mslxdff 0.1.203 → 0.1.207

package/src/upstream-engine/sdk/responses.js

@@ -1,176 +1,176 @@
-// Responses API 适配器（@ai-sdk/openai）：models.dev 里模型级 npm=@ai-sdk/openai 的模型
-// （如 muse-spark*，打 /zen/v1/responses）。与 chat 适配器同契约：输出 OpenAI chat SSE 帧。
-// 复用 attempt.js 的 baseURL 解析 / 错误映射 / 流序列化，不复制实现。
-// 见 .scratch/ai-sdk-upstream/SPEC-p3-responses.md 与 docs/adr/0017。
-import { toModelPrompt, toModelTools, toModelToolChoice, toModelParams } from "./convert.js";
-import { sdkBaseFromUrl, sanitizeHeaders, errorResponseFromSdkError, streamResponseFromParts } from "./attempt.js";
-import { ENGINE_MARKER } from "./chat.js";
-
-export const RESPONSES_CHAT_PATH = "/zen/v1/responses";
-
-// 降级阈值：加密思考块（encrypted_content）由上游按 caller（出口）签发，客户端跨出口回传
-// （经组员/换代理/直连切换）会被拒 400 "reasoning `encrypted_content` was not issued to this caller"。
-// 命中即剥掉加密态重试一次；仍失败原样返回，交上层转组员接力（保底不变差）。
-const ENC_CALLER_400_RE = /encrypted_content[^"]*was not issued to this caller/i;
-
-export function isEncryptedCallerError(text) {
-  return typeof text === "string" && ENC_CALLER_400_RE.test(text);
-}
-
-// 上游 encrypted reasoning 只在 output_item.done 里给，而 AI SDK 仅在有 summary 文本时才透出到 parts
-// （muse-spark 这类无 summary 的思考模型会被吞掉）。这里在 fetch 层 tee 一份原始 SSE 自行解析，
-// 侧信道把加密思考交给序列化器，收尾帧补发。
-function captureReasoningFetch(baseFetch, sink) {
-  const dbg = process.env.MSLXDFF_RESPONSES_DEBUG === "1";
-  return async (input, init) => {
-    const res = await baseFetch(input, init);
-    const ct = res.headers.get("content-type") || "";
-    if (dbg) console.log(`[capture] enter ct=${ct} hasBody=${Boolean(res.body)}`);
-    if (!ct.includes("text/event-stream") || !res.body) return res;
-    const [passthrough, probe] = res.body.tee();
-    (async () => {
-      const reader = probe.getReader();
-      const dec = new TextDecoder();
-      let buf = "";
-      try {
-        for (;;) {
-          const { done, value } = await reader.read();
-          if (done) break;
-          buf += dec.decode(value, { stream: true });
-          let nl;
-          while ((nl = buf.indexOf("\n")) >= 0) {
-            const line = buf.slice(0, nl).trim();
-            buf = buf.slice(nl + 1);
-            if (!line.startsWith("data:")) continue;
-            const d = line.slice(5).trim();
-            if (!d || d === "[DONE]") continue;
-            try {
-              const j = JSON.parse(d);
-              if (dbg && j.type === "response.output_item.done") console.log(`[capture] item.done type=${j.item?.type} enc=${typeof j.item?.encrypted_content}`);
-              if (j.type === "response.output_item.done" && j.item?.type === "reasoning" && typeof j.item.encrypted_content === "string") {
-                sink.reasoning = { id: j.item.id, encrypted: j.item.encrypted_content, summary: j.item.summary ?? [] };
-                if (dbg) console.log(`[capture] hit id=${String(j.item.id).slice(0, 24)} len=${j.item.encrypted_content.length}`);
-              }
-            } catch { /* ignore */ }
-          }
-        }
-      } catch { /* ignore */ }
-      if (dbg) console.log(`[capture] stream end captured=${sink.reasoning ? "yes" : "no"}`);
-    })();
-    return new Response(passthrough, { status: res.status, statusText: res.statusText, headers: res.headers });
-  };
-}
-
-let sdkPromise = null;
-
-export function loadOpenAISdk() {
-  if (!sdkPromise) sdkPromise = import("@ai-sdk/openai");
-  return sdkPromise;
-}
-
-export async function attemptOnceResponsesSdk({
-  url,
-  body,
-  headers = {},
-  providerName = "opencode",
-  marker = null,
-  clock = Date.now,
-  sdkLoader = loadOpenAISdk,
-  fetchImpl,
-} = {}) {
-  const t0 = clock();
-  const baseURL = sdkBaseFromUrl(url, "/responses");
-  if (!baseURL) throw new Error(`sdk-channel: responsesPath 非 /responses，暂不支持（${url}）`);
-  let createOpenAI;
-  try {
-    ({ createOpenAI } = await sdkLoader());
-  } catch (e) {
-    const err = new Error(`sdk-channel: @ai-sdk/openai 不可用（需 Node>=18 且已安装）: ${e?.message || e}`);
-    err._sdkLoadFailed = true;
-    throw err;
-  }
-  // headers 由调用方构造（含 Authorization），headers 优先于 apiKey 默认头
-  const captured = { reasoning: null };
-  const baseFetch = fetchImpl ?? ((u, i) => globalThis.fetch(u, i));
-  const provider = createOpenAI({
-    name: providerName,
-    baseURL,
-    apiKey: "public",
-    headers: sanitizeHeaders(headers),
-    fetch: captureReasoningFetch(baseFetch, captured),
-  });
-  const model = provider.responses(String(body?.model || ""));
-  const params = toModelParams(body, providerName);
-  // 无状态 + 加密思考回传（thinking 跨轮）：上游把思考以加密块发回，客户端持有并每轮带回。
-  // AI SDK responses 的 providerOptionsName 对非 azure 硬编码为 "openai"（dist/index.js:5240）。
-  const providerOptions = {
-    ...(params.providerOptions || {}),
-    openai: {
-      store: false,
-      include: ["reasoning.encrypted_content"],
-      reasoningSummary: "auto",
-      ...((params.providerOptions || {}).openai || {}),
-    },
-  };
-  let res;
-  let encRetry = false;
-  const streamOnce = (prompt) => model.doStream({
-    prompt,
-    ...params,
-    providerOptions,
-    tools: toModelTools(body?.tools),
-    toolChoice: toModelToolChoice(body?.tool_choice),
-  });
-  try {
-    res = await streamOnce(toModelPrompt(body?.messages));
-  } catch (e) {
-    let mapped = errorResponseFromSdkError(e, { marker });
-    if (mapped && mapped.status === 400) {
-      let txt = "";
-      try { txt = await mapped.clone().text(); } catch {}
-      if (isEncryptedCallerError(txt)) {
-        try {
-          res = await streamOnce(toModelPrompt(body?.messages, { dropEncrypted: true }));
-          mapped = null;
-          encRetry = true;
-        } catch (e2) {
-          const mapped2 = errorResponseFromSdkError(e2, { marker });
-          if (mapped2) return mapped2;
-          throw e2;
-        }
-      }
-    }
-    if (mapped) return mapped;
-    // 非 HTTP 类错误（网络断/超时/无 statusCode）经 errorResponseFromSdkError 映射为 null：
-    // 必须 rethrow，否则 res 保持 undefined 掉到下面 res.stream，把真实网络错误伪装成
-    // "Cannot read properties of undefined" 的 TypeError 502（2026-10-07 muse-spark 实测）。
-    // 不可无条件 throw：400(encrypted) 剥态重试成功也走到这里（res 已赋值、mapped 置 null）。
-    if (res === undefined) throw e;
-  }
-  const out = streamResponseFromParts(res.stream, { marker, clock, t0, captured });
-  if (encRetry) { try { out._t.encRetry = true; } catch {} }
-  return out;
-}
-
-export function createSdkResponses({
-  baseUrl,
-  chatPath = RESPONSES_CHAT_PATH,
-  buildHeaders,
-  providerName = "opencode",
-  marker = ENGINE_MARKER,
-  fetchImpl,
-} = {}) {
-  const base = String(baseUrl || "").replace(/\/+$/, "");
-  return {
-    async chat(body) {
-      return attemptOnceResponsesSdk({
-        url: `${base}${chatPath}`,
-        body,
-        headers: buildHeaders ? buildHeaders(body) : {},
-        providerName,
-        marker,
-        fetchImpl,
-      });
-    },
-  };
-}
+// Responses API 适配器（@ai-sdk/openai）：models.dev 里模型级 npm=@ai-sdk/openai 的模型
+// （如 muse-spark*，打 /zen/v1/responses）。与 chat 适配器同契约：输出 OpenAI chat SSE 帧。
+// 复用 attempt.js 的 baseURL 解析 / 错误映射 / 流序列化，不复制实现。
+// 见 .scratch/ai-sdk-upstream/SPEC-p3-responses.md 与 docs/adr/0017。
+import { toModelPrompt, toModelTools, toModelToolChoice, toModelParams } from "./convert.js";
+import { sdkBaseFromUrl, sanitizeHeaders, errorResponseFromSdkError, streamResponseFromParts } from "./attempt.js";
+import { ENGINE_MARKER } from "./chat.js";
+
+export const RESPONSES_CHAT_PATH = "/zen/v1/responses";
+
+// 降级阈值：加密思考块（encrypted_content）由上游按 caller（出口）签发，客户端跨出口回传
+// （经组员/换代理/直连切换）会被拒 400 "reasoning `encrypted_content` was not issued to this caller"。
+// 命中即剥掉加密态重试一次；仍失败原样返回，交上层转组员接力（保底不变差）。
+const ENC_CALLER_400_RE = /encrypted_content[^"]*was not issued to this caller/i;
+
+export function isEncryptedCallerError(text) {
+  return typeof text === "string" && ENC_CALLER_400_RE.test(text);
+}
+
+// 上游 encrypted reasoning 只在 output_item.done 里给，而 AI SDK 仅在有 summary 文本时才透出到 parts
+// （muse-spark 这类无 summary 的思考模型会被吞掉）。这里在 fetch 层 tee 一份原始 SSE 自行解析，
+// 侧信道把加密思考交给序列化器，收尾帧补发。
+function captureReasoningFetch(baseFetch, sink) {
+  const dbg = process.env.MSLXDFF_RESPONSES_DEBUG === "1";
+  return async (input, init) => {
+    const res = await baseFetch(input, init);
+    const ct = res.headers.get("content-type") || "";
+    if (dbg) console.log(`[capture] enter ct=${ct} hasBody=${Boolean(res.body)}`);
+    if (!ct.includes("text/event-stream") || !res.body) return res;
+    const [passthrough, probe] = res.body.tee();
+    (async () => {
+      const reader = probe.getReader();
+      const dec = new TextDecoder();
+      let buf = "";
+      try {
+        for (;;) {
+          const { done, value } = await reader.read();
+          if (done) break;
+          buf += dec.decode(value, { stream: true });
+          let nl;
+          while ((nl = buf.indexOf("\n")) >= 0) {
+            const line = buf.slice(0, nl).trim();
+            buf = buf.slice(nl + 1);
+            if (!line.startsWith("data:")) continue;
+            const d = line.slice(5).trim();
+            if (!d || d === "[DONE]") continue;
+            try {
+              const j = JSON.parse(d);
+              if (dbg && j.type === "response.output_item.done") console.log(`[capture] item.done type=${j.item?.type} enc=${typeof j.item?.encrypted_content}`);
+              if (j.type === "response.output_item.done" && j.item?.type === "reasoning" && typeof j.item.encrypted_content === "string") {
+                sink.reasoning = { id: j.item.id, encrypted: j.item.encrypted_content, summary: j.item.summary ?? [] };
+                if (dbg) console.log(`[capture] hit id=${String(j.item.id).slice(0, 24)} len=${j.item.encrypted_content.length}`);
+              }
+            } catch { /* ignore */ }
+          }
+        }
+      } catch { /* ignore */ }
+      if (dbg) console.log(`[capture] stream end captured=${sink.reasoning ? "yes" : "no"}`);
+    })();
+    return new Response(passthrough, { status: res.status, statusText: res.statusText, headers: res.headers });
+  };
+}
+
+let sdkPromise = null;
+
+export function loadOpenAISdk() {
+  if (!sdkPromise) sdkPromise = import("@ai-sdk/openai");
+  return sdkPromise;
+}
+
+export async function attemptOnceResponsesSdk({
+  url,
+  body,
+  headers = {},
+  providerName = "opencode",
+  marker = null,
+  clock = Date.now,
+  sdkLoader = loadOpenAISdk,
+  fetchImpl,
+} = {}) {
+  const t0 = clock();
+  const baseURL = sdkBaseFromUrl(url, "/responses");
+  if (!baseURL) throw new Error(`sdk-channel: responsesPath 非 /responses，暂不支持（${url}）`);
+  let createOpenAI;
+  try {
+    ({ createOpenAI } = await sdkLoader());
+  } catch (e) {
+    const err = new Error(`sdk-channel: @ai-sdk/openai 不可用（需 Node>=18 且已安装）: ${e?.message || e}`);
+    err._sdkLoadFailed = true;
+    throw err;
+  }
+  // headers 由调用方构造（含 Authorization），headers 优先于 apiKey 默认头
+  const captured = { reasoning: null };
+  const baseFetch = fetchImpl ?? ((u, i) => globalThis.fetch(u, i));
+  const provider = createOpenAI({
+    name: providerName,
+    baseURL,
+    apiKey: "public",
+    headers: sanitizeHeaders(headers),
+    fetch: captureReasoningFetch(baseFetch, captured),
+  });
+  const model = provider.responses(String(body?.model || ""));
+  const params = toModelParams(body, providerName);
+  // 无状态 + 加密思考回传（thinking 跨轮）：上游把思考以加密块发回，客户端持有并每轮带回。
+  // AI SDK responses 的 providerOptionsName 对非 azure 硬编码为 "openai"（dist/index.js:5240）。
+  const providerOptions = {
+    ...(params.providerOptions || {}),
+    openai: {
+      store: false,
+      include: ["reasoning.encrypted_content"],
+      reasoningSummary: "auto",
+      ...((params.providerOptions || {}).openai || {}),
+    },
+  };
+  let res;
+  let encRetry = false;
+  const streamOnce = (prompt) => model.doStream({
+    prompt,
+    ...params,
+    providerOptions,
+    tools: toModelTools(body?.tools),
+    toolChoice: toModelToolChoice(body?.tool_choice),
+  });
+  try {
+    res = await streamOnce(toModelPrompt(body?.messages));
+  } catch (e) {
+    let mapped = errorResponseFromSdkError(e, { marker });
+    if (mapped && mapped.status === 400) {
+      let txt = "";
+      try { txt = await mapped.clone().text(); } catch {}
+      if (isEncryptedCallerError(txt)) {
+        try {
+          res = await streamOnce(toModelPrompt(body?.messages, { dropEncrypted: true }));
+          mapped = null;
+          encRetry = true;
+        } catch (e2) {
+          const mapped2 = errorResponseFromSdkError(e2, { marker });
+          if (mapped2) return mapped2;
+          throw e2;
+        }
+      }
+    }
+    if (mapped) return mapped;
+    // 非 HTTP 类错误（网络断/超时/无 statusCode）经 errorResponseFromSdkError 映射为 null：
+    // 必须 rethrow，否则 res 保持 undefined 掉到下面 res.stream，把真实网络错误伪装成
+    // "Cannot read properties of undefined" 的 TypeError 502（2026-10-07 muse-spark 实测）。
+    // 不可无条件 throw：400(encrypted) 剥态重试成功也走到这里（res 已赋值、mapped 置 null）。
+    if (res === undefined) throw e;
+  }
+  const out = streamResponseFromParts(res.stream, { marker, clock, t0, captured });
+  if (encRetry) { try { out._t.encRetry = true; } catch {} }
+  return out;
+}
+
+export function createSdkResponses({
+  baseUrl,
+  chatPath = RESPONSES_CHAT_PATH,
+  buildHeaders,
+  providerName = "opencode",
+  marker = ENGINE_MARKER,
+  fetchImpl,
+} = {}) {
+  const base = String(baseUrl || "").replace(/\/+$/, "");
+  return {
+    async chat(body) {
+      return attemptOnceResponsesSdk({
+        url: `${base}${chatPath}`,
+        body,
+        headers: buildHeaders ? buildHeaders(body) : {},
+        providerName,
+        marker,
+        fetchImpl,
+      });
+    },
+  };
+}