open-agents-ai 0.103.72 → 0.103.74

package/dist/index.js

@@ -12394,6 +12394,7 @@ async function handleCmd(cmd) {
           }
           if (args.temperature !== undefined) riData.temperature = Number(args.temperature);
           if (args.max_tokens) riData.max_tokens = Number(args.max_tokens);
+          if (args.think !== undefined) riData.think = args.think === 'true' || args.think === true;
         } else {
           riData = { prompt: riPrompt };
         }
@@ -12984,6 +12985,7 @@ async function handleCmd(cmd) {
                   }
                   if (parsedReq.temperature !== undefined) chatBody.temperature = parsedReq.temperature;
                   if (parsedReq.max_tokens) chatBody.max_tokens = parsedReq.max_tokens;
+                  if (parsedReq.think !== undefined) chatBody.think = parsedReq.think;
 
                   var chatHeaders = { 'Content-Type': 'application/json' };
                   if (isPassthrough && endpointAuth) chatHeaders['Authorization'] = 'Bearer ' + endpointAuth;
@@ -18637,6 +18639,7 @@ var init_agenticRunner = __esm({
           deepContext: options?.deepContext ?? false,
           dynamicContext: options?.dynamicContext ?? "",
           streamEnabled: options?.streamEnabled ?? false,
+          thinking: options?.thinking ?? true,
           bruteForce: options?.bruteForce ?? true,
           bruteForceMaxCycles: options?.bruteForceMaxCycles ?? 100,
           modelTier: options?.modelTier ?? "large",
@@ -20953,10 +20956,12 @@ ${transcript}`
       baseUrl;
       model;
       apiKey;
-      constructor(baseUrl, model, apiKey) {
+      thinking;
+      constructor(baseUrl, model, apiKey, thinking) {
         this.baseUrl = normalizeBaseUrl(baseUrl);
         this.model = model;
         this.apiKey = apiKey ?? "";
+        this.thinking = thinking ?? true;
       }
       /** Build auth headers — all providers use standard Bearer token auth. */
       authHeaders() {
@@ -20973,7 +20978,7 @@ ${transcript}`
           tools: request.tools,
           temperature: request.temperature,
           max_tokens: request.maxTokens,
-          think: false
+          think: this.thinking
         };
         const resp = await fetch(`${this.baseUrl}/v1/chat/completions`, {
           method: "POST",
@@ -21023,7 +21028,7 @@ ${transcript}`
       }
       /**
        * SSE streaming variant — yields StreamChunks as tokens arrive.
-       * Uses `stream: true` and `think: true` so thinking tokens are visible.
+       * Uses `stream: true` and the current thinking setting.
        * The existing chatCompletion() method is completely unmodified.
        */
       async *chatCompletionStream(request) {
@@ -21035,7 +21040,7 @@ ${transcript}`
           max_tokens: request.maxTokens,
           stream: true,
           stream_options: { include_usage: true },
-          think: true
+          think: this.thinking
         };
         const resp = await fetch(`${this.baseUrl}/v1/chat/completions`, {
           method: "POST",
@@ -21121,11 +21126,13 @@ var init_nexusBackend = __esm({
       authKey;
       consecutiveFailures = 0;
       static MAX_CONSECUTIVE_FAILURES = 3;
-      constructor(sendFn, model, targetPeer, authKey) {
+      thinking;
+      constructor(sendFn, model, targetPeer, authKey, thinking) {
         this.sendFn = sendFn;
         this.model = model;
         this.targetPeer = targetPeer || "";
         this.authKey = authKey || "";
+        this.thinking = thinking ?? true;
       }
       async chatCompletion(request) {
         if (this.consecutiveFailures >= _NexusAgenticBackend.MAX_CONSECUTIVE_FAILURES) {
@@ -21146,6 +21153,7 @@ var init_nexusBackend = __esm({
         if (this.authKey) {
           daemonArgs.auth_key = this.authKey;
         }
+        daemonArgs.think = String(this.thinking);
         let rawResult;
         try {
           rawResult = await this.sendFn("remote_infer", daemonArgs, request.timeoutMs || 12e4);
@@ -26934,6 +26942,8 @@ function renderSlashHelp() {
     ["/endpoint <url>", "Set backend URL (auto-detects type)"],
     ["/endpoint <url> --auth <t>", "Set endpoint with Bearer auth"],
     ["/config", "Show current configuration"],
+    ["/parallel", "Show current Ollama parallel inference slots"],
+    ["/parallel <1-15>", "Set parallel slots (restarts Ollama, max 15)"],
     ["/update", "Check for updates and auto-install"],
     ["/update auto", "Enable auto-update after task completion (default)"],
     ["/update manual", "Disable auto-update (only update on /update)"],
@@ -26942,6 +26952,7 @@ function renderSlashHelp() {
     ["/voice clone <file>", "Set voice clone reference audio (wav/mp3/ogg/flac)"],
     ["/voice clone glados", "Generate clone ref from GLaDOS for LuxTTS"],
     ["/voice clone overwatch", "Generate clone ref from Overwatch for LuxTTS"],
+    ["/think", "Toggle thinking/reasoning mode (Qwen3, DeepSeek-R1, etc.)"],
     ["/stream", "Toggle real-time token streaming (pastel syntax highlighting)"],
     ["/dream", "Start dream mode \u2014 creative idle exploration (writes to .oa/dreams/)"],
     ["/dream deep", "Deep dream \u2014 multi-cycle exploration with sleep architecture"],
@@ -34092,6 +34103,9 @@ async function handleSlashCommand(input, ctx) {
     case "ep":
       await handleEndpoint(arg, ctx, hasLocal);
       return "handled";
+    case "parallel":
+      await handleParallel(arg, ctx);
+      return "handled";
     case "update":
     case "upgrade":
       await handleUpdate(arg, ctx);
@@ -34143,6 +34157,13 @@ async function handleSlashCommand(input, ctx) {
       renderInfo(`Token streaming: ${isOn ? "on" : "off"}${hasLocal ? " (project-local)" : ""}` + (isOn ? " \u2014 thinking tokens in grey italics, responses with pastel syntax highlighting" : ""));
       return "handled";
     }
+    case "think": {
+      const isOn = ctx.thinkToggle();
+      const save = hasLocal ? ctx.saveLocalSettings.bind(ctx) : ctx.saveSettings.bind(ctx);
+      save({ thinking: isOn });
+      renderInfo(`Thinking mode: ${isOn ? "on" : "off"}${hasLocal ? " (project-local)" : ""}` + (isOn ? " \u2014 models that support reasoning (Qwen3, DeepSeek-R1, etc.) will show their thinking chain" : " \u2014 reasoning chain suppressed, model responds directly"));
+      return "handled";
+    }
     case "tools": {
       const tools = listCustomToolFiles(ctx.repoRoot);
       if (tools.length === 0) {
@@ -35515,6 +35536,137 @@ async function handlePeerEndpoint(peerId, authKey, ctx, local) {
     renderWarning("Active task aborted \u2014 endpoint changed. Send a new prompt to continue.");
   }
 }
+async function handleParallel(arg, ctx) {
+  const { execSync: execSync29 } = await import("node:child_process");
+  const baseUrl = ctx.config.backendUrl || "http://localhost:11434";
+  const isRemote = ctx.config.backendType === "nexus";
+  if (isRemote) {
+    renderInfo("Parallel slots are configured on the remote provider's Ollama, not locally.");
+    renderInfo("The remote provider controls OLLAMA_NUM_PARALLEL on their end.");
+    return;
+  }
+  if (!arg) {
+    const envVal = process.env.OLLAMA_NUM_PARALLEL;
+    let runningInfo = "";
+    try {
+      const psResp = await fetch(baseUrl + "/api/ps", { signal: AbortSignal.timeout(3e3) });
+      if (psResp.ok) {
+        const ps = await psResp.json();
+        const models = ps.models || [];
+        if (models.length > 0) {
+          runningInfo = models.map((m) => `    ${c2.cyan(m.name)} \u2014 ${m.size_vram ? (m.size_vram / 1e9).toFixed(1) + " GB VRAM" : "CPU"}`).join("\n");
+        }
+      }
+    } catch {
+    }
+    let systemdVal = "";
+    try {
+      const out = execSync29("systemctl show ollama.service -p Environment 2>/dev/null || true", { encoding: "utf8" });
+      const match = out.match(/OLLAMA_NUM_PARALLEL=(\d+)/);
+      if (match)
+        systemdVal = match[1];
+    } catch {
+    }
+    const effective = envVal || systemdVal || "1 (default)";
+    console.log("");
+    console.log(`  ${c2.bold("Parallel Inference Slots")}`);
+    console.log("");
+    console.log(`  ${c2.dim("OLLAMA_NUM_PARALLEL:")} ${c2.bold(c2.cyan(effective))}`);
+    console.log(`  ${c2.dim("Endpoint:")} ${baseUrl}`);
+    if (runningInfo) {
+      console.log(`  ${c2.dim("Running models:")}`);
+      console.log(runningInfo);
+    }
+    console.log("");
+    console.log(`  ${c2.dim("Set with:")} ${c2.cyan("/parallel <1-15>")}`);
+    console.log(`  ${c2.dim("Each slot multiplies context memory usage per model.")}`);
+    console.log("");
+    return;
+  }
+  const n = parseInt(arg, 10);
+  if (isNaN(n) || n < 1 || n > 15) {
+    renderError("Parallel slots must be between 1 and 15. Usage: /parallel <1-15>");
+    return;
+  }
+  const isSystemd = (() => {
+    try {
+      const out = execSync29("systemctl is-active ollama.service 2>/dev/null", { encoding: "utf8" }).trim();
+      return out === "active" || out === "inactive";
+    } catch {
+      return false;
+    }
+  })();
+  if (isSystemd) {
+    renderInfo(`Setting OLLAMA_NUM_PARALLEL=${n} via systemd override...`);
+    try {
+      const overrideDir = "/etc/systemd/system/ollama.service.d";
+      const overrideFile = `${overrideDir}/parallel.conf`;
+      const overrideContent = `[Service]
+Environment="OLLAMA_NUM_PARALLEL=${n}"
+`;
+      execSync29(`sudo mkdir -p ${overrideDir}`, { stdio: "pipe" });
+      execSync29(`echo '${overrideContent}' | sudo tee ${overrideFile} > /dev/null`, { stdio: "pipe" });
+      execSync29("sudo systemctl daemon-reload", { stdio: "pipe" });
+      execSync29("sudo systemctl restart ollama.service", { stdio: "pipe" });
+      let ready = false;
+      for (let i = 0; i < 30 && !ready; i++) {
+        await new Promise((r) => setTimeout(r, 500));
+        try {
+          const resp = await fetch(baseUrl + "/api/tags", { signal: AbortSignal.timeout(2e3) });
+          if (resp.ok)
+            ready = true;
+        } catch {
+        }
+      }
+      if (ready) {
+        renderInfo(`Ollama restarted with OLLAMA_NUM_PARALLEL=${n}. ${n > 4 ? "Note: memory usage scales with parallel slots." : ""}`);
+      } else {
+        renderWarning(`Ollama service restarted but not responding yet. Check: sudo systemctl status ollama`);
+      }
+    } catch (e) {
+      renderError(`Failed to set systemd override: ${e.message || e}`);
+      renderInfo(`Manual steps:
+  sudo mkdir -p /etc/systemd/system/ollama.service.d
+  echo '[Service]\\nEnvironment="OLLAMA_NUM_PARALLEL=${n}"' | sudo tee /etc/systemd/system/ollama.service.d/parallel.conf
+  sudo systemctl daemon-reload && sudo systemctl restart ollama`);
+    }
+  } else {
+    renderInfo(`Setting OLLAMA_NUM_PARALLEL=${n}...`);
+    try {
+      try {
+        execSync29("pkill -f 'ollama serve' 2>/dev/null || true", { stdio: "pipe" });
+      } catch {
+      }
+      await new Promise((r) => setTimeout(r, 1e3));
+      process.env.OLLAMA_NUM_PARALLEL = String(n);
+      const { spawn: spawn19 } = await import("node:child_process");
+      const child = spawn19("ollama", ["serve"], {
+        stdio: "ignore",
+        detached: true,
+        env: { ...process.env, OLLAMA_NUM_PARALLEL: String(n) }
+      });
+      child.unref();
+      let ready = false;
+      for (let i = 0; i < 30 && !ready; i++) {
+        await new Promise((r) => setTimeout(r, 500));
+        try {
+          const resp = await fetch(baseUrl + "/api/tags", { signal: AbortSignal.timeout(2e3) });
+          if (resp.ok)
+            ready = true;
+        } catch {
+        }
+      }
+      if (ready) {
+        renderInfo(`Ollama restarted with OLLAMA_NUM_PARALLEL=${n}. ${n > 4 ? "Note: memory usage scales with parallel slots." : ""}`);
+      } else {
+        renderWarning(`Ollama restarted but not responding yet. Try: OLLAMA_NUM_PARALLEL=${n} ollama serve`);
+      }
+    } catch (e) {
+      renderError(`Failed to restart Ollama: ${e.message || e}`);
+      renderInfo(`Set manually: OLLAMA_NUM_PARALLEL=${n} ollama serve`);
+    }
+  }
+}
 async function handleUpdate(subcommand, ctx) {
   const repoRoot = ctx.repoRoot;
   if (subcommand === "auto") {
@@ -46501,7 +46653,7 @@ function formatDMNToolArgs(toolName, args) {
       return "";
   }
 }
-function startTask(task, config, repoRoot, voice, stream, taskStores, bruteForce, statusBar, sudoCallback, costTracker, onComplete, taskType, contextWindowSize, modelCaps, personality, deepContext, onCompaction, emotionEngine, flowEnabled, slashCommandHandler) {
+function startTask(task, config, repoRoot, voice, stream, taskStores, bruteForce, statusBar, sudoCallback, costTracker, onComplete, taskType, contextWindowSize, modelCaps, personality, deepContext, onCompaction, emotionEngine, flowEnabled, slashCommandHandler, thinkingEnabled) {
   const voiceStyleMap = {
     concise: 1,
     balanced: 3,
@@ -46549,9 +46701,9 @@ function startTask(task, config, repoRoot, voice, stream, taskStores, bruteForce
       return baseSendFn(cmd, args, timeout);
     };
     const targetPeer = config.backendUrl.startsWith("peer://") ? config.backendUrl.slice(7) : void 0;
-    backend = new NexusAgenticBackend(sendFn, config.model, targetPeer, config.apiKey);
+    backend = new NexusAgenticBackend(sendFn, config.model, targetPeer, config.apiKey, thinkingEnabled);
   } else {
-    backend = new OllamaAgenticBackend(config.backendUrl, config.model, config.apiKey);
+    backend = new OllamaAgenticBackend(config.backendUrl, config.model, config.apiKey, thinkingEnabled);
   }
   try {
     const endpointHistory = loadUsageHistory("endpoint", repoRoot);
@@ -46617,6 +46769,7 @@ function startTask(task, config, repoRoot, voice, stream, taskStores, bruteForce
     dynamicContext,
     modelTier,
     streamEnabled: stream?.enabled ?? false,
+    thinking: thinkingEnabled,
     bruteForce: bruteForce ?? true,
     bruteForceMaxCycles: 100,
     // effectively unlimited — no hard timeout, agent runs until complete or aborted
@@ -46730,6 +46883,8 @@ ${entry.fullContent}`
           "ep",
           "update",
           "upgrade",
+          "parallel",
+          "think",
           "telegram",
           "tg",
           "call",
@@ -47078,6 +47233,7 @@ async function startInteractive(config, repoPath) {
   if (savedSettings.dbPath)
     config = { ...config, dbPath: savedSettings.dbPath };
   let streamEnabled = savedSettings.stream ?? false;
+  let thinkingEnabled = savedSettings.thinking !== false;
   let bruteForceEnabled = savedSettings.bruteforce ?? true;
   let currentStyle = PRESET_NAMES.includes(savedSettings.style) ? savedSettings.style : "balanced";
   let deepContextEnabled = savedSettings.deepContext ?? false;
@@ -47759,6 +47915,10 @@ Rationale: ${proposal.rationale}${provenanceNote}`;
       streamEnabled = !streamEnabled;
       return streamEnabled;
     },
+    thinkToggle() {
+      thinkingEnabled = !thinkingEnabled;
+      return thinkingEnabled;
+    },
     bruteForceToggle() {
       bruteForceEnabled = !bruteForceEnabled;
       return bruteForceEnabled;
@@ -48912,7 +49072,7 @@ Execute this skill now. Follow the behavioral guidance above.`;
           }, bruteForceEnabled, statusBar, handleSudoRequest, costTracker, (summary, meta) => {
             lastCompletedSummary = summary;
             lastTaskMeta = meta ?? null;
-          }, currentTaskType, resolvedContextWindowSize, resolvedCaps, currentStyle, deepContextEnabled, compactionSNRCallback(), emotionEngine, flowEnabled, buildSlashCommandHandler());
+          }, currentTaskType, resolvedContextWindowSize, resolvedCaps, currentStyle, deepContextEnabled, compactionSNRCallback(), emotionEngine, flowEnabled, buildSlashCommandHandler(), thinkingEnabled);
           activeTask = task;
           showPrompt();
           await task.promise;
@@ -49144,7 +49304,7 @@ NEW TASK: ${fullInput}`;
       }, bruteForceEnabled, statusBar, handleSudoRequest, costTracker, (summary, meta) => {
         lastCompletedSummary = summary;
         lastTaskMeta = meta ?? null;
-      }, currentTaskType, resolvedContextWindowSize, resolvedCaps, currentStyle, deepContextEnabled, compactionSNRCallback(), emotionEngine, flowEnabled, buildSlashCommandHandler());
+      }, currentTaskType, resolvedContextWindowSize, resolvedCaps, currentStyle, deepContextEnabled, compactionSNRCallback(), emotionEngine, flowEnabled, buildSlashCommandHandler(), thinkingEnabled);
       activeTask = task;
       showPrompt();
       await task.promise;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.103.72",
+  "version": "0.103.74",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",