From 1419671f3fe5edbba76cb910735070d077f80750 Mon Sep 17 00:00:00 2001 From: Tianyi Cui <53024+tianyicui@users.noreply.github.com> Date: Tue, 11 Aug 2026 15:18:34 +0800 Subject: [PATCH] fix(session-export): wait for response pull capacity The ZIP loop checked desiredSize only after a push and responded to an overfull queue with setTimeout(0). A timer turn does not mean the consumer drained anything, so a slow or disconnected client still allowed the producer to enqueue the complete compressed archive while later artifact and attachment reads ran eagerly. Give the ReadableStream a 64 KiB byte queuing strategy and block the single producer on a pull-released capacity gate whenever desiredSize is non-positive. Cancellation wakes that gate through the existing producer signal; synchronous fflate output is therefore bounded to the queue high-water mark plus one input push. A regression test exhausts timer turns without consuming and proves the next media entry remains unread until response pulling begins, and the bilingual contracts now describe the real bound. --- ...026-08-10-web-session-log-export.i18n.yaml | 4 +- .../2026-08-10-web-session-log-export.md | 2 +- .../2026-08-10-web-session-log-export.zh.md | 2 +- packages/host/apiproxy/README.i18n.yaml | 4 +- packages/host/apiproxy/README.md | 2 +- packages/host/apiproxy/README.zh.md | 2 +- packages/host/apiproxy/src/session-export.ts | 90 ++++++++++++++----- .../apiproxy/tests/session-export.spec.ts | 34 ++++++- 8 files changed, 107 insertions(+), 33 deletions(-) diff --git a/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.i18n.yaml b/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.i18n.yaml index e2588d1529..937bb0df03 100644 --- a/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.i18n.yaml +++ b/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.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 .agents/notes/implemented/feature/2026-08-10-web-session-log-export.md -2026-08-10-web-session-log-export.md: 25aac00e1a3bd11d3f2a95770e520f53c348f8c2 -2026-08-10-web-session-log-export.zh.md: d00d3437ef53994d513a35829b9f5215ae5df417 +2026-08-10-web-session-log-export.md: 4568b5cf0e84a7efdf6e0e86d7e5955a2430f0f8 +2026-08-10-web-session-log-export.zh.md: 842330e30cc0a46579a823f80306ce88d6df1552 diff --git a/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.md b/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.md index 25aac00e1a..4568b5cf0e 100644 --- a/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.md +++ b/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.md @@ -10,7 +10,7 @@ The Trajectory view had no way to hand a debugging artifact to a human: the raw ## Decision -- **The export is a host-only download, not an RPC**: `GET /api/session.export?sessionId=…&includeDescendants=true` streams one ZIP attachment. Every file is a session's **stored artifact text verbatim**: `readRaw` on the persistence service reads the backend's own durable bytes (the JSONL backend decodes its physical zstd frames, or returns plaintext) — never a reconstruction from parsed events, so packed-chunk rows, key order, and line breaks survive byte-for-byte — under its original base name (`session.jsonl` at the root, `subagents//session.jsonl` for descendants). Compression runs on the host with fflate's streaming `Zip`/`ZipDeflate` API, each entry deflated in bounded chunks as it is produced, so the response is chunked as it is generated and the host never holds the whole archive in one buffer (at most one descendant's artifact text beyond the preloaded root), and production yields whenever the response queue fills, so a slow consumer bounds the accumulation (fflate's callback is synchronous — the drain point is the only backpressure). No manifest is written — every file is byte-identical to the durable artifact and self-describing through its own header line. +- **The export is a host-only download, not an RPC**: `GET /api/session.export?sessionId=…&includeDescendants=true` streams one ZIP attachment. Every file is a session's **stored artifact text verbatim**: `readRaw` on the persistence service reads the backend's own durable bytes (the JSONL backend decodes its physical zstd frames, or returns plaintext) — never a reconstruction from parsed events, so packed-chunk rows, key order, and line breaks survive byte-for-byte — under its original base name (`session.jsonl` at the root, `subagents//session.jsonl` for descendants). Compression runs on the host with fflate's streaming `Zip`/`ZipDeflate` API, each entry deflated in bounded chunks as it is produced, so the response is chunked as it is generated and the host never holds the whole archive in one buffer (at most one descendant's artifact text beyond the preloaded root). At the 64 KiB response byte high-water mark, production waits for consumer pull to restore capacity; fflate's synchronous callback can add at most one bounded input push beyond that queue bound. No manifest is written — every file is byte-identical to the durable artifact and self-describing through its own header line. - **Error vocabulary is HTTP-native**: missing services → 500, a backend without per-session raw artifacts → 501, missing root session → 404 (all decided before any byte streams), and a descendant without a stored artifact → the stream errors (fail-loud, never silent under-export). Request abort remains cancellation instead of being rewritten as 500; request and response-consumer cancellation converge on the producer signal, which reaches lineage and persistence reads and terminates the active compressor. The carrier (`toFetchHandler`) already applies the `/api` trust fence; the GET branch sits beside the existing SSE GET routes, and `ApiProxy.downloads.sessionLog` (host-only, no wire envelope, absent from `IApiClient`) implements it. - **The UI just downloads**: the 导出 button hands the endpoint directly to the browser's native download manager, so JavaScript neither fetches nor buffers the ZIP; the `session.log` RPC that an earlier iteration shipped was removed — the download endpoint is its only consumer, and the repo rule is no public interface without a current owner. The client bundle carries no archive implementation. - The 导出 button lives in the Trajectory toolbar; the plugin exposes `exportLog` through the view's inject face (components never touch ctx) and resolves the view tab label through the locale service (`轨迹` in Chinese, `Trajectory` in English). In-flight state disables the button during the handoff; a synchronous browser-handoff failure surfaces in a visible alert bar, while HTTP delivery is owned and reported by the browser. diff --git a/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.zh.md b/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.zh.md index d00d3437ef..842330e30c 100644 --- a/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.zh.md +++ b/.agents/notes/implemented/feature/2026-08-10-web-session-log-export.zh.md @@ -10,7 +10,7 @@ Trajectory 视图没有任何方式把调试工件交到人手里：原始会话 ## 决策 -- **导出是宿主侧的下载面，不是 RPC**：`GET /api/session.export?sessionId=…&includeDescendants=true` 流式返回一个 ZIP 附件。每个文件都是会话**存储工件的逐字原文**：持久化服务新增的 `readRaw` 读取后端自己的持久化字节（jsonl 后端解码其物理 zstd 帧，或直接返回明文）——绝非从解析后事件重建，因此 chunk 打包、键序、换行全部逐字节保留——放在其原始基础文件名下（根为 `session.jsonl`，子代理为 `subagents//session.jsonl`）。压缩在宿主侧用 fflate 的流式 `Zip`/`ZipDeflate` API 完成，每个条目按有界分块边产出边压缩，响应随生成分块写出，宿主从不把整个归档放进单个缓冲区（除预载的根外，最多同时持有一条后代的工件文本），且每当响应队列填满时生产会让出，慢消费者因此只产生有界的积压（fflate 的回调是同步的——让出点是唯一的背压手段）。不写清单——每个文件都与持久化工件逐字节一致，并通过自身 header 行自描述。 +- **导出是宿主侧的下载面，不是 RPC**：`GET /api/session.export?sessionId=…&includeDescendants=true` 流式返回一个 ZIP 附件。每个文件都是会话**存储工件的逐字原文**：持久化服务新增的 `readRaw` 读取后端自己的持久化字节（jsonl 后端解码其物理 zstd 帧，或直接返回明文）——绝非从解析后事件重建，因此 chunk 打包、键序、换行全部逐字节保留——放在其原始基础文件名下（根为 `session.jsonl`，子代理为 `subagents//session.jsonl`）。压缩在宿主侧用 fflate 的流式 `Zip`/`ZipDeflate` API 完成，每个条目按有界分块边产出边压缩，响应随生成分块写出，宿主从不把整个归档放进单个缓冲区（除预载的根外，最多同时持有一条后代的工件文本）。到达 64 KiB 响应字节高水位后，生产会等待 Consumer pull 恢复容量；fflate 的同步回调最多只会在该队列界限外再增加一次有界输入 push。不写清单——每个文件都与持久化工件逐字节一致，并通过自身 header 行自描述。 - **错误词汇是 HTTP 原生的**：服务缺失 → 500，后端不提供每会话原始工件 → 501，根会话缺失 → 404（三者都在任何字节流出前判定），后代缺少存储工件 → 流失败（fail-loud，绝不静默少导出）。请求中止会保持取消语义而不会改写成 500；请求取消与响应 Consumer 取消汇合到生产者 signal，该 signal 会传到血缘与持久化读取，并终止活跃压缩器。载体（`toFetchHandler`）已对 `/api` 应用信任围栏；GET 分支与既有 SSE GET 路由并列，由 `ApiProxy.downloads.sessionLog`（host-only、无 wire 信封、不在 `IApiClient` 上）实现。 - **UI 只负责下载**：「导出」按钮将端点直接交给浏览器原生下载管理器，因此 JavaScript 既不会 fetch 也不会缓冲 ZIP；早先迭代发布的 `session.log` RPC 已删除——下载端点是它唯一的消费者，仓库规则是不留无当前所有者的公共接口。客户端 bundle 不包含任何归档实现。 - 「导出」按钮位于 Trajectory 工具栏；插件通过视图的 inject face 暴露 `exportLog`（组件从不接触 ctx），并通过 locale 服务解析视图标签页标题（中文「轨迹」、英文 "Trajectory"）。进行中状态会在交接期间禁用按钮；同步的浏览器交接失败会在可见警示条中显示，而 HTTP 交付由浏览器负责并报告。 diff --git a/packages/host/apiproxy/README.i18n.yaml b/packages/host/apiproxy/README.i18n.yaml index bfafdbbc97..a0c5921be4 100644 --- a/packages/host/apiproxy/README.i18n.yaml +++ b/packages/host/apiproxy/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/host/apiproxy/README.md -README.md: c7b655816099d786a08ecfae6d794f35f2a9c8e3 -README.zh.md: 7577eb025fb84ac40de206e3ed780d92c2ab237a +README.md: 3c301b48cc92762fc1dff07a9442a1d48e66b1cc +README.zh.md: 79240941348783070b955162325fccf25c33aaae diff --git a/packages/host/apiproxy/README.md b/packages/host/apiproxy/README.md index c7b6558160..3c301b48cc 100644 --- a/packages/host/apiproxy/README.md +++ b/packages/host/apiproxy/README.md @@ -28,7 +28,7 @@ Question responses are validated against their pending request before the first `session.history`'s tail page (`beforeSeq` absent) additionally carries an optional `projections` block — the watermark snapshot of every unit registered on `ctx.sessionProjections` (`@deepseek-ai/dsh-session-projection`), with `asOfSeq` = the last event seq the values reflect (`-1` on an empty log). The gateway also subscribes to the registry's change feed and mints a `session/projection` mux frame per changed unit (`{sessionId, key, value, seq}` — live push state, never logged; clients hold one generic per-session value store under higher-seq-wins). The carrier holds zero domain knowledge (each value passed its unit's own schema inside the registry; the wire schemas keep `values`/`value` wide); loadOlder pages never carry the block, and a composition without the registry serves histories without either surface. -Session-log export is a host-only download surface, not an RPC: `GET /api/session.export?sessionId=…&includeDescendants=true` streams a ZIP whose files are each session's stored artifact text verbatim (the persistence backend's `readRaw` — exact durable bytes decoded from the physical encoding, never a reconstruction from parsed events), root under its original base name plus each subagent descendant under `subagents//`, and every image any included log references under `media/.` (read and verified from the attachment store; a shared image appears once). Each live root or descendant crosses the authoritative `SessionStore.flush` durability barrier immediately before its raw artifact read; cold sessions have no in-memory work to flush. Compression runs on the host with fflate's streaming Zip API, so the response is chunked as it is produced and the host never holds the whole archive in one buffer, and production yields whenever the response queue fills, so a slow consumer bounds the accumulation (fflate's callback is synchronous — the drain point is the only backpressure). Request abort and response-body cancellation stop lineage and artifact work, terminate the active compressor, and propagate as cancellation rather than an HTTP 500. It requires the persistence, session-query, and attachment services: a deployment without any answers 500, a persistence backend without per-session raw artifacts answers 501, a missing root session answers 404, and a descendant without a stored artifact or a referenced image that cannot be read fails the stream (fail-loud, never silent under-export). The carrier mounts the endpoint; `ApiProxy.downloads.sessionLog` implements it. +Session-log export is a host-only download surface, not an RPC: `GET /api/session.export?sessionId=…&includeDescendants=true` streams a ZIP whose files are each session's stored artifact text verbatim (the persistence backend's `readRaw` — exact durable bytes decoded from the physical encoding, never a reconstruction from parsed events), root under its original base name plus each subagent descendant under `subagents//`, and every image any included log references under `media/.` (read and verified from the attachment store; a shared image appears once). Each live root or descendant crosses the authoritative `SessionStore.flush` durability barrier immediately before its raw artifact read; cold sessions have no in-memory work to flush. Compression runs on the host with fflate's streaming Zip API, so the response is chunked as it is produced and the host never holds the whole archive in one buffer. Once the response queue reaches its 64 KiB byte high-water mark, production waits until consumer pull restores positive capacity; fflate's synchronous callback can overshoot that bound only by the output of one bounded input push. Request abort and response-body cancellation stop lineage and artifact work, terminate the active compressor, and propagate as cancellation rather than an HTTP 500. It requires the persistence, session-query, and attachment services: a deployment without any answers 500, a persistence backend without per-session raw artifacts answers 501, a missing root session answers 404, and a descendant without a stored artifact or a referenced image that cannot be read fails the stream (fail-loud, never silent under-export). The carrier mounts the endpoint; `ApiProxy.downloads.sessionLog` implements it. Session titles ride the generic projection pair like every other domain — the history-tail `projections` block plus `session/projection` frames under the `title` key. Titles do not join `session.list`; cold sessions remain metadata-only there until opening or resuming attaches their logs. `session.rename` accepts an explicit user title (resuming a cold session first), delegating to `ctx.sessionTitle.rename` — the accepted `session/title` event pins the title against automatic regeneration — and returns the normalized title plus its event seq so a client settles its `title` projection cell ahead of the push frame; a title that normalizes to empty returns `title-invalid`. diff --git a/packages/host/apiproxy/README.zh.md b/packages/host/apiproxy/README.zh.md index 7577eb025f..7924094134 100644 --- a/packages/host/apiproxy/README.zh.md +++ b/packages/host/apiproxy/README.zh.md @@ -28,7 +28,7 @@ Settings 分节中的 `reasoningEffort` 在 agent-default-model 插件配置中 `session.history` 的尾页（不带 `beforeSeq`）额外携带一个可选的 `projections` 块——`ctx.sessionProjections`（`@deepseek-ai/dsh-session-projection`）上每个已注册单元的水位线快照，`asOfSeq` = 这些值共同反映到的最后一个事件 seq（空日志为 `-1`）。网关还订阅注册表的变更流，为每个状态发生变化的单元生成一个 `session/projection` mux 帧（`{sessionId, key, value, seq}`——实时推送状态，绝不入日志；客户端按 seq 高者胜维护一个按会话的通用值仓）。载体不持有任何领域知识（每个值在注册表内部已过其单元自己的 schema；协议 schema 对 `values`/`value` 保持宽松）；loadOlder 页永不携带该块，未装注册表的组合则两个面都不提供。 -会话日志导出是宿主侧的下载面，不是 RPC：`GET /api/session.export?sessionId=…&includeDescendants=true` 流式返回一个 ZIP，其中每个文件都是会话存储工件的逐字原文（持久化后端的 `readRaw`——按物理编码解码的确切持久化字节，绝非从解析后事件重建），根会话放在其原始基础文件名下，每个子代理后代放在 `subagents//` 下，每个被任何包含的日志引用的图片放在 `media/.` 下（从附件存储读取并校验；共享图片只出现一次）。每个实时根会话或后代都会在读取原始工件前立即通过权威的 `SessionStore.flush` 持久性屏障；冷会话没有需要 flush 的内存工作。压缩在宿主侧用 fflate 的流式 Zip API 完成，响应边生成边分块写出，宿主从不把整个归档放进单个缓冲区，且每当响应队列填满时生产会让出，慢消费者因此只产生有界的积压（fflate 的回调是同步的——让出点是唯一的背压手段）。请求中止或响应 body 取消会停止血缘与工件工作、终止活跃压缩器，并继续按取消传播，而不会变成 HTTP 500。它要求同时挂载持久化、session-query 与附件服务：任一缺失应答 500，持久化后端不提供每会话原始工件时应答 501，根会话缺失时应答 404，后代缺少存储工件或引用的图片无法读取则整个流失败（fail-loud，绝不静默少导出）。端点由传输层挂载，`ApiProxy.downloads.sessionLog` 实现它。 +会话日志导出是宿主侧的下载面，不是 RPC：`GET /api/session.export?sessionId=…&includeDescendants=true` 流式返回一个 ZIP，其中每个文件都是会话存储工件的逐字原文（持久化后端的 `readRaw`——按物理编码解码的确切持久化字节，绝非从解析后事件重建），根会话放在其原始基础文件名下，每个子代理后代放在 `subagents//` 下，每个被任何包含的日志引用的图片放在 `media/.` 下（从附件存储读取并校验；共享图片只出现一次）。每个实时根会话或后代都会在读取原始工件前立即通过权威的 `SessionStore.flush` 持久性屏障；冷会话没有需要 flush 的内存工作。压缩在宿主侧用 fflate 的流式 Zip API 完成，响应边生成边分块写出，宿主从不把整个归档放进单个缓冲区。响应队列达到 64 KiB 字节高水位后，生产会等待 Consumer pull 恢复正容量；fflate 的同步回调最多只会让该界限多出一次有界输入 push 的输出。请求中止或响应 body 取消会停止血缘与工件工作、终止活跃压缩器，并继续按取消传播，而不会变成 HTTP 500。它要求同时挂载持久化、session-query 与附件服务：任一缺失应答 500，持久化后端不提供每会话原始工件时应答 501，根会话缺失时应答 404，后代缺少存储工件或引用的图片无法读取则整个流失败（fail-loud，绝不静默少导出）。端点由传输层挂载，`ApiProxy.downloads.sessionLog` 实现它。 会话标题与其他所有领域一样搭乘这对通用投影机制——历史尾页的 `projections` 块外加 `title` 键下的 `session/projection` 帧。标题不会加入 `session.list`；冷会话在其中仍只有元数据，直到打开或恢复操作附加其日志。`session.rename` 接受用户显式标题（冷会话先恢复），委托给 `ctx.sessionTitle.rename`——被接受的 `session/title` 事件将标题钉住、不再被自动生成覆盖——并返回规范化后的标题及其事件 seq，让 client 在推送帧到达前就结算自己的 `title` 投影格；规范化后为空的标题返回 `title-invalid`。 diff --git a/packages/host/apiproxy/src/session-export.ts b/packages/host/apiproxy/src/session-export.ts index be9bdd41ee..621bbfe6f5 100644 --- a/packages/host/apiproxy/src/session-export.ts +++ b/packages/host/apiproxy/src/session-export.ts @@ -13,10 +13,9 @@ * the active compressor. * Compression runs on the host with fflate's streaming Zip API, so the archive * bytes are produced incrementally and the host never holds the whole archive - * in one buffer; production yields to the consumer whenever the response queue - * fills past its high-water mark, so a slow consumer bounds the accumulation - * instead of piling up the whole archive (fflate's callback is synchronous — - * this drain point is the only backpressure available). + * in one buffer; production waits for consumer pull whenever the response queue + * reaches its byte high-water mark, so a slow consumer bounds accumulation to + * the configured queue plus one synchronous fflate push. * @module */ @@ -266,30 +265,66 @@ const PUSH_CHUNK_CODE_UNITS = 1 << 16 /** How many bytes of media one zip push carries (bounded memory; images are already size-capped). */ const PUSH_CHUNK_BYTES = 1 << 16 +/** Byte capacity retained by the response stream before ZIP production waits for pull. */ +const RESPONSE_HIGH_WATER_MARK_BYTES = 1 << 16 + +/** One producer waiter released only when ReadableStream pull restores capacity. */ +class ResponseCapacityGate { + private releasePending: (() => void) | undefined + + /** + * Wait until the response queue has positive byte capacity or cancellation wins. + * @param controller - response controller whose desired size owns capacity. + * @param signal - combined request/consumer cancellation. + */ + async wait( + controller: ReadableStreamDefaultController, + signal: AbortSignal, + ): Promise { + signal.throwIfAborted() + if (controller.desiredSize === null || controller.desiredSize > 0) return + await new Promise((resolve) => { + const release = (): void => { + this.releasePending = undefined + signal.removeEventListener('abort', release) + resolve() + } + this.releasePending = release + signal.addEventListener('abort', release, { once: true }) + }) + signal.throwIfAborted() + } + + /** Release the current producer waiter after a consumer pull. */ + pulled(): void { + this.releasePending?.() + } +} + /** * Push one media object's bytes into a deflate stream in bounded chunks, - * yielding to a slow consumer between chunks like the artifact path does. + * waiting for consumer capacity between chunks like the artifact path does. * @param deflate - the zip entry's deflate stream. * @param data - the stored image bytes. - * @param signal - optional cancellation; throws when aborted. + * @param controller - response queue controller. + * @param capacity - pull-driven response-capacity gate. + * @param signal - cancellation; throws when aborted. */ async function pushBinaryChunks( deflate: ZipDeflate, data: Uint8Array, controller: ReadableStreamDefaultController, - signal?: AbortSignal, + capacity: ResponseCapacityGate, + signal: AbortSignal, ): Promise { let offset = 0 do { - signal?.throwIfAborted() + signal.throwIfAborted() const end = Math.min(offset + PUSH_CHUNK_BYTES, data.byteLength) const finalChunk = end >= data.byteLength deflate.push(data.subarray(offset, end), finalChunk) offset = end - /* v8 ignore next 2 -- only fires when a slow consumer leaves the queue over-full */ - if (controller.desiredSize !== null && controller.desiredSize < 0) { - await new Promise(resolve => setTimeout(resolve, 0)) - } + await capacity.wait(controller, signal) } while (offset < data.byteLength) } @@ -299,19 +334,22 @@ async function pushBinaryChunks( * re-encodes as U+FFFD and would silently corrupt the exported artifact). * @param deflate - the zip entry's deflate stream. * @param content - the artifact text verbatim. - * @param signal - optional cancellation; throws when aborted. + * @param controller - response queue controller. + * @param capacity - pull-driven response-capacity gate. + * @param signal - cancellation; throws when aborted. */ async function pushArtifactChunks( deflate: ZipDeflate, content: string, controller: ReadableStreamDefaultController, - signal?: AbortSignal, + capacity: ResponseCapacityGate, + signal: AbortSignal, ): Promise { const encoder = new TextEncoder() let offset = 0 let finalChunk: boolean do { - signal?.throwIfAborted() + signal.throwIfAborted() let end = Math.min(offset + PUSH_CHUNK_CODE_UNITS, content.length) if (end < content.length && end - offset > 1) { // Back off one code unit when the boundary lands inside a surrogate @@ -322,10 +360,7 @@ async function pushArtifactChunks( finalChunk = end >= content.length deflate.push(encoder.encode(content.slice(offset, end)), finalChunk) offset = end - /* v8 ignore next 2 -- only fires when a slow consumer leaves the queue over-full */ - if (controller.desiredSize !== null && controller.desiredSize < 0) { - await new Promise(resolve => setTimeout(resolve, 0)) - } + await capacity.wait(controller, signal) } while (!finalChunk) } @@ -354,6 +389,7 @@ export function streamSessionLogZip( const producerSignal = AbortSignal.any([signal, consumerAbort.signal]) let zip: Zip | undefined let zipTerminated = false + const capacity = new ResponseCapacityGate() const terminateZip = (): void => { if (zip === undefined || zipTerminated) return zipTerminated = true @@ -362,9 +398,9 @@ export function streamSessionLogZip( return new ReadableStream({ start(controller) { // fflate invokes the callback synchronously per compressed chunk, so a - // single push can enqueue ahead of a slow consumer; pushArtifactChunks - // yields between chunks once the queue is over-full, bounding the - // accumulation to the queue high-water mark plus one push. + // single push can enqueue ahead of a slow consumer; the capacity gate + // waits for pull between pushes once the byte queue is full, bounding + // accumulation to the queue high-water mark plus one synchronous push. const archive = new Zip((error, data, final) => { /* v8 ignore next 3 -- fflate reports only internal zip failures, unreachable for valid inputs */ if (error) { @@ -382,9 +418,9 @@ export function streamSessionLogZip( const deflate = new ZipDeflate(entry.path, { level: 6 }) archive.add(deflate) if ('content' in entry) { - await pushArtifactChunks(deflate, entry.content, controller, producerSignal) + await pushArtifactChunks(deflate, entry.content, controller, capacity, producerSignal) } else { - await pushBinaryChunks(deflate, entry.data, controller, producerSignal) + await pushBinaryChunks(deflate, entry.data, controller, capacity, producerSignal) } } archive.end() @@ -397,11 +433,17 @@ export function streamSessionLogZip( } })() }, + pull() { + capacity.pulled() + }, cancel(reason) { consumerAbort.abort( reason instanceof Error ? reason : new Error('session log export stream cancelled'), ) terminateZip() }, + }, { + highWaterMark: RESPONSE_HIGH_WATER_MARK_BYTES, + size: chunk => chunk.byteLength, }) } diff --git a/packages/host/apiproxy/tests/session-export.spec.ts b/packages/host/apiproxy/tests/session-export.spec.ts index 3aa9b1a9d2..5a124db86c 100644 --- a/packages/host/apiproxy/tests/session-export.spec.ts +++ b/packages/host/apiproxy/tests/session-export.spec.ts @@ -5,7 +5,8 @@ * root → 404, missing descendant → errored stream). */ -import { describe, expect, it } from 'vitest' +import { randomBytes } from 'node:crypto' +import { describe, expect, it, vi } from 'vitest' import { Context } from '@deepseek-ai/cordis' import { unzipSync, strFromU8 } from 'fflate' import type { ImageAttachmentRef } from '@deepseek-ai/dsh-attachment' @@ -286,6 +287,37 @@ describe('session.export download endpoint', () => { expect(strFromU8(files['session.jsonl'] as Uint8Array)).toBe(root.content) }) + it('waits for response pull capacity before reading the next archive entry', async () => { + const root = artifact('session-root', undefined, [ + imageEventLine('after-root'), + randomBytes(512 * 1024).toString('base64'), + ].join('\n')) + let imageReads = 0 + const api = await buildApi({ 'session-root': root }, [], { + attachments: async (ref) => { + imageReads += 1 + return storedImage(String(ref.attachmentId), ref.mediaType) + }, + }) + vi.useFakeTimers() + let response: Response | undefined + try { + response = await toFetchHandler(api).fetch( + new Request('http://host/api/session.export?sessionId=session-root'), + ) + // Exhausting timer turns must not advance a producer whose byte queue is + // full; only a consumer pull can release it. + await vi.runAllTimersAsync() + expect(imageReads).toBe(0) + } finally { + vi.useRealTimers() + } + if (response === undefined) throw new Error('missing export response') + const files = unzipSync(await responseBytes(response)) + expect(imageReads).toBe(1) + expect(files['media/after-root.png']).toEqual(storedImage('after-root').data) + }) + it('exports an empty artifact as an empty zip entry', async () => { const root = { ...artifact('session-root'), content: '' } const api = await buildApi({ 'session-root': root })