Tianyi Cui
ab19fed77c
The first real LlmAdapter implementations, shipped as a deliberate pair:
same models and wire protocol, completely different internals, so the
StreamChunk protocol is verified across independent implementations.
- dsh-llm-deepseek: hand-rolled fetch + SSE parser + chunk-translation
state machine against the official chat-completions format (thinking
mode via top-level thinking/reasoning_effort; the empty-string
reasoning_content first chunk; usage attached to the finish chunk or
trailing; reasoning_content passback on tool-call turns; disjoint
cache-token accounting).
- dsh-llm-pi-ai: the same endpoint through @earendil-works/pi-ai,
mapping its event vocabulary (parsed tool arguments, in-stream error
events, folded reasoning tokens) onto the same chunks.
The agent loop now honors the in-band error path: an adapter that ends
its stream with finish {kind:error|aborted} (the only option for
adapters that can't throw mid-stream, like pi-ai) is translated into a
step error, so the turn ends error/aborted with a logged error event
instead of a normal completed assistant message. This makes the
StreamChunk error contract real for both adapters; docs/architecture.md
and the StreamChunk doc are updated accordingly.
New yarn test:e2e (vitest.e2e.config.ts, *.e2e.ts) runs key-gated
real-API matrices for both adapters across V4 Flash/Pro and all
thinking/effort levels; it self-skips without DEEPSEEK_API_KEY. Unit
suites run against local node:http mock SSE servers at 100% per-file
coverage.
dsh-llm
Provider-neutral LLM vocabulary and abstract service. This package defines the canonical language spoken by the agent loop, session logs, and every plugin.
Service: LlmService (ctx key: llm)
An adapter registry plus streaming / non-streaming call surfaces. Both call surfaces are interceptable via waterfall events.
Public API
ctx.llm.registerAdapter(models: string[], adapter: LlmAdapter): () => voidRegister an adapter for the given model names. Disposed with the calling fiber.ctx.llm.models(): string[]— model names with a registered adapter.ctx.llm.stream(options: GenerateOptions): AsyncIterable<StreamChunk>Stream one model call as raw chunks (token-level deltas).ctx.llm.streamBlocks(options: GenerateOptions): AsyncIterable<ContentBlock>Stream as completed content blocks (convenience view).ctx.llm.generate(options: GenerateOptions): Promise<GenerateResult>One model call, fully assembled.
Events
| Event | Mode | Purpose |
|---|---|---|
llm/stream |
waterfall | Intercept/wrap every streaming model call (retry, caching, routing) |
llm/generate |
waterfall | Intercept/wrap every non-streaming model call |
llm/adapter-change |
emit | An adapter was registered or unregistered |
Extension points
- Subclass
LlmAdapterand callctx.llm.registerAdapter(models, adapter)to add a new model provider. - Wrap
llm/streamorllm/generateviactx.on()waterfall listeners for caching, retry, logging, rate-limiting, etc.
Content-block vocabulary (types.ts)
Messages are arrays of typed content blocks: text, reasoning, tool-call,
tool-result, image. The union is derived from the merge-extensible
ContentBlockMap, so plugins can add block types via declaration merging.
Streaming is a raw chunk protocol (block-start, text-delta,
reasoning-delta, tool-call-delta, block-end, usage, finish).
BlockAssembler is the single shared implementation that assembles chunks into
blocks/messages.
Classes
LlmAdapter— abstract base class for provider adapters. The only required method isstream().BlockAssembler— incrementally assembles raw chunks into complete content blocks and an assistant message. Used by the agent loop (raw chunks for replay- assembled for history) and by
streamBlocks()/generate().
- assembled for history) and by
LlmError— typed error with acodestring (NO_ADAPTER,DUPLICATE_ADAPTER).
What is NOT here (TODO)
- DeepSeek V4 adapter — the first real adapter lands in a later phase.
- Streaming protocol review — the chunk protocol has
TODO(review)markers and needs careful review before the first real adapter (DeepSeek V4 wire format, partial JSON arguments, interleaved reasoning signatures, ...).