open-agents-ai 0.187.280 → 0.187.282

package/dist/index.js

@@ -328288,6 +328288,23 @@ function computeSignalFromText(text, confidence) {
 function truncateForLog(s2, n2) {
   return s2.length <= n2 ? s2 : s2.slice(0, n2 - 1) + "…";
 }
+function extractToolJson(text) {
+  const lines = text.split(/\r?\n/);
+  for (const line of lines) {
+    const t2 = line.trim();
+    if (!t2.startsWith("{") || !t2.endsWith("}")) continue;
+    try {
+      const obj = JSON.parse(t2);
+      if (typeof obj.tool === "string") {
+        const name10 = obj.tool;
+        const args = obj.args && typeof obj.args === "object" ? obj.args : {};
+        return { name: name10, args };
+      }
+    } catch {
+    }
+  }
+  return null;
+}
 var VAD_SILENCE_MS, MAX_SEGMENT_MS, MAX_CONTEXT_TURNS, SYSTEM_PROMPT2, MIN_SIGNAL_SCORE, NOISE_ONLY_RE, VoiceChatSession;
 var init_voicechat = __esm({
   "packages/cli/src/tui/voicechat.ts"() {
@@ -328295,7 +328312,7 @@ var init_voicechat = __esm({
     VAD_SILENCE_MS = 2e3;
     MAX_SEGMENT_MS = 6500;
     MAX_CONTEXT_TURNS = 20;
-    SYSTEM_PROMPT2 = `You are a voice assistant having a live spoken conversation. Keep responses extremely brief — 1-2 sentences max. You're speaking aloud, not writing. Be conversational, direct, and helpful. Don't use markdown, bullet points, or formatting — just natural speech. If you don't know something, say so briefly. Do not over-think — respond quickly and concisely.`;
+    SYSTEM_PROMPT2 = `You are a voice assistant having a live spoken conversation. Keep responses extremely brief — 1-2 sentences max. You're speaking aloud, not writing. Be conversational, direct, and helpful. Don't use markdown or formatting — just natural speech. Never invent environment facts (like cwd, OS, specs). If you need a precise fact from the main agent, output on a single line EXACTLY one JSON object with fields {"tool": string, "args": object} and nothing else; then wait for the tool result before answering. Prefer tools for factual queries; otherwise, answer directly.`;
     MIN_SIGNAL_SCORE = 0.4;
     NOISE_ONLY_RE = /^(?:[.·…\s,;:!?\-–—_()\[\]{}"'`]+|(?:uh|um|erm|hmm|mm+|uhh+|umm+)[\s.!?]*)+$/i;
     VoiceChatSession = class extends EventEmitter10 {
@@ -328305,6 +328322,9 @@ var init_voicechat = __esm({
       model;
       apiKey;
       runner;
+      verbose = false;
+      debugSnr = false;
+      toolRelay = null;
       // State machine
       _state = "IDLE";
       active = false;
@@ -328337,6 +328357,9 @@ var init_voicechat = __esm({
         this.model = opts.model;
         this.apiKey = opts.apiKey ?? "";
         this.runner = opts.runner ?? null;
+        this.verbose = Boolean(opts.verbose);
+        this.debugSnr = Boolean(opts.debugSnr);
+        this.toolRelay = opts.toolRelay ?? null;
         this.onStatus = opts.onStatus ?? (() => {
         });
         this.onUserSpeech = opts.onUserSpeech ?? (() => {
@@ -328376,7 +328399,7 @@ var init_voicechat = __esm({
         this.active = true;
         this.context = [{ role: "system", content: SYSTEM_PROMPT2 }];
         this.turnCount = 0;
-        this.onStatus("VoiceChat v2 active — state machine: LISTENING");
+        if (this.verbose) this.onStatus("VoiceChat active — LISTENING");
         this._onTranscript = (...args) => {
           let text;
           let isFinal;
@@ -328406,7 +328429,7 @@ var init_voicechat = __esm({
                 await this.listen.stop().catch(() => {
                 });
                 await this.listen.start();
-                this.onStatus("Mic auto-recovered — LISTENING");
+                if (this.verbose) this.onStatus("Mic auto-recovered — LISTENING");
               } catch {
               }
             }, 1e3);
@@ -328417,11 +328440,9 @@ var init_voicechat = __esm({
         try {
           await this.listen.start();
           this.setState("LISTENING");
-          this.onStatus("Mic active — LISTENING for speech...");
+          if (this.verbose) this.onStatus("Mic active — LISTENING for speech...");
         } catch (err) {
-          this.onStatus(
-            `Mic failed: ${err instanceof Error ? err.message : String(err)}. VoiceChat active without mic.`
-          );
+          this.onStatus(`Mic failed: ${err instanceof Error ? err.message : String(err)}. VoiceChat active without mic.`);
           this.setState("LISTENING");
         }
       }
@@ -328456,7 +328477,7 @@ var init_voicechat = __esm({
         } catch {
         }
         this.setState("IDLE");
-        this.onStatus("VoiceChat ended");
+        if (this.verbose) this.onStatus("VoiceChat ended");
         this.emit("stopped");
       }
       // ---------------------------------------------------------------------------
@@ -328512,7 +328533,7 @@ var init_voicechat = __esm({
         }
         const score = this.lastSignalScore ?? computeSignalFromText(text);
         if (score < MIN_SIGNAL_SCORE || NOISE_ONLY_RE.test(text)) {
-          this.onStatus(`Ignoring low-signal utterance (SNR:${score.toFixed(2)}): ${truncateForLog(text, 48)}`);
+          if (this.debugSnr) this.onStatus(`Ignoring low-signal utterance (SNR:${score.toFixed(2)}): ${truncateForLog(text, 48)}`);
           this.emit("snrFiltered", { score, text });
           this.setState("LISTENING");
           this.captureBuffer = "";
@@ -328540,10 +328561,33 @@ var init_voicechat = __esm({
       async think() {
         if (!this.active) return;
         this.setState("THINKING");
-        this.onStatus("Thinking...");
+        if (this.verbose) this.onStatus("Thinking...");
         this.abortController = new AbortController();
         try {
-          const response = await this.streamOllamaInference(this.abortController.signal);
+          if (this.toolRelay?.contextSnapshot) {
+            try {
+              const snap = await Promise.resolve(this.toolRelay.contextSnapshot());
+              if (snap && snap.trim()) {
+                this.context.push({ role: "system", content: `Context snapshot (read-only):
+${snap.trim()}` });
+              }
+            } catch {
+            }
+          }
+          let response = await this.streamOllamaInference(this.abortController.signal);
+          const toolReq = extractToolJson(response);
+          if (toolReq && this.toolRelay) {
+            const { name: name10, args } = toolReq;
+            let toolOutput = "";
+            try {
+              toolOutput = await this.toolRelay.call(name10, args);
+            } catch (e2) {
+              toolOutput = `Tool ${name10} failed: ${e2 instanceof Error ? e2.message : String(e2)}`;
+            }
+            this.context.push({ role: "system", content: `Tool ${name10} result (authoritative):
+${toolOutput}` });
+            response = await this.streamOllamaInference(this.abortController.signal);
+          }
           if (!this.active) return;
           if (response.trim()) {
             this.context.push({ role: "assistant", content: response.trim() });
@@ -328574,7 +328618,7 @@ var init_voicechat = __esm({
         }
         if (this.active) {
           this.setState("LISTENING");
-          this.onStatus("LISTENING...");
+          if (this.verbose) this.onStatus("LISTENING...");
         }
       }
       /**
@@ -333127,36 +333171,87 @@ Respond concisely and safely. Remember: you are talking to the general public.`;
         model: currentConfig.model,
         apiKey: currentConfig.apiKey,
         runner: summaryRunner,
+        toolRelay: {
+          async call(name10, args) {
+            try {
+              if (name10 === "voice_env") {
+                const os8 = __require("node:os");
+                const p2 = __require("node:process");
+                const info = {
+                  cwd: repoRoot,
+                  platform: os8.platform(),
+                  arch: os8.arch(),
+                  cpu: (os8.cpus() || [])[0]?.model || "unknown",
+                  memGB: Math.round(os8.totalmem() / (1024 * 1024 * 1024)),
+                  node: p2.version,
+                  model: currentConfig.model
+                };
+                return JSON.stringify(info, null, 2);
+              }
+              if (name10 === "voice_status") {
+                const status = activeTask ? {
+                  active: true,
+                  toolCalls: activeTask.toolCallCount,
+                  filesTouched: Array.from(activeTask.filesTouched).slice(-20)
+                } : { active: false };
+                return JSON.stringify(status, null, 2);
+              }
+              if (name10 === "voice_list_files") {
+                const baseDir = String(args?.dir ?? ".");
+                const { readdirSync: readdirSync31, statSync: statSync25 } = __require("node:fs");
+                const { join: join106, resolve: resolve40 } = __require("node:path");
+                const base3 = baseDir.startsWith("/") ? baseDir : resolve40(join106(repoRoot, baseDir));
+                const items = readdirSync31(base3).slice(0, 200).map((f2) => {
+                  const s2 = statSync25(join106(base3, f2));
+                  return { name: f2, dir: s2.isDirectory(), size: s2.size };
+                });
+                return JSON.stringify({ dir: base3, items }, null, 2);
+              }
+              if (name10 === "voice_read_file") {
+                const { readFileSync: readFileSync68 } = __require("node:fs");
+                const { join: join106, resolve: resolve40 } = __require("node:path");
+                const rel = String(args?.path || "");
+                const max = Math.max(0, Math.min(8192, Number(args?.max) || 2048));
+                const full = rel.startsWith("/") ? rel : resolve40(join106(repoRoot, rel));
+                const buf = readFileSync68(full);
+                const txt = buf.toString("utf8");
+                return txt.length > max ? txt.slice(0, max) + `
+... [truncated ${txt.length - max} chars]` : txt;
+              }
+              return `Unknown tool: ${name10}`;
+            } catch (e2) {
+              return `Error: ${e2?.message || String(e2)}`;
+            }
+          },
+          contextSnapshot() {
+            const parts = [];
+            parts.push(`cwd: ${repoRoot}`);
+            if (activeTask) {
+              parts.push(`active: yes, toolCalls: ${activeTask.toolCallCount}, filesTouched: ${activeTask.filesTouched.size}`);
+            } else {
+              parts.push("active: no");
+            }
+            return parts.join("\n");
+          }
+        },
+        verbose: false,
+        debugSnr: false,
         onStatus(msg) {
           writeContent(() => renderInfo2(`[voicechat] ${msg}`));
         },
         onUserSpeech(text) {
           writeContent(() => renderInfo2(`\x1B[38;5;45m[you]\x1B[0m ${text}`));
         },
-        onPartialTranscript(text) {
-          writeContent(() => {
-            process.stdout.write(`\r\x1B[2K\x1B[38;5;243m  [hearing] ${text.slice(0, 70)}\x1B[0m`);
-          });
+        // Suppressed to keep main loop quiet
+        onPartialTranscript(_text) {
         },
         onAgentSpeech(text) {
           writeContent(() => renderInfo2(`\x1B[38;5;178m[agent]\x1B[0m ${text.slice(0, 120)}`));
         },
-        onStateChange(state) {
-          writeContent(() => renderInfo2(`\x1B[38;5;243m[voicechat] ${state}\x1B[0m`));
+        // Keep state changes silent
+        onStateChange(_state2) {
         }
       });
-      _voiceChatSession2.on("snr", (e2) => {
-        const s2 = typeof e2?.score === "number" ? Math.max(0, Math.min(1, e2.score)) : null;
-        if (s2 !== null) {
-          writeContent(() => {
-            process.stdout.write(`\r\x1B[2K\x1B[38;5;243m  [hearing] (snr:${s2.toFixed(2)})\x1B[0m`);
-          });
-        }
-      });
-      _voiceChatSession2.on("snrFiltered", (e2) => {
-        const s2 = typeof e2?.score === "number" ? e2.score.toFixed(2) : "?";
-        writeContent(() => renderInfo2(`\x1B[38;5;243m[voicechat]\x1B[0m dropped low-signal utterance (SNR:${s2})`));
-      });
       await _voiceChatSession2.start();
     },
     async voiceChatStop() {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.187.280",
+  "version": "0.187.282",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",