clawmux 0.3.13 → 0.3.15

package/README.md

@@ -83,6 +83,8 @@ ClawMux registers as a single provider `clawmux` in OpenClaw with model `auto`.
 openclaw provider clawmux
 ```
 
+`clawmux init` manages the `clawmux` provider entry in `openclaw.json` (and the per-agent `models.json` caches) for you. It sets `api` to match your MEDIUM tier's API format and computes the correct `baseUrl` from that — `http://localhost:<port>/v1` for OpenAI-style APIs (where the upstream SDK appends `/chat/completions` or `/responses`) and `http://localhost:<port>` for everything else. Do not hand-edit these fields; rerun `clawmux init` after changing the MEDIUM model in `~/.openclaw/clawmux.json`.
+
 ## How It Works
 
 ```
@@ -109,7 +111,7 @@ ClawMux resolves each model's context window using this priority chain:
 
 1. **~/.openclaw/clawmux.json** `routing.contextWindows` — explicit per-model override
 2. **openclaw.json** `models.providers[provider].models[].contextWindow` — user config
-3. **OpenClaw built-in catalog** — pi-ai model database (812+ models)
+3. **OpenClaw built-in catalog** — pi-ai model database (830+ models)
 4. **Default: 200,000 tokens**
 
 Compression threshold uses the **minimum** context window across all routing models, since compression happens before routing decides which model to use.

package/dist/cli.cjs

@@ -3,6 +3,7 @@ var __create = Object.create;
 var __getProtoOf = Object.getPrototypeOf;
 var __defProp = Object.defineProperty;
 var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
 var __hasOwnProp = Object.prototype.hasOwnProperty;
 function __accessProp(key) {
   return this[key];
@@ -29,6 +30,23 @@ var __toESM = (mod, isNodeMode, target) => {
     cache.set(mod, to);
   return to;
 };
+var __toCommonJS = (from) => {
+  var entry = (__moduleCache ??= new WeakMap).get(from), desc;
+  if (entry)
+    return entry;
+  entry = __defProp({}, "__esModule", { value: true });
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (var key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(entry, key))
+        __defProp(entry, key, {
+          get: __accessProp.bind(from, key),
+          enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable
+        });
+  }
+  __moduleCache.set(from, entry);
+  return entry;
+};
+var __moduleCache;
 var __commonJS = (cb, mod) => () => (mod || cb((mod = { exports: {} }).exports, mod), mod.exports);
 var __returnValue = (v) => v;
 function __exportSetter(name, newValue) {
@@ -112214,13 +112232,18 @@ var require_dist4 = __commonJS((exports2, module2) => {
 });
 
 // src/cli.ts
+var exports_cli = {};
+__export(exports_cli, {
+  resolveProviderBaseUrl: () => resolveProviderBaseUrl
+});
+module.exports = __toCommonJS(exports_cli);
 var import_promises8 = require("node:fs/promises");
 var import_node_path7 = require("node:path");
 var import_node_child_process = require("node:child_process");
 var import_node_os2 = require("node:os");
 
 // src/proxy/router.ts
-var VERSION = process.env.npm_package_version ?? "0.3.13";
+var VERSION = process.env.npm_package_version ?? "0.3.15";
 function jsonResponse(body, status = 200) {
   return new Response(JSON.stringify(body), {
     status,
@@ -115091,6 +115114,9 @@ function djb2Hash(text) {
   return (hash >>> 0).toString(16);
 }
 function extractText(content) {
+  if (content === null || content === undefined) {
+    return "";
+  }
   if (typeof content === "string") {
     return content.slice(0, 200);
   }
@@ -115103,6 +115129,9 @@ function extractText(content) {
   return concatenated.slice(0, 200);
 }
 function contentLength(content) {
+  if (content === null || content === undefined) {
+    return 0;
+  }
   if (typeof content === "string") {
     return content.length;
   }
@@ -115178,7 +115207,7 @@ class EscalationMemory {
 }
 
 // src/routing/instruction-injector.ts
-var INJECT_FOR_TIERS = new Set(["LIGHT"]);
+var INJECT_FOR_TIERS = new Set(["LIGHT", "MEDIUM"]);
 var ESCALATION_INSTRUCTION = `If you cannot handle this request fully (due to complexity, missing context, or capability limits), output EXACTLY the following marker with no other text on that line: ${ESCALATE_SIGNAL}
 Do not explain. Do not ask permission. Just emit the marker and stop.`;
 function injectEscalationInstruction(messages) {
@@ -115194,6 +115223,12 @@ function injectEscalationInstruction(messages) {
       ...rest
     ];
   }
+  if (first.content === null) {
+    return [
+      { role: "system", content: ESCALATION_INSTRUCTION },
+      ...rest
+    ];
+  }
   if (typeof first.content === "string") {
     return [
       { role: "system", content: first.content + `
@@ -119730,8 +119765,7 @@ function piStreamToAnthropicSse(piStream) {
     }
   });
 }
-async function piStreamToAnthropicJson(piStream) {
-  const msg = await piStream.result();
+function anthropicMessageFromAssistant(msg) {
   if ((msg.stopReason === "error" || msg.stopReason === "aborted") && msg.errorMessage) {
     console.error(`[clawmux] pi-ai upstream error (${msg.stopReason}): ${msg.errorMessage}`);
   }
@@ -119769,6 +119803,10 @@ async function piStreamToAnthropicJson(piStream) {
     ...msg.errorMessage ? { error_message: msg.errorMessage } : {}
   };
 }
+async function piStreamToAnthropicJson(piStream) {
+  const msg = await piStream.result();
+  return anthropicMessageFromAssistant(msg);
+}
 
 // src/pi-bridge/event-to-openai-completions.ts
 var encoder3 = new TextEncoder;
@@ -119992,8 +120030,7 @@ function piStreamToOpenAiCompletionsSse(piStream) {
     }
   });
 }
-async function piStreamToOpenAiCompletionsJson(piStream) {
-  const msg = await piStream.result();
+function openAiCompletionsMessageFromAssistant(msg) {
   let textContent = "";
   const toolCalls = [];
   let reasoning = "";
@@ -120043,6 +120080,10 @@ async function piStreamToOpenAiCompletionsJson(piStream) {
     }
   };
 }
+async function piStreamToOpenAiCompletionsJson(piStream) {
+  const msg = await piStream.result();
+  return openAiCompletionsMessageFromAssistant(msg);
+}
 
 // src/pi-bridge/event-to-openai-responses.ts
 var encoder4 = new TextEncoder;
@@ -120205,8 +120246,7 @@ function piStreamToOpenAiResponsesSse(piStream) {
     }
   });
 }
-async function piStreamToOpenAiResponsesJson(piStream) {
-  const msg = await piStream.result();
+function openAiResponsesMessageFromAssistant(msg) {
   const output = [];
   for (const c of msg.content) {
     if (c.type === "text") {
@@ -120247,6 +120287,10 @@ async function piStreamToOpenAiResponsesJson(piStream) {
     }
   };
 }
+async function piStreamToOpenAiResponsesJson(piStream) {
+  const msg = await piStream.result();
+  return openAiResponsesMessageFromAssistant(msg);
+}
 
 // src/pi-bridge/event-to-google.ts
 var encoder5 = new TextEncoder;
@@ -120363,8 +120407,7 @@ function piStreamToGoogleSse(piStream) {
     }
   });
 }
-async function piStreamToGoogleJson(piStream) {
-  const msg = await piStream.result();
+function googleMessageFromAssistant(msg) {
   const parts = [];
   for (const c of msg.content) {
     if (c.type === "text") {
@@ -120396,6 +120439,10 @@ async function piStreamToGoogleJson(piStream) {
     modelVersion: msg.model || ""
   };
 }
+async function piStreamToGoogleJson(piStream) {
+  const msg = await piStream.result();
+  return googleMessageFromAssistant(msg);
+}
 
 // src/proxy/pipeline.ts
 registerAdapter(new AnthropicAdapter);
@@ -120632,35 +120679,12 @@ async function handleApiRequest(req, body, apiType, config, openclawConfig, auth
         }
         const detector = signalRouter.createSignalDetector();
         const detectionState = createSignalDetectionState();
-        if (wantsStream) {
-          const signalGen = detectSignalInStream(piStreamHandle, detector, detectionState, () => {});
-          const sseBody = piStreamToAnthropicSseFromGenerator(signalGen);
-          const response = new Response(sseBody, {
-            status: 200,
-            headers: { "content-type": "text/event-stream" }
-          });
-          response.clone().text().then(() => {
-            if (detectionState.signalDetected) {
-              abortCtrl.abort();
-              const nextTier = signalRouter.handleEscalation(messages, currentTier);
-              if (nextTier !== null) {
-                console.log(`[clawmux] [escalation] ${currentTier} → ${nextTier} (signal detected)`);
-                currentTier = nextTier;
-                return;
-              }
-            }
-            signalRouter.touchActivity(messages);
-            if (attempt > 0) {
-              signalRouter.recordSuccessfulEscalation(messages, currentTier);
-            }
-          });
-          return response;
-        }
-        const fullText = await collectPiStreamText(piStreamHandle, detector, detectionState);
+        const assistantMsg = await collectPiStreamWithSignalDetection(piStreamHandle, detector, detectionState);
         if (detectionState.signalDetected) {
+          abortCtrl.abort();
           const nextTier = signalRouter.handleEscalation(messages, currentTier);
           if (nextTier !== null) {
-            console.log(`[clawmux] [escalation] ${currentTier} → ${nextTier} (signal detected in non-streaming)`);
+            console.log(`[clawmux] [escalation] ${currentTier} → ${nextTier} (signal detected)`);
             currentTier = nextTier;
             continue;
           }
@@ -120669,7 +120693,15 @@ async function handleApiRequest(req, body, apiType, config, openclawConfig, auth
         if (attempt > 0) {
           signalRouter.recordSuccessfulEscalation(messages, currentTier);
         }
-        return new Response(JSON.stringify(fullText), {
+        if (wantsStream) {
+          const sseBody = buildSseForApiType(apiType, replayAssistantMessageAsEvents(assistantMsg));
+          return new Response(sseBody, {
+            status: 200,
+            headers: { "content-type": "text/event-stream" }
+          });
+        }
+        const jsonBody = buildJsonForApiType(apiType, assistantMsg);
+        return new Response(JSON.stringify(jsonBody), {
           status: 200,
           headers: { "content-type": "application/json" }
         });
@@ -120915,164 +120947,63 @@ async function yieldPiAiResponse(piStreamHandle, apiType, wantsStream) {
   }
   throw new Error(`Unsupported pi-ai apiType: ${apiType}`);
 }
-function piStreamToAnthropicSseFromGenerator(gen) {
-  const encoder6 = new TextEncoder;
-  function sseFrame2(event, data) {
-    return encoder6.encode(`event: ${event}
-data: ${JSON.stringify(data)}
-
-`);
-  }
-  return new ReadableStream({
-    async start(controller) {
-      try {
-        let messageStarted = false;
-        const openBlocks = new Map;
-        const ensureMessageStart = (model) => {
-          if (messageStarted)
-            return;
-          messageStarted = true;
-          controller.enqueue(sseFrame2("message_start", {
-            type: "message_start",
-            message: {
-              id: "msg_" + Date.now().toString(36),
-              type: "message",
-              role: "assistant",
-              content: [],
-              model,
-              stop_reason: null,
-              stop_sequence: null,
-              usage: { input_tokens: 0, output_tokens: 0 }
-            }
-          }));
-        };
-        for await (const event of gen) {
-          if (event.type === "start") {
-            ensureMessageStart(event.partial.model || "");
-          } else if (event.type === "text_start") {
-            ensureMessageStart(event.partial.model || "");
-            openBlocks.set(event.contentIndex, "text");
-            controller.enqueue(sseFrame2("content_block_start", {
-              type: "content_block_start",
-              index: event.contentIndex,
-              content_block: { type: "text", text: "" }
-            }));
-          } else if (event.type === "text_delta") {
-            if (openBlocks.get(event.contentIndex) !== "text") {
-              ensureMessageStart(event.partial.model || "");
-              openBlocks.set(event.contentIndex, "text");
-              controller.enqueue(sseFrame2("content_block_start", {
-                type: "content_block_start",
-                index: event.contentIndex,
-                content_block: { type: "text", text: "" }
-              }));
-            }
-            controller.enqueue(sseFrame2("content_block_delta", {
-              type: "content_block_delta",
-              index: event.contentIndex,
-              delta: { type: "text_delta", text: event.delta }
-            }));
-          } else if (event.type === "text_end") {
-            if (openBlocks.get(event.contentIndex) === "text") {
-              controller.enqueue(sseFrame2("content_block_stop", {
-                type: "content_block_stop",
-                index: event.contentIndex
-              }));
-              openBlocks.delete(event.contentIndex);
-            }
-          } else if (event.type === "done") {
-            for (const [idx] of openBlocks) {
-              controller.enqueue(sseFrame2("content_block_stop", {
-                type: "content_block_stop",
-                index: idx
-              }));
-            }
-            openBlocks.clear();
-            controller.enqueue(sseFrame2("message_delta", {
-              type: "message_delta",
-              delta: { stop_reason: "end_turn", stop_sequence: null },
-              usage: { output_tokens: event.message.usage?.output ?? 0 }
-            }));
-            controller.enqueue(sseFrame2("message_stop", { type: "message_stop" }));
-          } else if (event.type === "error") {
-            for (const [idx] of openBlocks) {
-              controller.enqueue(sseFrame2("content_block_stop", {
-                type: "content_block_stop",
-                index: idx
-              }));
-            }
-            openBlocks.clear();
-            controller.enqueue(sseFrame2("error", {
-              type: "error",
-              error: {
-                type: "api_error",
-                message: event.error?.errorMessage ?? "Unknown error"
-              }
-            }));
-          }
-        }
-        controller.close();
-      } catch (err) {
-        const msg = err instanceof Error ? err.message : String(err);
-        controller.enqueue(sseFrame2("error", {
-          type: "error",
-          error: { type: "api_error", message: msg }
-        }));
-        controller.close();
+function buildSseForApiType(apiType, gen) {
+  if (apiType === "anthropic-messages")
+    return piStreamToAnthropicSse(gen);
+  if (apiType === "openai-completions")
+    return piStreamToOpenAiCompletionsSse(gen);
+  if (apiType === "openai-responses")
+    return piStreamToOpenAiResponsesSse(gen);
+  if (apiType === "google-generative-ai")
+    return piStreamToGoogleSse(gen);
+  return piStreamToAnthropicSse(gen);
+}
+function buildJsonForApiType(apiType, msg) {
+  if (apiType === "anthropic-messages")
+    return anthropicMessageFromAssistant(msg);
+  if (apiType === "openai-completions")
+    return openAiCompletionsMessageFromAssistant(msg);
+  if (apiType === "openai-responses")
+    return openAiResponsesMessageFromAssistant(msg);
+  if (apiType === "google-generative-ai")
+    return googleMessageFromAssistant(msg);
+  return anthropicMessageFromAssistant(msg);
+}
+async function drainSignalGenerator(gen) {
+  for await (const _event of gen) {}
+}
+async function* replayAssistantMessageAsEvents(msg) {
+  const partial = msg;
+  yield { type: "start", partial };
+  for (let i2 = 0;i2 < msg.content.length; i2++) {
+    const block = msg.content[i2];
+    if (block.type === "text") {
+      yield { type: "text_start", contentIndex: i2, partial };
+      if (block.text.length > 0) {
+        yield { type: "text_delta", contentIndex: i2, delta: block.text, partial };
       }
-    }
-  });
-}
-async function collectPiStreamText(piStreamHandle, detector, state) {
-  const msg = await piStreamHandle.result();
-  const fullText = msg.content.filter((c) => c.type === "text").map((c) => c.text).join("");
-  const results = detector.feedChunk(fullText);
-  for (const r2 of results) {
-    if (r2.type === "signal_detected") {
-      state.signalDetected = true;
-      break;
-    }
-  }
-  detector.flush();
-  const STOP_REASON_MAP2 = {
-    stop: "end_turn",
-    length: "max_tokens",
-    toolUse: "tool_use",
-    error: "end_turn",
-    aborted: "end_turn"
-  };
-  const blocks = [];
-  for (const c of msg.content) {
-    if (c.type === "text") {
-      blocks.push({ type: "text", text: c.text });
-    } else if (c.type === "thinking") {
-      blocks.push({
-        type: "thinking",
-        thinking: c.thinking,
-        ...c.thinkingSignature ? { signature: c.thinkingSignature } : {}
-      });
-    } else if (c.type === "toolCall") {
-      blocks.push({
-        type: "tool_use",
-        id: c.id,
-        name: c.name,
-        input: c.arguments ?? {}
-      });
+      yield { type: "text_end", contentIndex: i2, content: block.text, partial };
+    } else if (block.type === "thinking") {
+      yield { type: "thinking_start", contentIndex: i2, partial };
+      if (block.thinking.length > 0) {
+        yield { type: "thinking_delta", contentIndex: i2, delta: block.thinking, partial };
+      }
+      yield { type: "thinking_end", contentIndex: i2, content: block.thinking, partial };
+    } else if (block.type === "toolCall") {
+      yield { type: "toolcall_start", contentIndex: i2, partial };
+      const argsJson = JSON.stringify(block.arguments ?? {});
+      if (argsJson.length > 0 && argsJson !== "{}") {
+        yield { type: "toolcall_delta", contentIndex: i2, delta: argsJson, partial };
+      }
+      yield { type: "toolcall_end", contentIndex: i2, toolCall: block, partial };
     }
   }
-  return {
-    id: "msg_" + Date.now().toString(36),
-    type: "message",
-    role: "assistant",
-    content: blocks,
-    model: msg.model || "",
-    stop_reason: STOP_REASON_MAP2[msg.stopReason] ?? "end_turn",
-    stop_sequence: null,
-    usage: {
-      input_tokens: msg.usage?.input ?? 0,
-      output_tokens: msg.usage?.output ?? 0
-    }
-  };
+  yield { type: "done", reason: msg.stopReason, message: msg };
+}
+async function collectPiStreamWithSignalDetection(piStreamHandle, detector, state) {
+  const signalGen = detectSignalInStream(piStreamHandle, detector, state, () => {});
+  await drainSignalGenerator(signalGen);
+  return await piStreamHandle.result();
 }
 var ROUTE_MAPPINGS = [
   { apiType: "anthropic-messages", key: "/v1/messages" },
@@ -121235,6 +121166,13 @@ function resolveProviderApi(mediumModel, openclawProviders) {
     return api;
   return PROVIDER_API_FALLBACK;
 }
+function resolveProviderBaseUrl(apiType, port) {
+  const origin = `http://localhost:${port}`;
+  if (apiType === "openai-completions" || apiType === "openai-responses") {
+    return `${origin}/v1`;
+  }
+  return origin;
+}
 async function fileExistsLocal(path4) {
   try {
     await import_promises8.access(path4);
@@ -121513,7 +121451,7 @@ async function update() {
     process.exit(1);
   }
 }
-async function fixAgentModelsJson(homeDir, providerKey, correctBaseUrl) {
+async function fixAgentModelsJson(homeDir, providerKey, servicePort, fallbackApiType) {
   const agentsDir = import_node_path7.join(homeDir, ".openclaw", "agents");
   let agentDirs;
   try {
@@ -121531,13 +121469,15 @@ async function fixAgentModelsJson(homeDir, providerKey, correctBaseUrl) {
       if (!agentProviders?.[providerKey])
         continue;
       const entry = agentProviders[providerKey];
+      const agentApiType = typeof entry["api"] === "string" && entry["api"].length > 0 ? entry["api"] : fallbackApiType;
+      const correctBaseUrl = resolveProviderBaseUrl(agentApiType, servicePort);
       const current = String(entry["baseUrl"] ?? "");
       if (current === correctBaseUrl)
         continue;
       entry["baseUrl"] = correctBaseUrl;
       await import_promises8.writeFile(modelsPath, JSON.stringify(data, null, 2) + `
 `);
-      console.log(`  fixed  agent ${agentId} ${providerKey} baseUrl: ${current} → ${correctBaseUrl}`);
+      console.log(`  fixed  agent ${agentId} ${providerKey} baseUrl: ${current} → ${correctBaseUrl} (api=${agentApiType})`);
     } catch {}
   }
 }
@@ -121596,31 +121536,42 @@ async function init() {
   } catch {
     console.log(`[info] clawmux.json not readable, using default api: ${providerApi}`);
   }
+  const servicePort = process.env.CLAWMUX_PORT ?? "3456";
+  const providerBaseUrl = resolveProviderBaseUrl(providerApi, servicePort);
   if (providers[PROVIDER_KEY]) {
     const existing = providers[PROVIDER_KEY];
-    if (existing["api"] !== providerApi) {
+    const existingApi = existing["api"];
+    const existingBaseUrl = existing["baseUrl"];
+    let dirty = false;
+    if (existingApi !== providerApi) {
       existing["api"] = providerApi;
+      dirty = true;
+    }
+    if (existingBaseUrl !== providerBaseUrl) {
+      existing["baseUrl"] = providerBaseUrl;
+      dirty = true;
+    }
+    if (dirty) {
       await import_promises8.writeFile(openclawConfigPath, JSON.stringify(config, null, 2) + `
 `);
-      console.log(`  updated ${PROVIDER_KEY} provider api → ${providerApi}`);
+      console.log(`  updated ${PROVIDER_KEY} provider (api=${providerApi}, baseUrl=${providerBaseUrl})`);
     } else {
-      console.log(`  skip  ${PROVIDER_KEY} (already exists, api=${providerApi})`);
+      console.log(`  skip  ${PROVIDER_KEY} (already exists, api=${providerApi}, baseUrl=${providerBaseUrl})`);
     }
   } else {
     providers[PROVIDER_KEY] = {
-      baseUrl: "http://localhost:3456",
+      baseUrl: providerBaseUrl,
       api: providerApi,
       models: [{ id: "auto", name: "ClawMux Auto Router" }]
     };
     await import_promises8.writeFile(openclawConfigPath, JSON.stringify(config, null, 2) + `
 `);
-    console.log(`  added ${PROVIDER_KEY} provider to openclaw.json (api=${providerApi})`);
+    console.log(`  added ${PROVIDER_KEY} provider to openclaw.json (api=${providerApi}, baseUrl=${providerBaseUrl})`);
   }
-  await fixAgentModelsJson(homeDir, PROVIDER_KEY, "http://localhost:3456");
-  const port = process.env.CLAWMUX_PORT ?? "3456";
+  await fixAgentModelsJson(homeDir, PROVIDER_KEY, servicePort, providerApi);
   if (!noService) {
     console.log("");
-    await installService(port, process.cwd());
+    await installService(servicePort, process.cwd());
   }
   console.log(`
 [info] ClawMux setup complete!`);

package/dist/index.cjs

@@ -112238,7 +112238,7 @@ __export(exports_src2, {
 module.exports = __toCommonJS(exports_src2);
 
 // src/proxy/router.ts
-var VERSION = process.env.npm_package_version ?? "0.3.13";
+var VERSION = process.env.npm_package_version ?? "0.3.15";
 function jsonResponse(body, status = 200) {
   return new Response(JSON.stringify(body), {
     status,
@@ -115109,6 +115109,9 @@ function djb2Hash(text) {
   return (hash >>> 0).toString(16);
 }
 function extractText(content) {
+  if (content === null || content === undefined) {
+    return "";
+  }
   if (typeof content === "string") {
     return content.slice(0, 200);
   }
@@ -115121,6 +115124,9 @@ function extractText(content) {
   return concatenated.slice(0, 200);
 }
 function contentLength(content) {
+  if (content === null || content === undefined) {
+    return 0;
+  }
   if (typeof content === "string") {
     return content.length;
   }
@@ -115196,7 +115202,7 @@ class EscalationMemory {
 }
 
 // src/routing/instruction-injector.ts
-var INJECT_FOR_TIERS = new Set(["LIGHT"]);
+var INJECT_FOR_TIERS = new Set(["LIGHT", "MEDIUM"]);
 var ESCALATION_INSTRUCTION = `If you cannot handle this request fully (due to complexity, missing context, or capability limits), output EXACTLY the following marker with no other text on that line: ${ESCALATE_SIGNAL}
 Do not explain. Do not ask permission. Just emit the marker and stop.`;
 function injectEscalationInstruction(messages) {
@@ -115212,6 +115218,12 @@ function injectEscalationInstruction(messages) {
       ...rest
     ];
   }
+  if (first.content === null) {
+    return [
+      { role: "system", content: ESCALATION_INSTRUCTION },
+      ...rest
+    ];
+  }
   if (typeof first.content === "string") {
     return [
       { role: "system", content: first.content + `
@@ -119748,8 +119760,7 @@ function piStreamToAnthropicSse(piStream) {
     }
   });
 }
-async function piStreamToAnthropicJson(piStream) {
-  const msg = await piStream.result();
+function anthropicMessageFromAssistant(msg) {
   if ((msg.stopReason === "error" || msg.stopReason === "aborted") && msg.errorMessage) {
     console.error(`[clawmux] pi-ai upstream error (${msg.stopReason}): ${msg.errorMessage}`);
   }
@@ -119787,6 +119798,10 @@ async function piStreamToAnthropicJson(piStream) {
     ...msg.errorMessage ? { error_message: msg.errorMessage } : {}
   };
 }
+async function piStreamToAnthropicJson(piStream) {
+  const msg = await piStream.result();
+  return anthropicMessageFromAssistant(msg);
+}
 
 // src/pi-bridge/event-to-openai-completions.ts
 var encoder3 = new TextEncoder;
@@ -120010,8 +120025,7 @@ function piStreamToOpenAiCompletionsSse(piStream) {
     }
   });
 }
-async function piStreamToOpenAiCompletionsJson(piStream) {
-  const msg = await piStream.result();
+function openAiCompletionsMessageFromAssistant(msg) {
   let textContent = "";
   const toolCalls = [];
   let reasoning = "";
@@ -120061,6 +120075,10 @@ async function piStreamToOpenAiCompletionsJson(piStream) {
     }
   };
 }
+async function piStreamToOpenAiCompletionsJson(piStream) {
+  const msg = await piStream.result();
+  return openAiCompletionsMessageFromAssistant(msg);
+}
 
 // src/pi-bridge/event-to-openai-responses.ts
 var encoder4 = new TextEncoder;
@@ -120223,8 +120241,7 @@ function piStreamToOpenAiResponsesSse(piStream) {
     }
   });
 }
-async function piStreamToOpenAiResponsesJson(piStream) {
-  const msg = await piStream.result();
+function openAiResponsesMessageFromAssistant(msg) {
   const output = [];
   for (const c of msg.content) {
     if (c.type === "text") {
@@ -120265,6 +120282,10 @@ async function piStreamToOpenAiResponsesJson(piStream) {
     }
   };
 }
+async function piStreamToOpenAiResponsesJson(piStream) {
+  const msg = await piStream.result();
+  return openAiResponsesMessageFromAssistant(msg);
+}
 
 // src/pi-bridge/event-to-google.ts
 var encoder5 = new TextEncoder;
@@ -120381,8 +120402,7 @@ function piStreamToGoogleSse(piStream) {
     }
   });
 }
-async function piStreamToGoogleJson(piStream) {
-  const msg = await piStream.result();
+function googleMessageFromAssistant(msg) {
   const parts = [];
   for (const c of msg.content) {
     if (c.type === "text") {
@@ -120414,6 +120434,10 @@ async function piStreamToGoogleJson(piStream) {
     modelVersion: msg.model || ""
   };
 }
+async function piStreamToGoogleJson(piStream) {
+  const msg = await piStream.result();
+  return googleMessageFromAssistant(msg);
+}
 
 // src/proxy/pipeline.ts
 registerAdapter(new AnthropicAdapter);
@@ -120650,35 +120674,12 @@ async function handleApiRequest(req, body, apiType, config, openclawConfig, auth
         }
         const detector = signalRouter.createSignalDetector();
         const detectionState = createSignalDetectionState();
-        if (wantsStream) {
-          const signalGen = detectSignalInStream(piStreamHandle, detector, detectionState, () => {});
-          const sseBody = piStreamToAnthropicSseFromGenerator(signalGen);
-          const response = new Response(sseBody, {
-            status: 200,
-            headers: { "content-type": "text/event-stream" }
-          });
-          response.clone().text().then(() => {
-            if (detectionState.signalDetected) {
-              abortCtrl.abort();
-              const nextTier = signalRouter.handleEscalation(messages, currentTier);
-              if (nextTier !== null) {
-                console.log(`[clawmux] [escalation] ${currentTier} → ${nextTier} (signal detected)`);
-                currentTier = nextTier;
-                return;
-              }
-            }
-            signalRouter.touchActivity(messages);
-            if (attempt > 0) {
-              signalRouter.recordSuccessfulEscalation(messages, currentTier);
-            }
-          });
-          return response;
-        }
-        const fullText = await collectPiStreamText(piStreamHandle, detector, detectionState);
+        const assistantMsg = await collectPiStreamWithSignalDetection(piStreamHandle, detector, detectionState);
         if (detectionState.signalDetected) {
+          abortCtrl.abort();
           const nextTier = signalRouter.handleEscalation(messages, currentTier);
           if (nextTier !== null) {
-            console.log(`[clawmux] [escalation] ${currentTier} → ${nextTier} (signal detected in non-streaming)`);
+            console.log(`[clawmux] [escalation] ${currentTier} → ${nextTier} (signal detected)`);
             currentTier = nextTier;
             continue;
           }
@@ -120687,7 +120688,15 @@ async function handleApiRequest(req, body, apiType, config, openclawConfig, auth
         if (attempt > 0) {
           signalRouter.recordSuccessfulEscalation(messages, currentTier);
         }
-        return new Response(JSON.stringify(fullText), {
+        if (wantsStream) {
+          const sseBody = buildSseForApiType(apiType, replayAssistantMessageAsEvents(assistantMsg));
+          return new Response(sseBody, {
+            status: 200,
+            headers: { "content-type": "text/event-stream" }
+          });
+        }
+        const jsonBody = buildJsonForApiType(apiType, assistantMsg);
+        return new Response(JSON.stringify(jsonBody), {
           status: 200,
           headers: { "content-type": "application/json" }
         });
@@ -120933,164 +120942,63 @@ async function yieldPiAiResponse(piStreamHandle, apiType, wantsStream) {
   }
   throw new Error(`Unsupported pi-ai apiType: ${apiType}`);
 }
-function piStreamToAnthropicSseFromGenerator(gen) {
-  const encoder6 = new TextEncoder;
-  function sseFrame2(event, data) {
-    return encoder6.encode(`event: ${event}
-data: ${JSON.stringify(data)}
-
-`);
-  }
-  return new ReadableStream({
-    async start(controller) {
-      try {
-        let messageStarted = false;
-        const openBlocks = new Map;
-        const ensureMessageStart = (model) => {
-          if (messageStarted)
-            return;
-          messageStarted = true;
-          controller.enqueue(sseFrame2("message_start", {
-            type: "message_start",
-            message: {
-              id: "msg_" + Date.now().toString(36),
-              type: "message",
-              role: "assistant",
-              content: [],
-              model,
-              stop_reason: null,
-              stop_sequence: null,
-              usage: { input_tokens: 0, output_tokens: 0 }
-            }
-          }));
-        };
-        for await (const event of gen) {
-          if (event.type === "start") {
-            ensureMessageStart(event.partial.model || "");
-          } else if (event.type === "text_start") {
-            ensureMessageStart(event.partial.model || "");
-            openBlocks.set(event.contentIndex, "text");
-            controller.enqueue(sseFrame2("content_block_start", {
-              type: "content_block_start",
-              index: event.contentIndex,
-              content_block: { type: "text", text: "" }
-            }));
-          } else if (event.type === "text_delta") {
-            if (openBlocks.get(event.contentIndex) !== "text") {
-              ensureMessageStart(event.partial.model || "");
-              openBlocks.set(event.contentIndex, "text");
-              controller.enqueue(sseFrame2("content_block_start", {
-                type: "content_block_start",
-                index: event.contentIndex,
-                content_block: { type: "text", text: "" }
-              }));
-            }
-            controller.enqueue(sseFrame2("content_block_delta", {
-              type: "content_block_delta",
-              index: event.contentIndex,
-              delta: { type: "text_delta", text: event.delta }
-            }));
-          } else if (event.type === "text_end") {
-            if (openBlocks.get(event.contentIndex) === "text") {
-              controller.enqueue(sseFrame2("content_block_stop", {
-                type: "content_block_stop",
-                index: event.contentIndex
-              }));
-              openBlocks.delete(event.contentIndex);
-            }
-          } else if (event.type === "done") {
-            for (const [idx] of openBlocks) {
-              controller.enqueue(sseFrame2("content_block_stop", {
-                type: "content_block_stop",
-                index: idx
-              }));
-            }
-            openBlocks.clear();
-            controller.enqueue(sseFrame2("message_delta", {
-              type: "message_delta",
-              delta: { stop_reason: "end_turn", stop_sequence: null },
-              usage: { output_tokens: event.message.usage?.output ?? 0 }
-            }));
-            controller.enqueue(sseFrame2("message_stop", { type: "message_stop" }));
-          } else if (event.type === "error") {
-            for (const [idx] of openBlocks) {
-              controller.enqueue(sseFrame2("content_block_stop", {
-                type: "content_block_stop",
-                index: idx
-              }));
-            }
-            openBlocks.clear();
-            controller.enqueue(sseFrame2("error", {
-              type: "error",
-              error: {
-                type: "api_error",
-                message: event.error?.errorMessage ?? "Unknown error"
-              }
-            }));
-          }
-        }
-        controller.close();
-      } catch (err) {
-        const msg = err instanceof Error ? err.message : String(err);
-        controller.enqueue(sseFrame2("error", {
-          type: "error",
-          error: { type: "api_error", message: msg }
-        }));
-        controller.close();
+function buildSseForApiType(apiType, gen) {
+  if (apiType === "anthropic-messages")
+    return piStreamToAnthropicSse(gen);
+  if (apiType === "openai-completions")
+    return piStreamToOpenAiCompletionsSse(gen);
+  if (apiType === "openai-responses")
+    return piStreamToOpenAiResponsesSse(gen);
+  if (apiType === "google-generative-ai")
+    return piStreamToGoogleSse(gen);
+  return piStreamToAnthropicSse(gen);
+}
+function buildJsonForApiType(apiType, msg) {
+  if (apiType === "anthropic-messages")
+    return anthropicMessageFromAssistant(msg);
+  if (apiType === "openai-completions")
+    return openAiCompletionsMessageFromAssistant(msg);
+  if (apiType === "openai-responses")
+    return openAiResponsesMessageFromAssistant(msg);
+  if (apiType === "google-generative-ai")
+    return googleMessageFromAssistant(msg);
+  return anthropicMessageFromAssistant(msg);
+}
+async function drainSignalGenerator(gen) {
+  for await (const _event of gen) {}
+}
+async function* replayAssistantMessageAsEvents(msg) {
+  const partial = msg;
+  yield { type: "start", partial };
+  for (let i2 = 0;i2 < msg.content.length; i2++) {
+    const block = msg.content[i2];
+    if (block.type === "text") {
+      yield { type: "text_start", contentIndex: i2, partial };
+      if (block.text.length > 0) {
+        yield { type: "text_delta", contentIndex: i2, delta: block.text, partial };
       }
-    }
-  });
-}
-async function collectPiStreamText(piStreamHandle, detector, state) {
-  const msg = await piStreamHandle.result();
-  const fullText = msg.content.filter((c) => c.type === "text").map((c) => c.text).join("");
-  const results = detector.feedChunk(fullText);
-  for (const r2 of results) {
-    if (r2.type === "signal_detected") {
-      state.signalDetected = true;
-      break;
-    }
-  }
-  detector.flush();
-  const STOP_REASON_MAP2 = {
-    stop: "end_turn",
-    length: "max_tokens",
-    toolUse: "tool_use",
-    error: "end_turn",
-    aborted: "end_turn"
-  };
-  const blocks = [];
-  for (const c of msg.content) {
-    if (c.type === "text") {
-      blocks.push({ type: "text", text: c.text });
-    } else if (c.type === "thinking") {
-      blocks.push({
-        type: "thinking",
-        thinking: c.thinking,
-        ...c.thinkingSignature ? { signature: c.thinkingSignature } : {}
-      });
-    } else if (c.type === "toolCall") {
-      blocks.push({
-        type: "tool_use",
-        id: c.id,
-        name: c.name,
-        input: c.arguments ?? {}
-      });
+      yield { type: "text_end", contentIndex: i2, content: block.text, partial };
+    } else if (block.type === "thinking") {
+      yield { type: "thinking_start", contentIndex: i2, partial };
+      if (block.thinking.length > 0) {
+        yield { type: "thinking_delta", contentIndex: i2, delta: block.thinking, partial };
+      }
+      yield { type: "thinking_end", contentIndex: i2, content: block.thinking, partial };
+    } else if (block.type === "toolCall") {
+      yield { type: "toolcall_start", contentIndex: i2, partial };
+      const argsJson = JSON.stringify(block.arguments ?? {});
+      if (argsJson.length > 0 && argsJson !== "{}") {
+        yield { type: "toolcall_delta", contentIndex: i2, delta: argsJson, partial };
+      }
+      yield { type: "toolcall_end", contentIndex: i2, toolCall: block, partial };
     }
   }
-  return {
-    id: "msg_" + Date.now().toString(36),
-    type: "message",
-    role: "assistant",
-    content: blocks,
-    model: msg.model || "",
-    stop_reason: STOP_REASON_MAP2[msg.stopReason] ?? "end_turn",
-    stop_sequence: null,
-    usage: {
-      input_tokens: msg.usage?.input ?? 0,
-      output_tokens: msg.usage?.output ?? 0
-    }
-  };
+  yield { type: "done", reason: msg.stopReason, message: msg };
+}
+async function collectPiStreamWithSignalDetection(piStreamHandle, detector, state) {
+  const signalGen = detectSignalInStream(piStreamHandle, detector, state, () => {});
+  await drainSignalGenerator(signalGen);
+  return await piStreamHandle.result();
 }
 var ROUTE_MAPPINGS = [
   { apiType: "anthropic-messages", key: "/v1/messages" },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "clawmux",
-  "version": "0.3.13",
+  "version": "0.3.15",
   "description": "Smart model routing + context compression proxy for OpenClaw",
   "type": "module",
   "bin": {