mslxdff 0.1.199 → 0.1.203

package/src/providers/zcode/sse.js

@@ -1,196 +1,196 @@
-// Anthropic SSE → OpenAI 形状转换（流式 / 聚合两种出口）。
-// 承诺字段：文本增量、thinking 增量（reasoning_content）、tool_use 增量（tool_calls）、usage、finish_reason。
-// thinking 跨轮回填：首个 thinking block 的 signature 随 reasoning_content 挂在 assistant 历史消息上（必须，
-// 上游要求 signature 与 thinking 原文一起回传；与 reasoning_content_signature 拼写不同的上游见下兼容）。
-const ENCODER = new TextEncoder();
-const DECODER = new TextDecoder();
-const STOP_REASONS = { end_turn: "stop", max_tokens: "length", tool_use: "tool_calls", stop_sequence: "stop", pause_turn: "stop" };
-const newId = () => `chatcmpl-zcode-${Math.random().toString(36).slice(2, 10)}${Date.now().toString(36)}`;
-
-function dataPayload(block) {
-  const lines = String(block || "").split("\n").filter((l) => l.startsWith("data:"));
-  if (!lines.length) return null;
-  const joined = lines.map((l) => l.slice(5).trimStart()).join("\n").trim();
-  if (!joined || joined === "[DONE]") return null;
-  try {
-    return JSON.parse(joined);
-  } catch {
-    return null;
-  }
-}
-
-export function createAnthropicTranslator({ model = "zcode", id = newId(), created = Math.floor(Date.now() / 1000) } = {}) {
-  let buffer = "";
-  let roleSent = false;
-  let tailSent = false;
-  let stopReason = null;
-  let error = null;
-  let inputTokens = null;
-  let outputTokens = null;
-  let text = "";
-  // thinking 全量文本（聚合出口回填 reasoning_content）与首个 signature（跨轮必需，见文件头）
-  let thinkingText = "";
-  let thinkingSig = null;
-  let toolCount = 0;
-  const toolIndexByBlock = new Map();
-  const toolCalls = [];
-
-  const usage = () => {
-    if (inputTokens == null && outputTokens == null) return undefined;
-    const prompt = inputTokens ?? 0;
-    const completion = outputTokens ?? 0;
-    return { prompt_tokens: prompt, completion_tokens: completion, total_tokens: prompt + completion };
-  };
-
-  const chunkLine = (delta, finish = null, u) => {
-    const chunk = { id, object: "chat.completion.chunk", created, model, choices: [{ index: 0, delta, finish_reason: finish }] };
-    if (u) chunk.usage = u;
-    return `data: ${JSON.stringify(chunk)}\n\n`;
-  };
-
-  function handle(evt) {
-    const out = [];
-    if (!evt || typeof evt !== "object") return out;
-    if (evt.type === "message_start") {
-      const it = evt.message?.usage?.input_tokens ?? evt.usage?.input_tokens;
-      if (it != null) inputTokens = Number(it) || 0;
-      if (!roleSent) {
-        roleSent = true;
-        out.push(chunkLine({ role: "assistant", content: "" }));
-      }
-    } else if (evt.type === "content_block_start") {
-      const block = evt.content_block || {};
-      if (block.type === "thinking") {
-        // 首个 thinking 块的 signature 留给聚合出口的跨轮回填（后续 thinking 块没有也必须回传空串占位——
-        // codearts 的 DeepSeek 同款要求，见 src/providers/codearts/chat.js）
-        if (thinkingSig === null && typeof block.signature === "string") thinkingSig = block.signature;
-        if (typeof block.thinking === "string" && block.thinking) thinkingText += block.thinking;
-      } else if (block.type === "tool_use") {
-        const idx = toolCount++;
-        toolIndexByBlock.set(Number(evt.index), idx);
-        toolCalls.push({ id: String(block.id || `toolu_${idx}`), name: String(block.name || ""), args: "" });
-        out.push(chunkLine({ tool_calls: [{ index: idx, id: String(block.id || ""), type: "function", function: { name: String(block.name || ""), arguments: "" } }] }));
-      }
-    } else if (evt.type === "content_block_delta") {
-      const d = evt.delta || {};
-      if (d.type === "text_delta" && typeof d.text === "string") {
-        text += d.text;
-        out.push(chunkLine({ content: d.text }));
-      } else if (d.type === "thinking_delta" && typeof d.thinking === "string") {
-        thinkingText += d.thinking;
-        if (typeof d.signature === "string" && d.signature && thinkingSig === null) thinkingSig = d.signature;
-        out.push(chunkLine({ reasoning_content: d.thinking }));
-      } else if (d.type === "input_json_delta" && typeof d.partial_json === "string") {
-        const idx = toolIndexByBlock.has(Number(evt.index)) ? toolIndexByBlock.get(Number(evt.index)) : Math.max(0, toolCount - 1);
-        if (toolCalls[idx]) toolCalls[idx].args += d.partial_json;
-        out.push(chunkLine({ tool_calls: [{ index: idx, function: { arguments: d.partial_json } }] }));
-      }
-    } else if (evt.type === "message_delta") {
-      const r = evt.delta?.stop_reason;
-      if (r) stopReason = STOP_REASONS[r] || "stop";
-      if (evt.usage?.output_tokens != null) outputTokens = Number(evt.usage.output_tokens) || 0;
-    } else if (evt.type === "message_stop") {
-      out.push(chunkLine({}, stopReason || "stop", usage()));
-      out.push("data: [DONE]\n\n");
-      tailSent = true;
-    } else if (evt.type === "error") {
-      error = { kind: "server", code: 0, message: String(evt.error?.message || "上游流内错误") };
-      out.push(`data: ${JSON.stringify({ error: { message: error.message, type: String(evt.error?.type || "upstream_error") } })}\n\n`);
-      out.push("data: [DONE]\n\n");
-      tailSent = true;
-    }
-    return out;
-  }
-
-  function feed(textChunk) {
-    buffer += String(textChunk ?? "").replace(/\r\n/g, "\n");
-    const out = [];
-    let idx;
-    while ((idx = buffer.indexOf("\n\n")) >= 0) {
-      const block = buffer.slice(0, idx);
-      buffer = buffer.slice(idx + 2);
-      const payload = dataPayload(block);
-      if (payload) out.push(...handle(payload));
-    }
-    return out;
-  }
-
-  function flush() {
-    if (tailSent) return [];
-    tailSent = true;
-    const out = [];
-    if (!roleSent) out.push(chunkLine({ role: "assistant", content: "" }));
-    out.push(chunkLine({}, stopReason || "stop", usage()));
-    out.push("data: [DONE]\n\n");
-    return out;
-  }
-
-  function result() {
-    const message = { role: "assistant", content: text || null };
-    // thinking 聚合回填：跨轮必需的 signature 与原文一起带回（见文件头；codearts DeepSeek 同款）
-    if (thinkingText) message.reasoning_content = thinkingText;
-    if (thinkingSig !== null) message.reasoning_content_signature = thinkingSig;
-    if (toolCalls.length) {
-      message.tool_calls = toolCalls.map((t) => ({ id: t.id, type: "function", function: { name: t.name, arguments: t.args || "{}" } }));
-      if (!text) message.content = null;
-    }
-    return {
-      error,
-      openAi: {
-        id,
-        object: "chat.completion",
-        created,
-        model,
-        choices: [{ index: 0, message, finish_reason: stopReason || "stop" }],
-        usage: usage() || { prompt_tokens: 0, completion_tokens: 0, total_tokens: 0 },
-      },
-    };
-  }
-
-  return { feed, flush, result };
-}
-
-export function anthropicToOpenAiStream(upstream, options = {}) {
-  const translator = createAnthropicTranslator(options);
-  if (!upstream || typeof upstream.getReader !== "function") {
-    return new ReadableStream({
-      start(controller) {
-        for (const s of [...translator.feed(String(upstream ?? "")), ...translator.flush()]) controller.enqueue(ENCODER.encode(s));
-        controller.close();
-      },
-    });
-  }
-  const reader = upstream.getReader();
-  return new ReadableStream({
-    async pull(controller) {
-      // 注意：pull 必须「有产出或收流」才返回——若本次不 enqueue，流规范不会再触发下一次 pull，
-      // 消费端 read() 会永久挂起（首个分片落在事件中间时必现）。故此处循环读到有产出为止。
-      for (;;) {
-        let next;
-        try {
-          next = await reader.read();
-        } catch {
-          next = { done: true };
-        }
-        if (next?.done) {
-          for (const s of translator.flush()) controller.enqueue(ENCODER.encode(s));
-          controller.close();
-          return;
-        }
-        const t = DECODER.decode(next.value, { stream: true });
-        const out = translator.feed(t);
-        if (out.length) {
-          for (const s of out) controller.enqueue(ENCODER.encode(s));
-          return;
-        }
-      }
-    },
-  });
-}
-
-export async function aggregateAnthropicToOpenAi(sseText, options = {}) {
-  const translator = createAnthropicTranslator(options);
-  translator.feed(String(sseText ?? ""));
-  translator.flush();
-  return translator.result();
-}
+// Anthropic SSE → OpenAI 形状转换（流式 / 聚合两种出口）。
+// 承诺字段：文本增量、thinking 增量（reasoning_content）、tool_use 增量（tool_calls）、usage、finish_reason。
+// thinking 跨轮回填：首个 thinking block 的 signature 随 reasoning_content 挂在 assistant 历史消息上（必须，
+// 上游要求 signature 与 thinking 原文一起回传；与 reasoning_content_signature 拼写不同的上游见下兼容）。
+const ENCODER = new TextEncoder();
+const DECODER = new TextDecoder();
+const STOP_REASONS = { end_turn: "stop", max_tokens: "length", tool_use: "tool_calls", stop_sequence: "stop", pause_turn: "stop" };
+const newId = () => `chatcmpl-zcode-${Math.random().toString(36).slice(2, 10)}${Date.now().toString(36)}`;
+
+function dataPayload(block) {
+  const lines = String(block || "").split("\n").filter((l) => l.startsWith("data:"));
+  if (!lines.length) return null;
+  const joined = lines.map((l) => l.slice(5).trimStart()).join("\n").trim();
+  if (!joined || joined === "[DONE]") return null;
+  try {
+    return JSON.parse(joined);
+  } catch {
+    return null;
+  }
+}
+
+export function createAnthropicTranslator({ model = "zcode", id = newId(), created = Math.floor(Date.now() / 1000) } = {}) {
+  let buffer = "";
+  let roleSent = false;
+  let tailSent = false;
+  let stopReason = null;
+  let error = null;
+  let inputTokens = null;
+  let outputTokens = null;
+  let text = "";
+  // thinking 全量文本（聚合出口回填 reasoning_content）与首个 signature（跨轮必需，见文件头）
+  let thinkingText = "";
+  let thinkingSig = null;
+  let toolCount = 0;
+  const toolIndexByBlock = new Map();
+  const toolCalls = [];
+
+  const usage = () => {
+    if (inputTokens == null && outputTokens == null) return undefined;
+    const prompt = inputTokens ?? 0;
+    const completion = outputTokens ?? 0;
+    return { prompt_tokens: prompt, completion_tokens: completion, total_tokens: prompt + completion };
+  };
+
+  const chunkLine = (delta, finish = null, u) => {
+    const chunk = { id, object: "chat.completion.chunk", created, model, choices: [{ index: 0, delta, finish_reason: finish }] };
+    if (u) chunk.usage = u;
+    return `data: ${JSON.stringify(chunk)}\n\n`;
+  };
+
+  function handle(evt) {
+    const out = [];
+    if (!evt || typeof evt !== "object") return out;
+    if (evt.type === "message_start") {
+      const it = evt.message?.usage?.input_tokens ?? evt.usage?.input_tokens;
+      if (it != null) inputTokens = Number(it) || 0;
+      if (!roleSent) {
+        roleSent = true;
+        out.push(chunkLine({ role: "assistant", content: "" }));
+      }
+    } else if (evt.type === "content_block_start") {
+      const block = evt.content_block || {};
+      if (block.type === "thinking") {
+        // 首个 thinking 块的 signature 留给聚合出口的跨轮回填（后续 thinking 块没有也必须回传空串占位——
+        // codearts 的 DeepSeek 同款要求，见 src/providers/codearts/chat.js）
+        if (thinkingSig === null && typeof block.signature === "string") thinkingSig = block.signature;
+        if (typeof block.thinking === "string" && block.thinking) thinkingText += block.thinking;
+      } else if (block.type === "tool_use") {
+        const idx = toolCount++;
+        toolIndexByBlock.set(Number(evt.index), idx);
+        toolCalls.push({ id: String(block.id || `toolu_${idx}`), name: String(block.name || ""), args: "" });
+        out.push(chunkLine({ tool_calls: [{ index: idx, id: String(block.id || ""), type: "function", function: { name: String(block.name || ""), arguments: "" } }] }));
+      }
+    } else if (evt.type === "content_block_delta") {
+      const d = evt.delta || {};
+      if (d.type === "text_delta" && typeof d.text === "string") {
+        text += d.text;
+        out.push(chunkLine({ content: d.text }));
+      } else if (d.type === "thinking_delta" && typeof d.thinking === "string") {
+        thinkingText += d.thinking;
+        if (typeof d.signature === "string" && d.signature && thinkingSig === null) thinkingSig = d.signature;
+        out.push(chunkLine({ reasoning_content: d.thinking }));
+      } else if (d.type === "input_json_delta" && typeof d.partial_json === "string") {
+        const idx = toolIndexByBlock.has(Number(evt.index)) ? toolIndexByBlock.get(Number(evt.index)) : Math.max(0, toolCount - 1);
+        if (toolCalls[idx]) toolCalls[idx].args += d.partial_json;
+        out.push(chunkLine({ tool_calls: [{ index: idx, function: { arguments: d.partial_json } }] }));
+      }
+    } else if (evt.type === "message_delta") {
+      const r = evt.delta?.stop_reason;
+      if (r) stopReason = STOP_REASONS[r] || "stop";
+      if (evt.usage?.output_tokens != null) outputTokens = Number(evt.usage.output_tokens) || 0;
+    } else if (evt.type === "message_stop") {
+      out.push(chunkLine({}, stopReason || "stop", usage()));
+      out.push("data: [DONE]\n\n");
+      tailSent = true;
+    } else if (evt.type === "error") {
+      error = { kind: "server", code: 0, message: String(evt.error?.message || "上游流内错误") };
+      out.push(`data: ${JSON.stringify({ error: { message: error.message, type: String(evt.error?.type || "upstream_error") } })}\n\n`);
+      out.push("data: [DONE]\n\n");
+      tailSent = true;
+    }
+    return out;
+  }
+
+  function feed(textChunk) {
+    buffer += String(textChunk ?? "").replace(/\r\n/g, "\n");
+    const out = [];
+    let idx;
+    while ((idx = buffer.indexOf("\n\n")) >= 0) {
+      const block = buffer.slice(0, idx);
+      buffer = buffer.slice(idx + 2);
+      const payload = dataPayload(block);
+      if (payload) out.push(...handle(payload));
+    }
+    return out;
+  }
+
+  function flush() {
+    if (tailSent) return [];
+    tailSent = true;
+    const out = [];
+    if (!roleSent) out.push(chunkLine({ role: "assistant", content: "" }));
+    out.push(chunkLine({}, stopReason || "stop", usage()));
+    out.push("data: [DONE]\n\n");
+    return out;
+  }
+
+  function result() {
+    const message = { role: "assistant", content: text || null };
+    // thinking 聚合回填：跨轮必需的 signature 与原文一起带回（见文件头；codearts DeepSeek 同款）
+    if (thinkingText) message.reasoning_content = thinkingText;
+    if (thinkingSig !== null) message.reasoning_content_signature = thinkingSig;
+    if (toolCalls.length) {
+      message.tool_calls = toolCalls.map((t) => ({ id: t.id, type: "function", function: { name: t.name, arguments: t.args || "{}" } }));
+      if (!text) message.content = null;
+    }
+    return {
+      error,
+      openAi: {
+        id,
+        object: "chat.completion",
+        created,
+        model,
+        choices: [{ index: 0, message, finish_reason: stopReason || "stop" }],
+        usage: usage() || { prompt_tokens: 0, completion_tokens: 0, total_tokens: 0 },
+      },
+    };
+  }
+
+  return { feed, flush, result };
+}
+
+export function anthropicToOpenAiStream(upstream, options = {}) {
+  const translator = createAnthropicTranslator(options);
+  if (!upstream || typeof upstream.getReader !== "function") {
+    return new ReadableStream({
+      start(controller) {
+        for (const s of [...translator.feed(String(upstream ?? "")), ...translator.flush()]) controller.enqueue(ENCODER.encode(s));
+        controller.close();
+      },
+    });
+  }
+  const reader = upstream.getReader();
+  return new ReadableStream({
+    async pull(controller) {
+      // 注意：pull 必须「有产出或收流」才返回——若本次不 enqueue，流规范不会再触发下一次 pull，
+      // 消费端 read() 会永久挂起（首个分片落在事件中间时必现）。故此处循环读到有产出为止。
+      for (;;) {
+        let next;
+        try {
+          next = await reader.read();
+        } catch {
+          next = { done: true };
+        }
+        if (next?.done) {
+          for (const s of translator.flush()) controller.enqueue(ENCODER.encode(s));
+          controller.close();
+          return;
+        }
+        const t = DECODER.decode(next.value, { stream: true });
+        const out = translator.feed(t);
+        if (out.length) {
+          for (const s of out) controller.enqueue(ENCODER.encode(s));
+          return;
+        }
+      }
+    },
+  });
+}
+
+export async function aggregateAnthropicToOpenAi(sseText, options = {}) {
+  const translator = createAnthropicTranslator(options);
+  translator.feed(String(sseText ?? ""));
+  translator.flush();
+  return translator.result();
+}

package/src/readline-compat.js

@@ -1,25 +1,25 @@
-// readline 统一走回调版 `node:readline`（question 包成 Promise），
-// 顺带给 -chat 一个人话版本门（项目要求 Node >=18，ADR-0024）。
-// 保留理由：回调版在 18+ 行为一致且无 import 期风险（`node:readline/promises`
-// 若被误用于旧运行时会直接 ERR_UNKNOWN_BUILTIN_MODULE，连报错都来不及打）。
-import readline from "node:readline";
-import { MIN_NODE_MAJOR, nodeMajor } from "./compat.js";
-
-export function createInterface(opts) {
-  const rl = readline.createInterface(opts);
-  const ask = rl.question.bind(rl);
-  rl.question = (query) => new Promise((resolve) => ask(query, (ans) => resolve(ans)));
-  return rl;
-}
-
-export { nodeMajor };
-
-// -chat 版本门：项目要求 Node >=18（ADR-0024），不满足直接给人话 + 升级指引。
-// 返回 true=通过，false=已打印升级指引。
-export function assertChatNode({ min = MIN_NODE_MAJOR } = {}) {
-  const major = nodeMajor();
-  if (major >= min) return true;
-  console.error(`Node 版本过旧（当前 v${process.versions.node}），-chat 需要 Node ${min}+（推荐 20+）。`);
-  console.error("先升级 Node 再重试：nvm install 20 && nvm use 20，或到 https://nodejs.org/ 下 LTS。");
-  return false;
-}
+// readline 统一走回调版 `node:readline`（question 包成 Promise），
+// 顺带给 -chat 一个人话版本门（项目要求 Node >=18，ADR-0024）。
+// 保留理由：回调版在 18+ 行为一致且无 import 期风险（`node:readline/promises`
+// 若被误用于旧运行时会直接 ERR_UNKNOWN_BUILTIN_MODULE，连报错都来不及打）。
+import readline from "node:readline";
+import { MIN_NODE_MAJOR, nodeMajor } from "./compat.js";
+
+export function createInterface(opts) {
+  const rl = readline.createInterface(opts);
+  const ask = rl.question.bind(rl);
+  rl.question = (query) => new Promise((resolve) => ask(query, (ans) => resolve(ans)));
+  return rl;
+}
+
+export { nodeMajor };
+
+// -chat 版本门：项目要求 Node >=18（ADR-0024），不满足直接给人话 + 升级指引。
+// 返回 true=通过，false=已打印升级指引。
+export function assertChatNode({ min = MIN_NODE_MAJOR } = {}) {
+  const major = nodeMajor();
+  if (major >= min) return true;
+  console.error(`Node 版本过旧（当前 v${process.versions.node}），-chat 需要 Node ${min}+（推荐 20+）。`);
+  console.error("先升级 Node 再重试：nvm install 20 && nvm use 20，或到 https://nodejs.org/ 下 LTS。");
+  return false;
+}