp?: unknown }
@@ -509,10 +479,9 @@ describe('BasicCompactService.compactRegion', () => {
const session = multiTurnSession(3, 1) const session = multiTurnSession(3, 1)
const nodes = session.surface.nodes const nodes = session.surface.nodes
const result = await compactRegion(svc, session, nodes[0]!, nodes[1]!, 'm') await compactRegion(svc, session, nodes[0]!, nodes[1]!, 'm')
// Provenance (compact/summary) carries the RAW, unframed summary. // Provenance (compact/summary) carries the RAW, unframed summary.
expect(result.summary).toEqual([{ type: 'text', text: 'STRUCTURED SUMMARY' }])
const summaryEvent = session.events.findLast(e => e.type === 'compact/summary')! const summaryEvent = session.events.findLast(e => e.type === 'compact/summary')!
expect(summaryEvent.data).toMatchObject({ summary: [{ type: 'text', text: 'STRUCTURED SUMMARY' }] }) expect(summaryEvent.data).toMatchObject({ summary: [{ type: 'text', text: 'STRUCTURED SUMMARY' }] })
@@ -590,7 +559,6 @@ describe('BasicCompactService.compactIfNeeded', () => {
expect(result).not.toBeNull() expect(result).not.toBeNull()
expect(session.events.filter(e => e.type === 'compact/summary')).toHaveLength(1) expect(session.events.filter(e => e.type === 'compact/summary')).toHaveLength(1)
expect(svc.estimateTokens(session.deriveMessages(), '')).toBeLessThan(70)
}) })
it('walks tail→head and retains nodes within token budget', async () => { it('walks tail→head and retains nodes within token budget', async () => {
@@ -716,7 +684,6 @@ describe('BasicCompactService.compactIfNeeded', () => {
expect(result).not.toBeNull() expect(result).not.toBeNull()
expect(svc.summarizeCalls).toHaveLength(2) expect(svc.summarizeCalls).toHaveLength(2)
expect(session.events.filter(e => e.type === 'compact/summary')).toHaveLength(2) expect(session.events.filter(e => e.type === 'compact/summary')).toHaveLength(2)
expect(svc.estimateTokens(session.deriveMessages(), '')).toBeLessThan(50)
}) })
it('throws after the configured re-compaction attempts still leave the surface above threshold', async () => { it('throws after the configured re-compaction attempts still leave the surface above threshold', async () => {
@@ -801,53 +768,51 @@ describe('BasicCompactService blocking (compaction in progress)', () => {
describe('BasicCompactService token estimation (char/4 heuristic)', () => { describe('BasicCompactService token estimation (char/4 heuristic)', () => {
it('estimates text blocks with char/4 + overhead', () => { it('estimates text blocks with char/4 + overhead', () => {
const svc = new BasicCompactService(new Context(), cfg({ auto: false })) const svc = new InspectableCompactService(new Context(), cfg({ auto: false }))
// 'this is a somewhat longer text block' = 36 → ceil(36/4)+4 = 13; 'short' = 5 → 2+4 = 6 // 'this is a somewhat longer text block' = 36 → ceil(36/4)+4 = 13; 'short' = 5 → 2+4 = 6
const blocks: ContentBlock[] = [ const blocks: ContentBlock[] = [
{ type: 'text', text: 'this is a somewhat longer text block' }, { type: 'text', text: 'this is a somewhat longer text block' },
{ type: 'text', text: 'short' }, { type: 'text', text: 'short' },
] ]
expect(svc.estimateContentTokens(blocks)).toBe(19) expect(svc.estimateContent(blocks)).toBe(19)
}) })
it('estimates reasoning blocks same as text', () => { it('estimates reasoning blocks same as text', () => {
const svc = new BasicCompactService(new Context(), cfg({ auto: false })) const svc = new InspectableCompactService(new Context(), cfg({ auto: false }))
// 'thinking about this...' = 22 → ceil(22/4)+4 = 10 // 'thinking about this...' = 22 → ceil(22/4)+4 = 10
expect(svc.estimateContentTokens([{ type: 'reasoning', text: 'thinking about this...' }])).toBe(10) expect(svc.estimateContent([{ type: 'reasoning', text: 'thinking about this...' }])).toBe(10)
}) })
it('estimates tool-call blocks from name + arguments', () => { it('estimates tool-call blocks from name + arguments', () => {
const svc = new BasicCompactService(new Context(), cfg({ auto: false })) const svc = new InspectableCompactService(new Context(), cfg({ auto: false }))
// 'bash' = 4 → 1; '{"command":"ls"}' = 16 → 4; + 4 overhead = 9 // 'bash' = 4 → 1; '{"command":"ls"}' = 16 → 4; + 4 overhead = 9
expect(svc.estimateContentTokens([ expect(svc.estimateContent([
{ type: 'tool-call', id: CallId('c1'), name: 'bash', arguments: '{"command":"ls"}' }, { type: 'tool-call', id: CallId('c1'), name: 'bash', arguments: '{"command":"ls"}' },
])).toBe(9) ])).toBe(9)
}) })
it('estimates tool-result blocks recursively', () => { it('estimates tool-result blocks recursively', () => {
const svc = new BasicCompactService(new Context(), cfg({ auto: false })) const svc = new InspectableCompactService(new Context(), cfg({ auto: false }))
// inner text 5 → 2+4 = 6; outer 6 + 4 overhead = 10 // inner text 5 → 2+4 = 6; outer 6 + 4 overhead = 10
expect(svc.estimateContentTokens([ expect(svc.estimateContent([
{ type: 'tool-result', toolCallId: CallId('c1'), content: [{ type: 'text', text: 'hello' }], isError: false }, { type: 'tool-result', toolCallId: CallId('c1'), content: [{ type: 'text', text: 'hello' }], isError: false },
])).toBe(10) ])).toBe(10)
}) })
it('returns 0 for empty content blocks', () => { it('returns 0 for empty content blocks', () => {
const svc = new BasicCompactService(new Context(), cfg({ auto: false })) const svc = new InspectableCompactService(new Context(), cfg({ auto: false }))
expect(svc.estimateContentTokens([])).toBe(0) expect(svc.estimateContent([])).toBe(0)
}) })
it('honors a configured charsPerToken (fractional densities included)', () => { it('honors a configured charsPerToken (fractional densities included)', () => {
// 'this is a somewhat longer text block' = 36 chars. // 'this is a somewhat longer text block' = 36 chars.
const blocks: ContentBlock[] = [{ type: 'text', text: 'this is a somewhat longer text block' }] const blocks: ContentBlock[] = [{ type: 'text', text: 'this is a somewhat longer text block' }]
// charsPerToken 2: ceil(36/2)+4 = 22 — a CJK-density config doubles the estimate. // charsPerToken 2: ceil(36/2)+4 = 22 — a CJK-density config doubles the estimate.
const dense = new BasicCompactService(new Context(), cfg({ auto: false, charsPerToken: 2 })) const dense = new InspectableCompactService(new Context(), cfg({ auto: false, charsPerToken: 2 }))
expect(dense.estimateContentTokens(blocks)).toBe(22) expect(dense.estimateContent(blocks)).toBe(22)
// Fractional density is legal: ceil(36/1.5)+4 = 28. // Fractional density is legal: ceil(36/1.5)+4 = 28.
const fractional = new BasicCompactService(new Context(), cfg({ auto: false, charsPerToken: 1.5 })) const fractional = new InspectableCompactService(new Context(), cfg({ auto: false, charsPerToken: 1.5 }))
expect(fractional.estimateContentTokens(blocks)).toBe(28) expect(fractional.estimateContent(blocks)).toBe(28)
// The system-prompt term scales with the same knob: 36-char prompt at density 2 → ceil(36/2) = 18.
expect(dense.estimateTokens([], 'this is a somewhat longer text block')).toBe(18)
}) })
}) })
@@ -1013,17 +978,17 @@ function compactRegion(
model: string, model: string,
signal?: AbortSignal, signal?: AbortSignal,
) { ) {
return svc.compactRegion(session, start, end, stubAgent(session, model), signal) return svc.compactRegion(start, end, stubAgent(session, model), signal)
} }
function summarize(svc: BasicCompactService, text: string, model: string) { function summarize(svc: InspectableCompactService, text: string, model: string) {
return svc.summarize(text, stubAgent(new Session(SessionId('summary')), model)) return svc.summarizeForTest(text, stubAgent(new Session(SessionId('summary')), model))
} }
describe('BasicCompactService.summarize (real ctx.llm.stream)', () => { describe('BasicCompactService.summarize (real ctx.llm.stream)', () => {
it('summarizes via the registered adapter and returns its content', async () => { it('summarizes via the registered adapter and returns its content', async () => {
const { ctx, adapter } = await ctxWithModel('SUMMARY TEXT') const { ctx, adapter } = await ctxWithModel('SUMMARY TEXT')
const svc = new BasicCompactService(ctx, cfg({ auto: false, maxTokens: 512 })) const svc = new InspectableCompactService(ctx, cfg({ auto: false, maxTokens: 512 }))
const { summary, model, maxTokens } = await summarize(svc, 'User: hi\n\nAssistant: hello', 'test-model') const { summary, model, maxTokens } = await summarize(svc, 'User: hi\n\nAssistant: hello', 'test-model')
expect(summary).toEqual([{ type: 'text', text: 'SUMMARY TEXT' }]) expect(summary).toEqual([{ type: 'text', text: 'SUMMARY TEXT' }])
@@ -1041,7 +1006,7 @@ describe('BasicCompactService.summarize (real ctx.llm.stream)', () => {
it('uses maxTokens as the summarization provider cap', async () => { it('uses maxTokens as the summarization provider cap', async () => {
const { ctx, adapter } = await ctxWithModel('SUMMARY TEXT') const { ctx, adapter } = await ctxWithModel('SUMMARY TEXT')
const svc = new BasicCompactService(ctx, cfg({ const svc = new InspectableCompactService(ctx, cfg({
auto: false, auto: false,
maxTokens: 50, maxTokens: 50,
})) }))
@@ -1059,7 +1024,7 @@ describe('BasicCompactService.summarize (real ctx.llm.stream)', () => {
// synthesized user/message summary as an orphaned call. // synthesized user/message summary as an orphaned call.
{ type: 'tool-call', id: CallId('c1'), name: 'bash', arguments: '{}' }, { type: 'tool-call', id: CallId('c1'), name: 'bash', arguments: '{}' },
]) ])
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
const { summary } = await summarize(svc, 'User: hi', 'test-model') const { summary } = await summarize(svc, 'User: hi', 'test-model')
@@ -1068,26 +1033,26 @@ describe('BasicCompactService.summarize (real ctx.llm.stream)', () => {
it('throws when no text block remains after filtering', async () => { it('throws when no text block remains after filtering', async () => {
const { ctx } = await ctxWithBlocks([{ type: 'reasoning', text: 'private only' }]) const { ctx } = await ctxWithBlocks([{ type: 'reasoning', text: 'private only' }])
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
await expect(summarize(svc, 'User: hi', 'test-model')).rejects.toThrow(/no text summary content/) await expect(summarize(svc, 'User: hi', 'test-model')).rejects.toThrow(/no text summary content/)
}) })
it('throws when no model is provided', async () => { it('throws when no model is provided', async () => {
const { ctx } = await ctxWithModel('x') const { ctx } = await ctxWithModel('x')
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
await expect(summarize(svc, 'text', '')).rejects.toThrow(/no model available/) await expect(summarize(svc, 'text', '')).rejects.toThrow(/no model available/)
}) })
it('rethrows when the stream ends with a finish-error chunk', async () => { it('rethrows when the stream ends with a finish-error chunk', async () => {
const ctx = await ctxWithFinish({ kind: 'error', message: 'provider 401', code: 'UNAUTHORIZED' }) const ctx = await ctxWithFinish({ kind: 'error', message: 'provider 401', code: 'UNAUTHORIZED' })
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
await expect(summarize(svc, 'text', 'test-model')).rejects.toMatchObject({ message: 'provider 401', code: 'UNAUTHORIZED' }) await expect(summarize(svc, 'text', 'test-model')).rejects.toMatchObject({ message: 'provider 401', code: 'UNAUTHORIZED' })
}) })
it('rethrows a finish-error chunk without a code (code stays undefined)', async () => { it('rethrows a finish-error chunk without a code (code stays undefined)', async () => {
const ctx = await ctxWithFinish({ kind: 'error', message: 'opaque failure' }) const ctx = await ctxWithFinish({ kind: 'error', message: 'opaque failure' })
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
const error = await summarize(svc, 'text', 'test-model').then(() => null, (e: unknown) => e as Error & { code?: string }) const error = await summarize(svc, 'text', 'test-model').then(() => null, (e: unknown) => e as Error & { code?: string })
expect(error?.message).toBe('opaque failure') expect(error?.message).toBe('opaque failure')
expect(error?.code).toBeUndefined() expect(error?.code).toBeUndefined()
@@ -1095,13 +1060,13 @@ describe('BasicCompactService.summarize (real ctx.llm.stream)', () => {
it('rethrows when the stream ends with a finish-aborted chunk', async () => { it('rethrows when the stream ends with a finish-aborted chunk', async () => {
const ctx = await ctxWithFinish({ kind: 'aborted' }) const ctx = await ctxWithFinish({ kind: 'aborted' })
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
await expect(summarize(svc, 'text', 'test-model')).rejects.toMatchObject({ message: 'summarization stream aborted', code: 'ABORTED' }) await expect(summarize(svc, 'text', 'test-model')).rejects.toMatchObject({ message: 'summarization stream aborted', code: 'ABORTED' })
}) })
it('fails closed on a max-tokens finish (an incomplete checkpoint must not commit)', async () => { it('fails closed on a max-tokens finish (an incomplete checkpoint must not commit)', async () => {
const ctx = await ctxWithFinish({ kind: 'max-tokens' }) const ctx = await ctxWithFinish({ kind: 'max-tokens' })
const svc = new BasicCompactService(ctx, cfg({ auto: false })) const svc = new InspectableCompactService(ctx, cfg({ auto: false }))
await expect(summarize(svc, 'text', 'test-model')).rejects.toMatchObject({ code: 'MAX_TOKENS' }) await expect(summarize(svc, 'text', 'test-model')).rejects.toMatchObject({ code: 'MAX_TOKENS' })
}) })
@@ -1129,8 +1094,9 @@ describe('BasicCompactService.summarize (real ctx.llm.stream)', () => {
const session = multiTurnSession(2, 1) const session = multiTurnSession(2, 1)
const nodes = session.surface.nodes const nodes = session.surface.nodes
const result = await compactRegion(svc, session, nodes[0]!, nodes[1]!, 'test-model') await compactRegion(svc, session, nodes[0]!, nodes[1]!, 'test-model')
expect(result.summary).toEqual([{ type: 'text', text: 'CONDENSED' }]) const summaryEvent = session.events.findLast(e => e.type === 'compact/summary')!
expect(summaryEvent.data.summary).toEqual([{ type: 'text', text: 'CONDENSED' }])
// The raw summary is wrapped in the checkpoint framing on the surface. // The raw summary is wrapped in the checkpoint framing on the surface.
expect(session.deriveMessages()[0]!.content).toContainEqual({ type: 'text', text: 'CONDENSED' }) expect(session.deriveMessages()[0]!.content).toContainEqual({ type: 'text', text: 'CONDENSED' })
}) })
@@ -1392,10 +1358,10 @@ describe('BasicCompactService edge cases', () => {
}) })
it('estimates unknown block types via JSON length (default branch)', () => { it('estimates unknown block types via JSON length (default branch)', () => {
const svc = new BasicCompactService(new Context(), cfg({ auto: false })) const svc = new InspectableCompactService(new Context(), cfg({ auto: false }))
// A block whose type is none of the known kinds — exercises the default arm. // A block whose type is none of the known kinds — exercises the default arm.
const unknown = { type: 'custom-widget', payload: 'some data' } as unknown as ContentBlock const unknown = { type: 'custom-widget', payload: 'some data' } as unknown as ContentBlock
expect(svc.estimateContentTokens([unknown])).toBeGreaterThan(0) expect(svc.estimateContent([unknown])).toBeGreaterThan(0)
}) })
it('auto-compaction reports bounded retry exhaustion after committing a smaller summary', async () => { it('auto-compaction reports bounded retry exhaustion after committing a smaller summary', async () => {
@@ -1604,14 +1570,15 @@ describe('BasicCompactService positional range (surface seqs are not monotonic a
// First compaction: shadow the two oldest surface nodes. // First compaction: shadow the two oldest surface nodes.
const nodes0 = session.surface.nodes const nodes0 = session.surface.nodes
const first = await compactRegion(svc, session, nodes0[0]!, nodes0[1]!, 'm') await compactRegion(svc, session, nodes0[0]!, nodes0[1]!, 'm')
const firstSummarySeq = session.events.findLast(e => e.type === 'compact/summary')!.seq
// The summary node now sits at the head with a seq HIGHER than the // The summary node now sits at the head with a seq HIGHER than the
// retained older nodes that follow it — the non-monotonic surface. (The // retained older nodes that follow it — the non-monotonic surface. (The
// head is the user/message replace node, appended after the compact/summary // head is the user/message replace node, appended after the compact/summary
// provenance event, so its seq is at least first.summarySeq.) // provenance event.
const nodes1 = session.surface.nodes const nodes1 = session.surface.nodes
expect(nodes1[0]!).toBeGreaterThanOrEqual(first.summarySeq) expect(nodes1[0]!).toBeGreaterThan(firstSummarySeq)
expect(nodes1[0]!).toBeGreaterThan(nodes1[1]!) expect(nodes1[0]!).toBeGreaterThan(nodes1[1]!)
// Second compaction: shadow [summary(head) … turn-2's step end]. The start // Second compaction: shadow [summary(head) … turn-2's step end]. The start
@@ -1623,13 +1590,14 @@ describe('BasicCompactService positional range (surface seqs are not monotonic a
const endSeq = nodes1[2]! const endSeq = nodes1[2]!
expect(startSeq).toBeGreaterThan(endSeq) expect(startSeq).toBeGreaterThan(endSeq)
const second = await compactRegion(svc, session, startSeq, endSeq, 'm') const second = await compactRegion(svc, session, startSeq, endSeq, 'm')
const secondSummarySeq = session.events.findLast(e => e.type === 'compact/summary')!.seq
// Exactly the three nodes at surface positions [0..2] are shadowed, in // Exactly the three nodes at surface positions [0..2] are shadowed, in
// surface order — the positional slice, regardless of their seq values. // surface order — the positional slice, regardless of their seq values.
expect(second.shadowedSeqs).toEqual([nodes1[0]!, nodes1[1]!, nodes1[2]!]) expect(second.shadowedSeqs).toEqual([nodes1[0]!, nodes1[1]!, nodes1[2]!])
// The surface still derives cleanly: a new head replace node + the rest. // The surface still derives cleanly: a new head replace node + the rest.
const finalNodes = session.surface.nodes const finalNodes = session.surface.nodes
expect(finalNodes[0]!).toBeGreaterThanOrEqual(second.summarySeq) expect(finalNodes[0]!).toBeGreaterThan(secondSummarySeq)
expect(session.deriveMessages().length).toBe(finalNodes.length) expect(session.deriveMessages().length).toBe(finalNodes.length)
}) })
@@ -1679,8 +1647,9 @@ describe('BasicCompactService llm inject (real plugin-load path)', () => {
const svc = ctx.compact as BasicCompactService const svc = ctx.compact as BasicCompactService
const session = multiTurnSession(2, 1) const session = multiTurnSession(2, 1)
const nodes = session.surface.nodes const nodes = session.surface.nodes
const result = await compactRegion(svc, session, nodes[0]!, nodes[1]!, 'test-model') await compactRegion(svc, session, nodes[0]!, nodes[1]!, 'test-model')
expect(result.summary).toEqual([{ type: 'text', text: 'CONDENSED' }]) const summaryEvent = session.events.findLast(e => e.type === 'compact/summary')!
expect(summaryEvent.data.summary).toEqual([{ type: 'text', text: 'CONDENSED' }])
// Tear the fiber down so this test owns no leaked registration; the // Tear the fiber down so this test owns no leaked registration; the
// dedicated cleanup assertion lives in the "HMR safety" suite. // dedicated cleanup assertion lives in the "HMR safety" suite.

View File

@@ -19,7 +19,7 @@ Both methods are **abstract** — the backend owns the entire strategy (token es
| Member | Semantics | | Member | Semantics |
|---|---| |---|---|
| `compactIfNeeded(agent, fullSystemPrompt, sessionPrefix, signal)` | Estimate the surface-derived history size; if over the backend's threshold, compact an older range via `compactRegion`, keeping recent context intact. Returns the `CompactionResult`, or `null` if nothing needed compacting. All parameters required — the loop's `agent/pre-step` checkpoint supplies the agent, assembled `fullSystemPrompt`, composed `sessionPrefix` (request-only messages every request carries but the derived history omits — the pressure estimate must count them), and turn `signal`. A backend's summarization request is a direct `ctx.llm.stream()` call (not a loop step), so per-call interception happens at `llm/stream`. | | `compactIfNeeded(agent, fullSystemPrompt, sessionPrefix, signal)` | Estimate the surface-derived history size; if over the backend's threshold, compact an older range via `compactRegion`, keeping recent context intact. Returns the `CompactionResult`, or `null` if nothing needed compacting. All parameters required — the loop's `agent/pre-step` checkpoint supplies the agent, assembled `fullSystemPrompt`, composed `sessionPrefix` (request-only messages every request carries but the derived history omits — the pressure estimate must count them), and turn `signal`. A backend's summarization request is a direct `ctx.llm.stream()` call (not a loop step), so per-call interception happens at `llm/stream`. |
| `compactRegion(session, start, end, agent, signal?)` | Forcibly summarize surface nodes `[start, end]` (inclusive seqs) into a single replacement node. **Throws** if a compaction is already in progress, if `start`/`end` aren't surface nodes, or if `start` is positioned after `end` on the surface. The range is a SURFACE-POSITION span, not a numeric seq interval — after a prior replace lands a fresh high-seq summary node at the shadowed range's position, surface order no longer tracks seq order. | | `compactRegion(start, end, agent, signal?)` | Forcibly summarize surface nodes `[start, end]` (inclusive seqs) from `agent.session` into a single replacement node. **Throws** if a compaction is already in progress, if `start`/`end` aren't surface nodes, or if `start` is positioned after `end` on the surface. The range is a SURFACE-POSITION span, not a numeric seq interval — after a prior replace lands a fresh high-seq summary node at the shadowed range's position, surface order no longer tracks seq order. |
`compactIfNeeded` takes a required `signal`; `compactRegion`'s is optional. A backend that summarizes via `ctx.llm.stream()` **must** forward it into the call's `GenerateOptions.signal`, so an abort or fiber dispose tears down the in-flight summarization instead of leaving an orphaned model call running past the cancellation. The session being compacted comes from the agent context; the turn that the `compact/*` events belong to is recoverable from the log (the currently-open turn), so the backend stamps it from the log rather than trusting a caller-supplied value. `compactIfNeeded` takes a required `signal`; `compactRegion`'s is optional. A backend that summarizes via `ctx.llm.stream()` **must** forward it into the call's `GenerateOptions.signal`, so an abort or fiber dispose tears down the in-flight summarization instead of leaving an orphaned model call running past the cancellation. The session being compacted comes from the agent context; the turn that the `compact/*` events belong to is recoverable from the log (the currently-open turn), so the backend stamps it from the log rather than trusting a caller-supplied value.

View File

@@ -14,7 +14,8 @@
* implementation (deferred) / consumer (a `/compact` tool, deferred) — modeled * implementation (deferred) / consumer (a `/compact` tool, deferred) — modeled
* on the bash trio. Unlike `dsh-bash`, this interface necessarily * on the bash trio. Unlike `dsh-bash`, this interface necessarily
* depends on `dsh-session` and `dsh-llm`: the contract's verbs are defined over * depends on `dsh-session` and `dsh-llm`: the contract's verbs are defined over
* a `Session` and its output is the `ContentBlock` vocabulary. That deviation * an agent-owned `Session` and the durable summary event uses the
* `ContentBlock` vocabulary. That deviation
* from the "interface depends only on cordis" guidance is intentional and * from the "interface depends only on cordis" guidance is intentional and
* recorded in the [compaction capability-seam RFC](../../../../docs/rfc/implemented/feature/2026-06-18-compaction-capability-seam.md). * recorded in the [compaction capability-seam RFC](../../../../docs/rfc/implemented/feature/2026-06-18-compaction-capability-seam.md).
* *
@@ -132,10 +133,10 @@ export abstract class CompactService extends Service {
* open (unclosed) tail step is invalid — its tool-calls have no results yet. * open (unclosed) tail step is invalid — its tool-calls have no results yet.
* `dsh-session` exports `isToolPairingBalanced` for this check. * `dsh-session` exports `isToolPairingBalanced` for this check.
* *
* @param session - the session whose surface is mutated.
* @param start - inclusive seq of the first surface node to compact. * @param start - inclusive seq of the first surface node to compact.
* @param end - inclusive seq of the last surface node to compact. * @param end - inclusive seq of the last surface node to compact.
* @param agent - agent context used by router-aware summarizers. * @param agent - agent context whose session is mutated and whose routing
* options are used by summarizers.
* @param signal - optional cancellation signal. A backend that summarizes via * @param signal - optional cancellation signal. A backend that summarizes via
* `ctx.llm.stream()` MUST forward this into the call's `GenerateOptions.signal` * `ctx.llm.stream()` MUST forward this into the call's `GenerateOptions.signal`
* so an abort/dispose tears down the in-flight summarization rather than * so an abort/dispose tears down the in-flight summarization rather than
@@ -146,10 +147,10 @@ export abstract class CompactService extends Service {
* prior replace can leave the surface non-monotonic in seq order), or if * prior replace can leave the surface non-monotonic in seq order), or if
* either boundary is not a balanced tool-pairing cut (would split a step's * either boundary is not a balanced tool-pairing cut (would split a step's
* tool-call/result pair). * tool-call/result pair).
* @returns what the compaction did (the replaced range and its summary node). * @returns the replaced range and its token accounting. The durable
* `compact/summary` event owns the summary and bookkeeping-event identity.
*/ */
abstract compactRegion( abstract compactRegion(
session: Session,
start: number, start: number,
end: number, end: number,
agent: CompactAgentContext, agent: CompactAgentContext,

View File

@@ -49,14 +49,6 @@ declare module '@deepseek-ai/dsh-session' {
/** Result of a successful compaction operation. */ /** Result of a successful compaction operation. */
export interface CompactionResult { export interface CompactionResult {
/** The seq of the appended `compact/start` event. */
startSeq: number
/** The seq of the appended `compact/summary` event. */
summarySeq: number
/** The seq of the appended `compact/end` event. */
endSeq: number
/** The summary content blocks produced by the backend. */
summary: ContentBlock[]
/** /**
* The surface-boundary pair that was shadowed: the seqs of the first * The surface-boundary pair that was shadowed: the seqs of the first
* (`start`) and last (`end`) surface nodes of the replaced range. A * (`start`) and last (`end`) surface nodes of the replaced range. A

View File

@@ -27,28 +27,24 @@ class StubCompactService extends CompactService {
} }
override async compactRegion( override async compactRegion(
session: Session,
start: number, start: number,
end: number, end: number,
_agent: CompactAgentContext, agent: CompactAgentContext,
signal?: AbortSignal, signal?: AbortSignal,
): Promise<CompactionResult> { ): Promise<CompactionResult> {
this.lastSignal = signal this.lastSignal = signal
const session = agent.session
// Minimal stub honoring the lock + log-only event contract. // Minimal stub honoring the lock + log-only event contract.
const startEvent = session.append('compact/start', { turn: 0 }) session.append('compact/start', { turn: 0 })
const summaryEvent = session.append('compact/summary', { session.append('compact/summary', {
summary: [{ type: 'text', text: 'stub' }], summary: [{ type: 'text', text: 'stub' }],
shadowedRange: { start, end }, shadowedRange: { start, end },
shadowedSeqs: [], shadowedSeqs: [],
shadowedTokenCount: 0, shadowedTokenCount: 0,
model: 'stub', model: 'stub',
}) })
const endEvent = session.append('compact/end', { turn: 0 }) session.append('compact/end', { turn: 0 })
return { return {
startSeq: startEvent.seq,
summarySeq: summaryEvent.seq,
endSeq: endEvent.seq,
summary: [{ type: 'text', text: 'stub' }],
shadowedRange: { start, end }, shadowedRange: { start, end },
shadowedSeqs: [], shadowedSeqs: [],
shadowedTokenCount: 0, shadowedTokenCount: 0,
@@ -88,7 +84,7 @@ describe('CompactService seam', () => {
const svc = new StubCompactService(ctx) const svc = new StubCompactService(ctx)
const session = new Session(SessionId('s')) const session = new Session(SessionId('s'))
const result = await svc.compactRegion(session, 0, 0, stubAgent(session, 'm')) const result = await svc.compactRegion(0, 0, stubAgent(session, 'm'))
const startEvent = session.events.find(e => e.type === 'compact/start') const startEvent = session.events.find(e => e.type === 'compact/start')
expect(startEvent).toBeDefined() expect(startEvent).toBeDefined()
@@ -96,8 +92,9 @@ describe('CompactService seam', () => {
// verify the runtime value is absent. // verify the runtime value is absent.
const raw = startEvent as unknown as { surfaceOp?: unknown } const raw = startEvent as unknown as { surfaceOp?: unknown }
expect(raw.surfaceOp).toBeUndefined() expect(raw.surfaceOp).toBeUndefined()
expect(result.summarySeq).toBeGreaterThan(result.startSeq) expect(result.shadowedRange).toEqual({ start: 0, end: 0 })
expect(result.endSeq).toBeGreaterThan(result.summarySeq) expect(session.events.filter(e => e.type.startsWith('compact/')).map(e => e.type))
.toEqual(['compact/start', 'compact/summary', 'compact/end'])
}) })
it('threads the cancellation signal through to the backend', async () => { it('threads the cancellation signal through to the backend', async () => {
@@ -106,7 +103,7 @@ describe('CompactService seam', () => {
const session = new Session(SessionId('s')) const session = new Session(SessionId('s'))
const controller = new AbortController() const controller = new AbortController()
await svc.compactRegion(session, 0, 0, stubAgent(session, 'm'), controller.signal) await svc.compactRegion(0, 0, stubAgent(session, 'm'), controller.signal)
expect(svc.lastSignal).toBe(controller.signal) expect(svc.lastSignal).toBe(controller.signal)
await svc.compactIfNeeded(stubAgent(session), '', [], controller.signal) await svc.compactIfNeeded(stubAgent(session), '', [], controller.signal)

View File

@@ -109,7 +109,7 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
summary: 'Abstract compaction service.', summary: 'Abstract compaction service.',
methods: [ methods: [
'abstract compactIfNeeded( agent: CompactAgentContext, fullSystemPrompt: string, sessionPrefix: readonly Message[], signal: AbortSignal, ): Promise<CompactionResult | null>', 'abstract compactIfNeeded( agent: CompactAgentContext, fullSystemPrompt: string, sessionPrefix: readonly Message[], signal: AbortSignal, ): Promise<CompactionResult | null>',
'abstract compactRegion( session: Session, start: number, end: number, agent: CompactAgentContext, signal?: AbortSignal, ): Promise<CompactionResult>', 'abstract compactRegion( start: number, end: number, agent: CompactAgentContext, signal?: AbortSignal, ): Promise<CompactionResult>',
], ],
}, },
{ {
@@ -624,7 +624,7 @@ export const TYPE_API: readonly TypeApiEntry[] = [
}, },
{ {
name: 'CompactionResult', name: 'CompactionResult',
declaration: 'export interface CompactionResult {\n startSeq: number;\n summarySeq: number;\n endSeq: number;\n summary: ContentBlock[];\n shadowedRange: {\n start: number;\n end: number;\n };\n shadowedSeqs: number[];\n shadowedTokenCount: number;\n}', declaration: 'export interface CompactionResult {\n shadowedRange: {\n start: number;\n end: number;\n };\n shadowedSeqs: number[];\n shadowedTokenCount: number;\n}',
}, },
{ {
name: 'ConfinedArgv', name: 'ConfinedArgv',