* feat(agent-core): compress oversized images before sending to the model
Downsample images to a 2000px longest-edge and per-image byte budget at the
single prompt-ingestion chokepoint (the prompt/steer RPC) and on tool results
(ReadMediaFile, MCP), so every client transport — CLI, web, desktop, ACP, SDK —
is covered uniformly inside the core. PNG screenshots stay lossless and only
degrade to JPEG when the byte budget cannot otherwise be met. Best-effort: the
original image is sent unchanged if compression fails.
* fix(agent-core): serialize prompt/steer RPCs to avoid a turn-claim race
The prompt/steer RPC handlers await image compression before turn.launch()
synchronously claims the active turn, so two overlapping calls could both
compress first — letting the faster-to-compress one win the turn and strand the
other on agent_busy. Run these two RPCs through a per-agent serialization chain
so they claim in submit order; cancel and the other RPCs stay immediate.
* fix: update flake.nix pnpmDeps hash for the jimp dependency
Adding jimp to the workspace changed pnpm-lock.yaml, so the pnpmDeps
fixed-output hash was stale and the nix build failed. Update it to the value
the CI nix build reported.
* fix(agent-core): guard image compression against decompression bombs
A tiny-byte, huge-dimension image (e.g. a solid 30000x30000 PNG) would be fully
decoded into a multi-gigabyte bitmap by Jimp before any resize — an OOM vector
the byte budget never catches. Skip compression when the sniffed pixel count
exceeds MAX_DECODE_PIXELS (~100 MP), before the decode; oversized images pass
through uncompressed as they did before compression existed.
* fix(agent-core): cap decode byte size before compressing images
Compression runs before downstream size caps (e.g. the 10MB MCP per-part
limit), so a huge or invalid base64 image from an MCP tool was Buffer.from-
decoded — and handed to Jimp — just to be dropped afterward. Add a
MAX_DECODE_BYTES ceiling (64MB, overridable) checked before the base64 decode
and before Jimp, the byte-side complement to the pixel-count guard; oversized
payloads pass through uncompressed.
* refactor(agent-core): compress images at ingestion, not on the turn RPC
Move image compression off the prompt/steer RPC path and back to each ingestion
site (CLI paste, server upload resolution, ACP conversion; ReadMediaFile and MCP
already compressed at their producers). Compressing on the RPC control path put
an async step before the synchronous turn-claim, which spawned a series of
races: prompt/steer interleaving, and — with a cancel arriving mid-compression —
an ineffective abort that let a cancelled prompt launch anyway.
Treating compression as a pure input-stage transform (done while the content
part is built, before it ever enters the agent loop) removes those races
structurally: rpc.prompt/steer are plain synchronous handlers again, and the
serialization/cancel-window machinery is gone. Records stay compressed, resume
stays consistent, and coverage degrades gracefully (a new client that skips
compression just sends a larger image, as before this feature).
* fix: compress inline base64 prompts and honor ACP cancels mid-compression
Two contained ingestion-site follow-ups:
- server: resolvePromptMediaFiles now also compresses images submitted as an
inline `{ kind: 'base64' }` source, not just uploaded files, so the REST
inline-base64 path gets the same downsampling.
- acp-adapter: AcpSession tracks a pending-abort flag while prompt() awaits
image compression (before any turn exists). A session/cancel in that window
flips it, so the prompt returns `cancelled` instead of launching a turn the
client already stopped.
* fix(acp-adapter): cover all concurrent pre-turn prompts on cancel
The pending-abort marker was a single session field, so with two
`session/prompt` requests compressing large inline images at once the later
one overwrote it and a `session/cancel` could mark only one — the other
launched after the client had cancelled. Track a token per in-flight prompt in
a set and flip them all on cancel so every pre-turn prompt is covered.
* chore(node-sdk): declare jimp as a devDependency
The SDK re-exports the image compressor, whose lazy `import('jimp')` (inside
the bundled agent-core code) is inlined into the published dist. jimp was
resolved only transitively via agent-core, so declare it as an explicit build
input here — matching the CLI — to make the bundling reliable rather than
phantom. It stays a devDependency: jimp is bundled, not a runtime dependency.
|
||
|---|---|---|
| .. | ||
| scratch | ||
| src | ||
| test | ||
| AGENTS.md | ||
| CHANGELOG.md | ||
| package.json | ||
| README.md | ||
| SECURITY.md | ||
| tsconfig.json | ||
| tsdown.config.ts | ||
| vitest.config.ts | ||
@moonshot-ai/server
Local REST + WebSocket server that exposes the Kimi Code SDK over a stable wire
protocol. It hosts agent-core sessions and serves them under a single
/api/v1 prefix. This package is private — it is not published on its own;
it ships inside the kimi CLI (apps/kimi-code) and is launched via
kimi server run.
What it does
- Hosts
agent-coresessions, prompts, tools, approvals, questions, and workspaces in process. - Exposes them over REST (Fastify) and WebSocket (
ws) under/api/v1. - Serves the built-in web UI (
apps/kimi-web) as static assets when awebAssetsDiris provided. - Publishes machine-readable contract docs:
/openapi.json,/asyncapi.json.
Running it
# From the repo root — dev server with auto-restart
pnpm dev:server
pnpm dev:server:restart
# Checks
pnpm --filter @moonshot-ai/server typecheck # tsc --noEmit
pnpm --filter @moonshot-ai/server test # vitest run
pnpm --filter @moonshot-ai/server build # tsdown
The public entry point is startServer(opts) in src/start.ts, which returns a
RunningServer. In production the CLI command kimi server run
(apps/kimi-code/src/cli/sub/server/run.ts) imports and calls it. This package
has no dev script of its own — always start it from the repo root or via the
CLI.
By default the server listens on 127.0.0.1:58627; e2e clients target it with
KIMI_SERVER_URL (default http://127.0.0.1:58627).
Architecture
apps/kimi-code (CLI) apps/kimi-web (browser)
│ │
└──────────┬───────────────────┘
│ REST + WebSocket, /api/v1
┌──────────▼───────────┐
│ @moonshot-ai/server │
│ Fastify REST │
│ ws gateway │
│ DI container │ ← @moonshot-ai/agent-core
│ agent-core sessions │ ← @moonshot-ai/agent-core
└──────────────────────┘
- REST (
src/routes/): domain modules aggregated byregisterApiV1Routes.ts. Routes are declared withmiddleware/defineRoute.ts, which bundles Zod validators with the OpenAPI response schema. - WebSocket (
src/ws/,src/services/gateway/): per-sessionseq,server_hello/ack/event/resync_requiredframes, replay and fan-out. - DI (
src/services/serviceCollection.ts): seeds the container from@moonshot-ai/agent-core(getSingletonServiceDescriptors()) and layers in server-owned gateways plusIApprovalService/IQuestionServiceimplementations. - OS service managers (
src/svc/): launchd / systemd / schtasks backends forkimi server install/start.
Wire protocol notes
- Envelope: every REST response is
{ code, msg, data, request_id }and the HTTP status is effectively always 200 — checkcode(0 = ok), not the status. :actionendpoints: some routes use an:id:actionsuffix (e.g./sessions/{id}:undo); the suffix is parsed byroutes/action-suffix.ts.- Single-instance lock: a running server acquires a lock; a second start on
the same home throws
ServerLockedError. Tests pass a uniquelockPath/port.
Related packages
@moonshot-ai/agent-core— the agent engine the server hosts, including the in-process DI service layer it wires together.@moonshot-ai/protocol— wire types and the AsyncAPI document.@moonshot-ai/node-sdk— typed in-process facade for user code (KimiHarness,Session); prefer it over hand-rolling REST/WS calls.@moonshot-ai/server-e2e— wire-level e2e client and scenarios against a running server.
Development
For conventions, gotchas, and the boot wiring order, see
packages/server/AGENTS.md. For the service naming and
registration rules, see
packages/services/AGENTS.md.