dsh-llm-deepseek/packages/llm/llm-deepseek官方

English | 中文

DeepSeek chat-completions adapter for the harness LLM seam: direct fetch + SSE (framed by eventsource-parser) translating the official wire format (source of truth: the API docs — guides/thinking_mode, guides/tool_calls, api/create-chat-completion) into the StreamChunk protocol.

A second, library-backed implementation of the same seam exists in @deepseek-ai/dsh-llm-pi-ai. This package owns the deepseek-official provider route — deliberately distinct from pi-ai's catalog name deepseek, so one composition can mount both DeepSeek paths side by side; registering another adapter for deepseek-official itself still throws LlmError('DUPLICATE_ADAPTER').

The package root exposes the Cordis plugin contract and DeepSeekAdapter; wire serialization, SSE parsing, and chunk translation helpers are not part of that root contract.

Config

- id: llm-deepseek
  name: '@deepseek-ai/dsh-llm-deepseek'
  config:
    apiKeyEnv: DEEPSEEK_API_KEY  # default; resolved per request via ctx.credentials, then the environment
    baseURL: https://api.deepseek.com # optional; $DEEPSEEK_BASE_URL then the public API when omitted
    thinking: enabled        # optional; provider default is enabled
    reasoningEffort: high    # optional; off | low | high | max — omitted ⇒ high
    maxTokens: 256000        # optional positive per-request output cap; this is the default
    streamIdleTimeoutMs: 300000 # optional; positive finite Node timer delay; five-minute default
    maxRequestImageBytes: 20971520 # optional positive integer; 20 MiB base64-payload default
    retryPolicy:             # optional; omission uses normal mode with five retries
      mode: always           # normal | always
      backoff:
        initialDelayMs: 500
        maxDelayMs: 10000
        jitterRatio: 0.1
    defaultContextWindow: 1000000 # optional positive-integer fallback; this is the default
    models:                  # optional; defaults to V4 Flash and V4 Pro
      - id: deepseek-v4-flash
        name: DeepSeek-V4-Flash
      - id: private-vision
        name: Private Vision
        inputModalities: [text, image]
      - id: private-reasoner
        description: Company-hosted reasoning model
        contextWindow: 512000

The plugin registers the single provider route deepseek-official together with its resolved retryPolicy; omission resolves to normal mode with five retries. A request selects it with provider: deepseek-official; its model is passed through as the wire model string, so changing DeepSeek models does not require lifecycle-time registration. Omitting models advertises deepseek-v4-flash and deepseek-v4-pro, each with a 1,000,000-token context window; an explicit list replaces those defaults, while models: [] advertises none. Vision models are not advertised by default until their endpoint rollout is complete, but a deployment can add one with inputModalities: [text, image]. Catalog entries are exposed through ctx.llm.listModels('deepseek-official') for clients such as ACP editors and the Web selector, but remain advisory: unlisted model ids still pass through unchanged. An omitted entry name defaults to its id, and omitted inputModalities means text only.

An image-capable catalog entry may declare inputModalities: [text, image]. The adapter resolves user and tool-result ImageBlock references through ctx.attachments, verifies the stored bytes, and sends transient data:<media-type>;base64,... image_url parts without changing the durable session message. Text-only and unlisted models reject image input before credential, attachment, or network I/O. System and assistant history remain image-free; tool-result images follow their string-only tool messages in a separate user message.

maxRequestImageBytes bounds accumulated base64 image payload and defaults to 20 MiB, leaving headroom below the official 30 MiB request-body limit for text, tools, and JSON framing. When history exceeds the bound, the oldest images become the fixed model-visible placeholder [image omitted to keep the request within its image limit; older images are omitted first. If this image is still needed, read its file again when a path is available; otherwise ask the user to attach it again.] until the request fits; omitted attachments are not read. Attachment admission continues to own per-image and per-message raw-byte, media, dimension, and pixel limits.

contextWindow is optional per configured model and is not exposed through the advisory catalog. ctx.llm.resolveModelInfo('deepseek-official', model).context returns an exact model value first, then defaultContextWindow for an entry without capacity or an unlisted pass-through id. The adapter default is 1,000,000; pressure-sensitive plugins therefore get deployment-owned capacity without treating the model selector as authoritative. Registering another adapter for deepseek-official throws LlmError('DUPLICATE_ADAPTER').

maxTokens is the adapter-configured output cap for conversation requests and defaults to 256,000. A catalog entry may carry its own maxTokens, which wins for that model; an entry without one, and any unlisted pass-through id, resolve to the profile value, so adding a per-model cap changes one model rather than the route. Exact-model resolution exposes the winner as defaultMaxTokens; LlmRuntime materializes that value into GenerateOptions.maxTokens before the agent loop writes request/header, so the wire request remains reconstructable. An explicit request or AgentOptions.maxTokens value wins and is serialized as max_tokens. The adapter does not clamp this request budget against contextWindow; deployments with a smaller context or provider output limit must configure a compatible maxTokens.

The same exact-model result exposes ordered off, low, high, and max efforts under reasoning for every pass-through model when deployment policy permits thinking. reasoningEffort selects the deployment default and falls back to high when omitted. agent/request can replace it on each conversation step; the resolved value is logged in request/header. low, high, and max enable thinking and serialize as the same official top-level reasoning_effort value; adapter-owned off instead serializes thinking.type: disabled and omits reasoning_effort. An unsupported value fails with UNSUPPORTED_REASONING_EFFORT before network I/O.

thinking: disabled is a deployment lock that publishes only off with off as its default. Omitting reasoningEffort or configuring it as off is valid; configuring low, high, or max fails plugin loading, and a direct per-request attempt to enable thinking fails before network I/O. A request with GenerateOptions.purpose: 'session-title' also forces thinking disabled and omits the already-resolved effort, reserving its bounded output for visible title text without changing conversation or compaction defaults.

streamIdleTimeoutMs bounds each outstanding provider read, including the initial fetch, without counting time the consumer spends between chunks. DeepSeek SSE comments rearm an outstanding read as transport activity but never become StreamChunk values or session-log events. One stable abort signal reaches the request and body reader for the whole call; expiry stops the transport and throws LlmError('TIMEOUT'), while an earlier caller abort throws LlmError('ABORTED'). The adapter makes exactly one provider request per stream() call; it registers the configured policy as provider metadata, and dsh-llm-retry separately executes it at durable agent-step boundaries.

Dynamic configuration (settings + credentials)

Connection facts are not frozen at load. resolveAdapterOptions is the one explicit resolve step from raw config to validated facts, and the adapter re-reads them through a thunk once per operation: base URL, catalog, request defaults, image bound, and idle budget all take effect on the next request, while an in-flight stream keeps the facts it started with. Three optional seams feed that thunk:

• ctx.settings — the plugin registers the llm-deepseek namespace with this same Config schema and its cordis.yml entry as the composition base, so a llm-deepseek: section in the user settings document overrides any field without a restart. Without a mounted settings service the entry config alone drives the adapter, unchanged. A live settings snapshot that passes the schema but fails a beyond-schema bound (a duplicate catalog id, a broken thinking/effort pair) keeps the last good facts and logs the failure; the entry config itself still fails plugin load.
• ctx.credentials — the API key resolves per stream call, from the same resolved snapshot that supplies the endpoint. Configuration carries only apiKeyEnv, never a literal key: the reference resolves through the credential seam, and without a mounted seam through the trusted environment layers. Because credential facts travel with the connection facts, a settings snapshot the resolver rejects contributes neither its endpoint nor its key: the whole previous generation keeps serving. Every resolved key is format-checked before use, so a value no HTTP header can carry is refused with LlmError('INVALID_CREDENTIAL') naming the failing entry point — never any part of the key — instead of surfacing as an opaque fetch TypeError. A request with no key anywhere fails with MISSING_CREDENTIAL naming every configuration entry point, while the route stays registered and the catalog stays browsable — first-run onboarding is "browse models, store the key, prompt again", with no restart between.
• ctx.attachments — image requests resolve this service at request time, so Cordis load order does not freeze optional image availability. Absence rejects image input with UNSUPPORTED_CONTENT; text-only calls do not require the service.

The one registration-captured fact is the retry policy: when its resolved value changes, the plugin re-registers the route in place (same adapter instance, one synchronous section), so ctx.llm.providerRetryPolicy('deepseek-official') always reports the current policy.

The plugin also declares its route in the configurable-provider directory (ctx.llm.listConfigurableProviders()): provider deepseek-official, settings namespace llm-deepseek, empty settings path — the whole section is the profile. Configuration surfaces use that entry to offer this adapter alongside dormant pi-ai providers.

App attribution

Every request carries the shared attribution header from dsh-llm's attributionHeaders() - the mandatory User-Agent baseline identifying the harness (see dsh-llm § App attribution). Direct DeepSeek requests and OpenAI-compatible gateway requests get no provider-specific app-attribution headers under this adapter contract; OpenRouter app attribution is deferred to a future explicit OpenRouter adapter or mode. A request whose GenerateOptions.purpose is compaction (dsh-compaction-basic's auxiliary summarization call) additionally carries x-deepseek-harness-compact: 1, so the host can separate compaction traffic from conversation requests.

DeepSeek request identity is separate from app attribution. After credential resolution, every provider request carries x-deepseek-harness-user-id with the stable anonymous id from @deepseek-ai/dsh-anonymous-user-id; a request carrying GenerateOptions.sessionId also sends that exact value as x-deepseek-harness-session-id, while a direct call without a session omits the session header. Both headers go to the resolved baseURL, including a configured gateway, and remain outside the request body and model-visible content.

Wire-format notes

• Streaming only (stream_options.include_usage always on). usage may arrive attached to the finish chunk or as a trailing usage-only chunk — the translator defers both to [DONE], so usage always precedes finish and nothing follows finish.
• The adapter-owned off effort maps to thinking: {type: 'disabled'} and never crosses the wire as reasoning_effort: 'off'.
• The first thinking-mode chunk carries reasoning_content: "" — handled (no spurious reasoning block).
• Reasoning passback rule: every assistant turn that carried reasoning serializes reasoning_content back in history. Thinking mode requires it on tool-call turns; DeepSeek ignores it elsewhere, while a gateway re-encoding the conversation for another vendor recovers that turn's upstream thinking signature by hashing the replayed text.
• Image-capable user messages preserve text/image order. Tool-role content remains a string; consecutive tool-result images are grouped into the following user message with Attached image(s) from tool result:.
• Cache accounting: cacheReadTokens ← prompt_cache_hit_tokens / prompt_tokens_details.cached_tokens; DeepSeek reports no cache-write metric.

Errors

Non-2xx responses throw LlmError with stable codes: AUTH (401/403), QUOTA (a response whose provider details identify exhausted quota, balance, or credits), RATE_LIMIT (other 429s), CONTEXT_WINDOW_EXCEEDED (a 400 whose provider code, type, or message identifies context overflow), INVALID_REQUEST (other 400s and 413), SERVER (5xx), HTTP_<status> otherwise. Its serializable failure retains the HTTP status plus a valid positive Retry-After seconds/date delay and x-request-id / x-deepseek-request-id when present. Attachment reads retain their stable attachment failure code rather than becoming transport failures. A pre-response transport failure (DNS, refused connection, TLS, proxy) throws TRANSPORT naming the configured endpoint and chaining the original rejection as cause; caller aborts throw ABORTED, and the loop's cancellation signal remains authoritative. Protocol violations throw STREAM_CLOSED (no [DONE]) or MALFORMED_RESPONSE (bad JSON payload). Unknown wire finish_reasons (e.g. content_filter, insufficient_system_resource) become finish {kind: 'error', failure} chunks, and a completed stream whose stop (or absent) finish opened no content blocks becomes a finish {kind: 'error'} with code EMPTY_RESPONSE (retried by default policy).

Model Experience

DeepSeek request

What the model sees

The selected DeepSeek model receives the harness system prompt, message history, tool schemas, stop sequences, and call config without adapter-authored prompt prose. The vision model also receives retained user and tool-result images as base64 data URLs; an over-budget older image is represented by the documented placeholder. Reasoning content from a prior assistant turn is passed back verbatim, whether or not that turn called a tool.

Token effect

Provider tokenization governs exact text and image-token input. Reasoning passback carries every reasoned turn's chain of thought into later requests, while dropping over-budget images avoids paying those tokens again; cache-read usage is reported when available.

KV Cache effect

An unchanged assembled prefix, including deterministically encoded retained images and placeholders, is eligible for DeepSeek cache reuse, which this adapter reports in usage. A model-route change or any upstream prompt, schema, prefix, history, or image-budget change may prevent reuse from the first changed token; reasoning passback appends on every reasoned turn.

DeepSeek response

What the model sees

Reasoning, text, and raw-string tool arguments are translated into harness chunks for the loop to log and assemble.

Token effect

Generated tokens follow the request's logged reasoning effort and maxTokens; only loop-retained blocks affect later input.

KV Cache effect

Loop-retained response blocks append to the next request and preserve its earlier reusable prefix; dropped blocks have no later cache effect. Changing the provider or model selects a different cache domain.

Known Limitations and Deferred Work

• A settings models list replaces the composition list wholesale — settings-layer merging is per-field, and arrays are one field; per-entry catalog merging would need a keyed shape.
• tool_choice is not mapped — not part of the core vocabulary (MVP cut, shared with the pi-ai twin).
• Requests use raw fetch, not @cordisjs/plugin-http — no shared proxy/interception configuration; adoption is deferred until a second adapter wants it (TODO(http)).
• Plugin-added content block types are skipped — core text and supported image blocks are serialized, and empty tool output crosses the wire as the literal (no output).
• Images are input-only durable attachments — direct external URLs, the Files API, and assistant image output are not supported.