Merge pull request #688 from theglove44/fix/thinking-blocks-leak

fix(agents): strip <thought> and <antthinking> tags from output

CHANGELOG.md

@@ -13,6 +13,7 @@
 - Docker: allow optional home volume + extra bind mounts in `docker-setup.sh`. (#679) — thanks @gabriel-trigo.
 
 ### Fixes
+- Agents: strip `<thought>`/`<antthinking>` tags from hidden reasoning output and cover tag variants in tests. (#688) — thanks @theglove44.
 - Agents: recognize "usage limit" errors as rate limits for failover. (#687) — thanks @evalexpr.
 - CLI: avoid success message when daemon restart is skipped. (#685) — thanks @carlulsoe.
 - Gateway: disable the OpenAI-compatible `/v1/chat/completions` endpoint by default; enable via `gateway.http.endpoints.chatCompletions.enabled=true`.

src/agents/clawdbot-gateway-tool.test.ts

@@ -14,9 +14,7 @@ describe("gateway tool", () => {
     vi.useFakeTimers();
     const kill = vi.spyOn(process, "kill").mockImplementation(() => true);
     const previousStateDir = process.env.CLAWDBOT_STATE_DIR;
-    const stateDir = await fs.mkdtemp(
+    const stateDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-test-"));
-      path.join(os.tmpdir(), "clawdbot-test-"),
-    );
     process.env.CLAWDBOT_STATE_DIR = stateDir;
 
     try {

src/agents/pi-embedded-subscribe.test.ts

@@ -10,6 +10,12 @@ type StubSession = {
 type SessionEventHandler = (evt: unknown) => void;
 
 describe("subscribeEmbeddedPiSession", () => {
+  const THINKING_TAG_CASES = [
+    { tag: "think", open: "<think>", close: "</think>" },
+    { tag: "thinking", open: "<thinking>", close: "</thinking>" },
+    { tag: "thought", open: "<thought>", close: "</thought>" },
+    { tag: "antthinking", open: "<antthinking>", close: "</antthinking>" },
+  ] as const;
   it("filters to <final> and falls back when tags are malformed", () => {
     let handler: ((evt: unknown) => void) | undefined;
     const session: StubSession = {
@@ -167,7 +173,12 @@ describe("subscribeEmbeddedPiSession", () => {
     expect(onBlockReply.mock.calls[1][0].text).toBe("Final answer");
   });
 
-  it("promotes <think> tags to thinking blocks at write-time", () => {
+  it.each(
+    THINKING_TAG_CASES,
+  )("promotes <%s> tags to thinking blocks at write-time", ({
+    open,
+    close,
+  }) => {
     let handler: ((evt: unknown) => void) | undefined;
     const session: StubSession = {
       subscribe: (fn) => {
@@ -193,7 +204,7 @@ describe("subscribeEmbeddedPiSession", () => {
       content: [
         {
           type: "text",
-          text: "<think>\nBecause it helps\n</think>\n\nFinal answer",
+          text: `${open}\nBecause it helps\n${close}\n\nFinal answer`,
         },
       ],
     } as AssistantMessage;
@@ -212,7 +223,12 @@ describe("subscribeEmbeddedPiSession", () => {
     ]);
   });
 
-  it("streams <think> reasoning via onReasoningStream without leaking into final text", () => {
+  it.each(
+    THINKING_TAG_CASES,
+  )("streams <%s> reasoning via onReasoningStream without leaking into final text", ({
+    open,
+    close,
+  }) => {
     let handler: ((evt: unknown) => void) | undefined;
     const session: StubSession = {
       subscribe: (fn) => {
@@ -240,7 +256,7 @@ describe("subscribeEmbeddedPiSession", () => {
       message: { role: "assistant" },
       assistantMessageEvent: {
         type: "text_delta",
-        delta: "<think>\nBecause",
+        delta: `${open}\nBecause`,
       },
     });
 
@@ -249,7 +265,7 @@ describe("subscribeEmbeddedPiSession", () => {
       message: { role: "assistant" },
       assistantMessageEvent: {
         type: "text_delta",
-        delta: " it helps\n</think>\n\nFinal answer",
+        delta: ` it helps\n${close}\n\nFinal answer`,
       },
     });
 
@@ -258,7 +274,7 @@ describe("subscribeEmbeddedPiSession", () => {
       content: [
         {
           type: "text",
-          text: "<think>\nBecause it helps\n</think>\n\nFinal answer",
+          text: `${open}\nBecause it helps\n${close}\n\nFinal answer`,
         },
       ],
     } as AssistantMessage;
@@ -279,10 +295,9 @@ describe("subscribeEmbeddedPiSession", () => {
     ]);
   });
 
-  it.each([
+  it.each(
-    { tag: "think", open: "<think>", close: "</think>" },
+    THINKING_TAG_CASES,
-    { tag: "thinking", open: "<thinking>", close: "</thinking>" },
+  )("suppresses <%s> blocks across chunk boundaries", ({ open, close }) => {
-  ])("suppresses <%s> blocks across chunk boundaries", ({ open, close }) => {
     let handler: ((evt: unknown) => void) | undefined;
     const session: StubSession = {
       subscribe: (fn) => {

src/agents/pi-embedded-subscribe.ts

@@ -32,10 +32,11 @@ import {
   promoteThinkingTagsToBlocks,
 } from "./pi-embedded-utils.js";
 
-const THINKING_TAG_RE = /<\s*\/?\s*think(?:ing)?\s*>/gi;
+const THINKING_TAG_RE = /<\s*\/?\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
-const THINKING_OPEN_RE = /<\s*think(?:ing)?\s*>/i;
+const THINKING_OPEN_RE = /<\s*(?:think(?:ing)?|thought|antthinking)\s*>/i;
-const THINKING_CLOSE_RE = /<\s*\/\s*think(?:ing)?\s*>/i;
+const THINKING_CLOSE_RE = /<\s*\/\s*(?:think(?:ing)?|thought|antthinking)\s*>/i;
-const THINKING_TAG_SCAN_RE = /<\s*(\/?)\s*think(?:ing)?\s*>/gi;
+const THINKING_TAG_SCAN_RE =
+  /<\s*(\/?)\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
 const TOOL_RESULT_MAX_CHARS = 8000;
 const log = createSubsystemLogger("agent/embedded");
 const RAW_STREAM_ENABLED = process.env.CLAWDBOT_RAW_STREAM === "1";

src/agents/pi-embedded-utils.ts

@@ -52,12 +52,12 @@ export function splitThinkingTaggedText(
   // with a think tag (common for local/OpenAI-compat providers that emulate
   // reasoning blocks via tags).
   if (!trimmedStart.startsWith("<")) return null;
-  const openRe = /<\s*think(?:ing)?\s*>/i;
+  const openRe = /<\s*(?:think(?:ing)?|thought|antthinking)\s*>/i;
-  const closeRe = /<\s*\/\s*think(?:ing)?\s*>/i;
+  const closeRe = /<\s*\/\s*(?:think(?:ing)?|thought|antthinking)\s*>/i;
   if (!openRe.test(trimmedStart)) return null;
   if (!closeRe.test(text)) return null;
 
-  const scanRe = /<\s*(\/?)\s*think(?:ing)?\s*>/gi;
+  const scanRe = /<\s*(\/?)\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
   let inThinking = false;
   let cursor = 0;
   let thinkingStart = 0;
@@ -136,7 +136,7 @@ export function promoteThinkingTagsToBlocks(message: AssistantMessage): void {
 
 export function extractThinkingFromTaggedText(text: string): string {
   if (!text) return "";
-  const scanRe = /<\s*(\/?)\s*think(?:ing)?\s*>/gi;
+  const scanRe = /<\s*(\/?)\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
   let result = "";
   let lastIndex = 0;
   let inThinking = false;
@@ -157,8 +157,8 @@ export function extractThinkingFromTaggedStream(text: string): string {
   const closed = extractThinkingFromTaggedText(text);
   if (closed) return closed;
 
-  const openRe = /<\s*think(?:ing)?\s*>/gi;
+  const openRe = /<\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
-  const closeRe = /<\s*\/\s*think(?:ing)?\s*>/gi;
+  const closeRe = /<\s*\/\s*(?:think(?:ing)?|thought|antthinking)\s*>/gi;
   const openMatches = [...text.matchAll(openRe)];
   if (openMatches.length === 0) return "";
   const closeMatches = [...text.matchAll(closeRe)];

src/cli/update-cli.ts

@@ -165,9 +165,7 @@ export async function updateCommand(opts: UpdateCommandOptions): Promise<void> {
         try {
           await doctorCommand(defaultRuntime, { nonInteractive: true });
         } catch (err) {
-          defaultRuntime.log(
+          defaultRuntime.log(theme.warn(`Doctor failed: ${String(err)}`));
-            theme.warn(`Doctor failed: ${String(err)}`),
-          );
         } finally {
           delete process.env.CLAWDBOT_UPDATE_IN_PROGRESS;
         }