feat: Code Mode — the registry's mode config, the SDK codegen, and the run_code bridge

The dsh-tools half of the Code Mode RFC (its fourth, final change): the
registry gains its first config — mode: native | code | both — and OWNS how
its tools reach the model. 'code' contributes exactly one wire tool,
run_code, plus a lazy tools:sdk prompt section declaring every other tool
as a generated TypeScript API (jsonSchemaToTs: total over the defineTool
subset, unknown degradation, lexicographic byte-identical rendering);
'both' ships both representations; 'native' is byte-for-byte the old
behavior. Non-native modes fail every assembly loudly without a
typescript-language ctx.codeRuntime.

run_code's dispatch bridge: JSON-normalizes each binding argument before
dispatch (what dispatches is what the tool/code-dispatch event logs — the
append can never fail on payload shape; BigInt/circulars reject that one
call), serializes all program tool calls through a per-run queue (even
Promise.all — no concurrency-safety metadata yet), routes every sub-call
through tools/pre-execute → tools/post-execute (a deny rejects the
program-side promise), drops sub-call additionalContext (no safe outlet
mid-run; pinned), owns a run-scoped abort that follows the outer signal in
and fires on settlement (in-flight sub-dispatch aborted, queued abandoned,
queue drained before returning), and converts a failed run into
CodeRunFailedError → a structured isError carrying kind + captured logs.
tool/code-dispatch joins SessionEventMap by declaration merging (log-only;
deriveMessages ignores it).

The composed surface: the tools config forwards through agent-core and
both app packages; examples/code-agent + demo:code run the worker runtime
under mode code (keyless boot smoke + a with-key e2e proving the collapsed
[run_code] header, the dispatch events, and the file the program wrote);
two new snapshot scenarios (code-mode-turn, both-mode-turn) record the SDK
section, collapsed header, dispatch events, and result card — each its own
header-pinning class (the harness gains per-scenario config overlays and
per-class pins). Catalogs, graphs, cookbook, hooks-bridge notes, and the
RFC (moved to implemented/, restructured to decision-era headings) updated
in the same change.
This commit is contained in:
Tianyi Cui
2026-07-08 12:58:23 +08:00
parent aa2a7f9a8a
commit b59d245c7c
56 changed files with 2395 additions and 102 deletions

View File

@@ -38,7 +38,7 @@ import { basename, resolve } from 'node:path'
import { Context } from 'cordis'
import type { ToolSchema } from '@deepseek-ai/dsh-llm'
import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
import ToolRegistry from '@deepseek-ai/dsh-tools'
import ToolRegistry, { type Config as ToolsConfig } from '@deepseek-ai/dsh-tools'
import LocalBashExecutor from '@deepseek-ai/dsh-bash-local'
import LocalFileSystem from '@deepseek-ai/dsh-fs-local'
import WebService from '@deepseek-ai/dsh-web'
@@ -84,6 +84,13 @@ interface ToolPackage {
/** Plug the injected seams + the tool plugin onto a context that already
* carries `systemPrompt` + `tools`. */
mount: (ctx: Context) => Promise<void>
/**
* Config for the caller's `ToolRegistry` mount. The registry itself ships a
* model-facing tool (`run_code`, registered under a non-native `mode`), so
* ITS catalog entry boots the registry in the mode that surfaces it;
* every other entry uses the default (native) registry.
*/
toolsConfig?: ToolsConfig
/**
* A deployment note rendered after the package's tools, for a fact that
* booting the package alone cannot show. The registered tool NAME can be a
@@ -100,6 +107,20 @@ interface ToolPackage {
* guard proves it is exhaustive against the on-disk glob.
*/
const TOOL_PACKAGES: ToolPackage[] = [
{
pkg: '@deepseek-ai/dsh-tools',
dir: 'tools',
source: 'packages/core/tools/src/code-mode.ts',
requires: ['ctx.tools', 'ctx.codeRuntime (execution time)', 'ctx.systemPrompt'],
writes: ['tool/call', 'one tool/code-dispatch per bridged sub-call', 'tool/result'],
// The registry's OWN tool: run_code exists only under a non-native mode
// (the registry registers it in its constructor; the code runtime is read
// at assembly/execution time, so the schema harvest needs none mounted).
toolsConfig: { mode: 'code' },
async mount() {},
note:
'Registered by the tool registry itself under `mode: code` / `mode: both` (see the Code Mode RFC). Under `code` it is the ONLY wire tool; the other registered tools are declared to the model as a generated TypeScript SDK prompt section instead, and a program calls them through port-bridged bindings that dispatch through the ordinary tools/pre-execute → tools/post-execute pipeline, one at a time.',
},
{
pkg: '@deepseek-ai/dsh-tool-bash',
dir: 'tool-bash',
@@ -231,7 +252,7 @@ export async function collectToolCatalog(packages: ToolPackage[] = TOOL_PACKAGES
// fiber) — the repo's "dispose must reach quiescence" rule.
try {
await ctx.plugin(SystemPrompt)
await ctx.plugin(ToolRegistry)
await ctx.plugin(ToolRegistry, entry.toolsConfig ?? {})
await entry.mount(ctx)
const schemas = ctx.tools.schemas().sort((a, b) => a.name.localeCompare(b.name))
catalog.push({