Merge remote-tracking branch 'origin/master' into fix/stats

packages/client/connection/README.i18n.yaml

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write packages/client/connection/README.md
-README.md: a82bb55ab65df2732ad16248d2cc9aa15b60e94d
-README.zh.md: d4c19db8e74e3562ec55b5ef60675aef79b82008
+README.md: d3727df981ecbc022345def48b38fbc879e29cb2
+README.zh.md: 69af4cdba553399c0db0e0587e540c2a7f5375d2

packages/client/connection/README.md

@@ -23,3 +23,4 @@ None; this package neither assembles nor sends a provider request.
 ## Known Limitations and Deferred Work
 
 - **History resumes an unattached session** — opening history may create the host-side agent and add latency to the first open; there is no persistence-only read path.
+- **The `/api` bridge buffers each request body in memory** — `maxRequestBodyBytes` (default 160 MiB, sized for the default 100 MiB aggregate image limit after base64 expansion plus envelope headroom) is therefore also the per-request resident bound; a streaming body path would be needed to lower it without shrinking the image limits.

packages/client/connection/README.zh.md

@@ -23,3 +23,4 @@ node 半侧在桥接或 upgrade 前守卫 `/api` 下的每个入口（`src/api-r
 ## 已知限制与暂缓事项
 
 - **History 会恢复未附加的会话**：打开 history 可能创建宿主侧 agent，并增加首次打开的延迟；没有仅从持久化读取的路径。
+- **`/api` 桥把每个请求体整体缓冲在内存里**：`maxRequestBodyBytes`（默认 160 MiB，按默认 100 MiB 图片总量上限经 base64 膨胀加信封余量得出）因此同时是单请求的驻留内存上界；要降低它而不缩小图片限额，需要流式请求体路径。

packages/client/connection/src/client/fixture.ts

@@ -353,7 +353,7 @@ function fixtureUsage(turn: number, step: number): TokenUsage {
   }
 }
 
-/** fx-alpha history script: 74 turns (~150+ messages -> 4 pages at PAGE_MESSAGES=50),
+/** fx-alpha history script: 75 turns (~150+ messages -> 4 pages at PAGE_MESSAGES=50),
  *  mixing reasoning blocks / tool call+result / context. */
 function buildAlphaLog(): SessionEvent[] {
   const events: Record<string, unknown>[] = []
@@ -489,7 +489,7 @@ function buildAlphaLog(): SessionEvent[] {
     push({ type: 'step/end', data: { turn, step: 0 } })
     push({ type: 'turn/end', data: { turn, reason: { kind: 'completed' } } })
   }
-  // Turn 73: todo_write sample — the TodoRow toolview in the flow plus the
+  // Turn 74: todo_write sample — the TodoRow toolview in the flow plus the
   // todo/write snapshot event feeding the TodoPanel plan strip. Two items are
   // in_progress: this fixture chooses the parallel policy, so both surfaces
   // must render a parallel plan rather than the first active item alone.
@@ -548,20 +548,35 @@ function buildAlphaLog(): SessionEvent[] {
   toolTurn(70, 'web_search', '{"query":"deepseek harness architecture"}', 'Search results for deepseek harness architecture.')
   toolTurn(71, 'web_fetch', '{"url":"https://www.deepseek.com/blog/harness-architecture"}', '# Harness architecture\n\nEverything is a plugin.')
 
-  // Turn 72: user and assistant images share one durable fixture object.
-  // The todo turn remains last so its standing projection stays visible.
+  // Turn 72: max-tokens sample — the provider ends the turn at its output cap
+  // mid-sentence, so the chat flow must render the turn-max-tokens notice
+  // instead of ending silently. Ordered before the todo turn for the same
+  // standing-plan reason the bash turn is.
   push({ type: 'turn/start', data: { turn: 72 } })
+  push({ type: 'user/message', surfaceOp: 'append', data: userMessage(text('问题 72：请完整列出全部一百条条目。')) })
+  push({ type: 'step/start', data: { turn: 72, step: 0 } })
+  push({
+    type: 'assistant/message',
+    surfaceOp: 'append',
+    data: { turn: 72, step: 0, message: assistantMessage(text('条目 1：第一条。条目 2：第二条。条目 3：这一条写到一半被')) },
+  })
+  push({ type: 'step/end', data: { turn: 72, step: 0 } })
+  push({ type: 'turn/end', data: { turn: 72, reason: { kind: 'max-tokens' } } })
+
+  // Turn 73: user and assistant images share one durable fixture object.
+  // The todo turn remains last so its standing projection stays visible.
+  push({ type: 'turn/start', data: { turn: 73 } })
   push({
     type: 'user/message',
     surfaceOp: 'append',
     data: userMessage([{ type: 'image', attachment: FIXTURE_IMAGE_REF }, ...text('历史用户图片')]),
   })
-  push({ type: 'step/start', data: { turn: 72, step: 0 } })
+  push({ type: 'step/start', data: { turn: 73, step: 0 } })
   push({
     type: 'assistant/message',
     surfaceOp: 'append',
     data: {
-      turn: 72,
+      turn: 73,
       step: 0,
       message: assistantMessage(
         [...text('结构化模型图片：'), { type: 'image', attachment: FIXTURE_IMAGE_REF }],
@@ -569,11 +584,11 @@ function buildAlphaLog(): SessionEvent[] {
       ),
     },
   })
-  push({ type: 'step/end', data: { turn: 72, step: 0 } })
-  push({ type: 'turn/end', data: { turn: 72, reason: { kind: 'completed' } } })
+  push({ type: 'step/end', data: { turn: 73, step: 0 } })
+  push({ type: 'turn/end', data: { turn: 73, reason: { kind: 'completed' } } })
 
   const todoArgs = JSON.stringify({ todos: fixtureTodos })
-  toolTurn(73, 'todo_write', todoArgs, 'Updated todo list: 1 pending, 2 in progress, 1 completed.')
+  toolTurn(74, 'todo_write', todoArgs, 'Updated todo list: 1 pending, 2 in progress, 1 completed.')
   // The real tool appends the snapshot mid-execution — between tool/call and
   // tool/result — so the fixture reproduces that exact ordering (the last
   // toolTurn events run ... tool/call, tool/result, step/end, turn/end).
@@ -1052,6 +1067,18 @@ function projectionValuesOf(log: readonly SessionEvent[]): Record<string, unknow
   values['contextBreakdown'] = contextBreakdownOf(log)
   // Always present (session-stats unit composed): whole-log turn/step counts.
   values['sessionStats'] = sessionStatsOf(log)
+  // Always present (attachment service composed): the deployment image
+  // limits, constant per boot (mirrors the attachment-local defaults).
+  // Deliberate host divergence: the real gateway never pushes an imageLimits
+  // change frame (constant unit), but the fixture's uniform baseline replay
+  // frames every key here, incidentally exercising higher-seq-wins.
+  values['imageLimits'] = {
+    maxImageBytes: 5 * 1024 * 1024,
+    maxImagesPerMessage: 20,
+    maxMessageImageBytes: 100 * 1024 * 1024,
+    maxImagePixels: 40_000_000,
+    mediaTypes: ['image/png', 'image/jpeg', 'image/webp', 'image/gif'],
+  }
   return values
 }
 
@@ -1516,7 +1543,7 @@ function createFixtureWorld(options: FixtureOptions): FixtureWorld {
     ['my-agent', { trust: 'user', content: "- id: tool-read\n  name: '@deepseek-ai/dsh-tool-read'\n" }],
   ])
   let fixtureDefaultPreset = 'standard'
-  const nextTurn = new Map<SessionId, number>([[sid('fx-alpha'), 74]])
+  const nextTurn = new Map<SessionId, number>([[sid('fx-alpha'), 75]])
   let nextSession = 1
   let nextRpc = 1
   let attachedSessions = options.empty ? 0 : 1

packages/client/connection/src/http-bridge.ts

@@ -5,6 +5,12 @@
 
 import type { IncomingMessage, ServerResponse } from 'node:http'
 
+/** Default carrier cap for all HTTP RPC bodies: sized for the default
+ * aggregate image limit (100 MiB) after base64 expansion plus envelope
+ * headroom (~134.3 MiB required), rounded up for slack. The bridge buffers
+ * each body in memory, so this cap is also the per-request resident bound. */
+export const DEFAULT_MAX_REQUEST_BODY_BYTES = 160 * 1024 * 1024
+
 /** Transport-independent request handler consumed by the Host HTTP bridge. */
 export interface FetchHandler {
   /**
@@ -27,7 +33,7 @@ export async function bridge(
   req: IncomingMessage,
   res: ServerResponse,
   apiHandler: FetchHandler,
-  maxRequestBodyBytes = 32 * 1024 * 1024,
+  maxRequestBodyBytes = DEFAULT_MAX_REQUEST_BODY_BYTES,
 ): Promise<void> {
   const abort = new AbortController()
   // Client-disconnect detection MUST hang off the response, not the request:

packages/client/connection/src/index.ts

@@ -6,7 +6,7 @@ import type {} from '@deepseek-ai/dsh-attachment'
 import type { WebRoute, WebUpgradeRoute } from '@deepseek-ai/dsh-host-webserver'
 import { toFetchHandler } from '@deepseek-ai/dsh-host-apiproxy'
 import { API_PATH, HOST_EVENTS_PATH, MUX_EVENTS_PATH } from './api-path.ts'
-import { bridge } from './http-bridge.ts'
+import { bridge, DEFAULT_MAX_REQUEST_BODY_BYTES } from './http-bridge.ts'
 import { assertTrustedAuthority, isTrustedApiRequest } from './api-request-trust.ts'
 import { HostConnectionService } from './rpc-host.ts'
 import { rejectWebSocketUpgrade, WebSocketDownlinks } from './websocket-downlink.ts'
@@ -42,8 +42,6 @@ function assertImageBodyCapacity(ctx: Context, maxRequestBodyBytes: number): voi
     )
   }
 }
-/** Default carrier cap for all HTTP RPC bodies. */
-const DEFAULT_MAX_REQUEST_BODY_BYTES = 32 * 1024 * 1024
 
 /** Services required before providing Connection; API Proxy is an optional `/api` fallback. */
 export const inject = ['httpServer']

packages/client/connection/tests/fixture.client.spec.ts

@@ -168,6 +168,13 @@ describe('createFixtureApi', () => {
         sessionStats: {
           turns: 0, steps: 0, llmMs: 0, toolMs: 0, ttftMs: 0, ttftSteps: 0, decodeMs: 0, decodeTokens: 0,
         },
+        imageLimits: {
+          maxImageBytes: 5 * 1024 * 1024,
+          maxImagesPerMessage: 20,
+          maxMessageImageBytes: 100 * 1024 * 1024,
+          maxImagePixels: 40_000_000,
+          mediaTypes: ['image/png', 'image/jpeg', 'image/webp', 'image/gif'],
+        },
       } },
     })
   })
@@ -357,7 +364,7 @@ describe('createFixtureApi', () => {
       const envelopes: RpcRequest<MuxFrame>[] = []
       for await (const envelope of api.events.mux(req({}), abort.signal)) {
         envelopes.push(envelope)
-        if (envelopes.length >= 12) abort.abort()
+        if (envelopes.length >= 13) abort.abort()
       }
       return envelopes
     }
@@ -380,10 +387,14 @@ describe('createFixtureApi', () => {
     expect((first[8]?.payload as { value: { messageTokens: number } }).value.messageTokens).toBeGreaterThan(0)
     expect(first[9]?.payload).toMatchObject({ type: 'session/projection', sessionId: 'fx-alpha', key: 'sessionStats' })
     expect((first[9]?.payload as { value: { turns: number; steps: number } }).value.steps).toBeGreaterThan(0)
-    expect(first[10]?.payload).toMatchObject({ type: 'approval/requested', toolName: 'dangerous_tool' })
-    expect(second[10]?.rpcId).toBe(first[10]?.rpcId) // stable rpcId across replays (host replay semantics)
-    expect(first[11]?.payload).toMatchObject({ type: 'question/requested', sessionId: 'fx-alpha' })
-    expect(second[11]?.rpcId).toBe(first[11]?.rpcId)
+    expect(first[10]?.payload).toMatchObject({
+      type: 'session/projection', sessionId: 'fx-alpha', key: 'imageLimits',
+      value: { maxImagesPerMessage: 20, maxImageBytes: 5 * 1024 * 1024 },
+    })
+    expect(first[11]?.payload).toMatchObject({ type: 'approval/requested', toolName: 'dangerous_tool' })
+    expect(second[11]?.rpcId).toBe(first[11]?.rpcId) // stable rpcId across replays (host replay semantics)
+    expect(first[12]?.payload).toMatchObject({ type: 'question/requested', sessionId: 'fx-alpha' })
+    expect(second[12]?.rpcId).toBe(first[12]?.rpcId)
   })
 
   it('steer with no replay in flight falls through to a fresh queued turn; non-text blocks stringify empty', async () => {