@goondocks/myco 0.4.3 → 0.4.4

package/dist/{chunk-OUFSLZTX.js → chunk-KDWBZSOB.js}

@@ -1,15 +1,16 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 import {
   getPluginVersion
-} from "./chunk-2GJFTIWX.js";
+} from "./chunk-TK2ZYIAL.js";
 import {
   AgentRegistry
-} from "./chunk-BNIYWCST.js";
+} from "./chunk-EQVQEFOA.js";
 import {
   DAEMON_CLIENT_TIMEOUT_MS,
   DAEMON_HEALTH_CHECK_TIMEOUT_MS,
-  DAEMON_HEALTH_RETRY_DELAYS
-} from "./chunk-JBD5KP5G.js";
+  DAEMON_HEALTH_RETRY_DELAYS,
+  DAEMON_STALE_GRACE_PERIOD_MS
+} from "./chunk-TDLQBGKA.js";
 
 // src/hooks/client.ts
 import fs from "fs";
@@ -68,9 +69,16 @@ var DaemonClient = class {
   /**
    * Check if the daemon is running a stale version.
    * Returns true if the daemon's version doesn't match the current plugin version.
+   * Skips the check if daemon.json was written recently (grace period) to prevent
+   * rapid restart loops from concurrent hooks or session reloads.
    */
   async isStale() {
     try {
+      const jsonPath = path.join(this.vaultDir, "daemon.json");
+      const stat = fs.statSync(jsonPath);
+      if (Date.now() - stat.mtimeMs < DAEMON_STALE_GRACE_PERIOD_MS) {
+        return false;
+      }
       const info = this.readDaemonJson();
       if (!info) return false;
       const res = await fetch(`http://127.0.0.1:${info.port}/health`, {
@@ -101,11 +109,15 @@ var DaemonClient = class {
     }
   }
   /**
-   * Ensure the daemon is running the current version. Spawns it if unhealthy
-   * or restarts it if the version is stale. Returns true if healthy after this call.
+   * Ensure the daemon is running. Spawns it if unhealthy.
+   * When checkStale is true (default), also restarts a healthy daemon if its
+   * version doesn't match the current plugin version. Use checkStale: false
+   * for hooks that just need the daemon alive (e.g., stop) without triggering
+   * version-driven restarts.
    */
-  async ensureRunning() {
-    if (await this.isStale()) {
+  async ensureRunning(opts) {
+    const checkStale = opts?.checkStale ?? true;
+    if (checkStale && await this.isStale()) {
       this.killDaemon();
       await new Promise((r) => setTimeout(r, 200));
     } else if (await this.isHealthy()) {
@@ -166,4 +178,4 @@ var DaemonClient = class {
 export {
   DaemonClient
 };
-//# sourceMappingURL=chunk-OUFSLZTX.js.map
+//# sourceMappingURL=chunk-KDWBZSOB.js.map

package/dist/chunk-KDWBZSOB.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/hooks/client.ts"],"sourcesContent":["import fs from 'node:fs';\nimport path from 'node:path';\nimport { spawn } from 'node:child_process';\nimport { DAEMON_CLIENT_TIMEOUT_MS, DAEMON_HEALTH_CHECK_TIMEOUT_MS, DAEMON_HEALTH_RETRY_DELAYS, DAEMON_STALE_GRACE_PERIOD_MS } from '../constants.js';\nimport { AgentRegistry } from '../agents/registry.js';\nimport { getPluginVersion } from '../version.js';\n\ninterface DaemonInfo {\n  pid: number;\n  port: number;\n}\n\ninterface HealthResponse {\n  myco: boolean;\n  version?: string;\n}\n\ninterface ClientResult {\n  ok: boolean;\n  data?: any;\n}\n\nexport class DaemonClient {\n  private vaultDir: string;\n\n  constructor(vaultDir: string) {\n    this.vaultDir = vaultDir;\n  }\n\n  async post(endpoint: string, body: unknown): Promise<ClientResult> {\n    try {\n      const info = this.readDaemonJson();\n      if (!info) return { ok: false };\n\n      const res = await fetch(`http://127.0.0.1:${info.port}${endpoint}`, {\n        method: 'POST',\n        headers: { 'Content-Type': 'application/json' },\n        body: JSON.stringify(body),\n        signal: AbortSignal.timeout(DAEMON_CLIENT_TIMEOUT_MS),\n      });\n\n      if (!res.ok) return { ok: false };\n      const data = await res.json();\n      return { ok: true, data };\n    } catch {\n      return { ok: false };\n    }\n  }\n\n  async get(endpoint: string): Promise<ClientResult> {\n    try {\n      const info = this.readDaemonJson();\n      if (!info) return { ok: false };\n\n      const res = await fetch(`http://127.0.0.1:${info.port}${endpoint}`, {\n        signal: AbortSignal.timeout(DAEMON_CLIENT_TIMEOUT_MS),\n      });\n\n      if (!res.ok) return { ok: false };\n      const data = await res.json();\n      return { ok: true, data };\n    } catch {\n      return { ok: false };\n    }\n  }\n\n  async isHealthy(): Promise<boolean> {\n    try {\n      const info = this.readDaemonJson();\n      if (!info) return false;\n\n      const res = await fetch(`http://127.0.0.1:${info.port}/health`, {\n        signal: AbortSignal.timeout(DAEMON_HEALTH_CHECK_TIMEOUT_MS),\n      });\n      if (!res.ok) return false;\n      const data = await res.json() as HealthResponse;\n      return data.myco === true;\n    } catch {\n      return false;\n    }\n  }\n\n  /**\n   * Check if the daemon is running a stale version.\n   * Returns true if the daemon's version doesn't match the current plugin version.\n   * Skips the check if daemon.json was written recently (grace period) to prevent\n   * rapid restart loops from concurrent hooks or session reloads.\n   */\n  private async isStale(): Promise<boolean> {\n    try {\n      const jsonPath = path.join(this.vaultDir, 'daemon.json');\n      const stat = fs.statSync(jsonPath);\n      if (Date.now() - stat.mtimeMs < DAEMON_STALE_GRACE_PERIOD_MS) {\n        return false;\n      }\n\n      const info = this.readDaemonJson();\n      if (!info) return false;\n\n      const res = await fetch(`http://127.0.0.1:${info.port}/health`, {\n        signal: AbortSignal.timeout(DAEMON_HEALTH_CHECK_TIMEOUT_MS),\n      });\n      if (!res.ok) return false;\n      const data = await res.json() as HealthResponse;\n      if (!data.myco) return false;\n\n      // No version in response = old daemon that predates this check\n      if (!data.version) return true;\n\n      return data.version !== getPluginVersion();\n    } catch {\n      return false;\n    }\n  }\n\n  /**\n   * Kill the running daemon process.\n   */\n  private killDaemon(): void {\n    try {\n      const info = this.readDaemonJson();\n      if (!info) return;\n      process.kill(info.pid, 'SIGTERM');\n    } catch { /* already dead */ }\n    try {\n      fs.unlinkSync(path.join(this.vaultDir, 'daemon.json'));\n    } catch { /* already gone */ }\n  }\n\n  /**\n   * Ensure the daemon is running. Spawns it if unhealthy.\n   * When checkStale is true (default), also restarts a healthy daemon if its\n   * version doesn't match the current plugin version. Use checkStale: false\n   * for hooks that just need the daemon alive (e.g., stop) without triggering\n   * version-driven restarts.\n   */\n  async ensureRunning(opts?: { checkStale?: boolean }): Promise<boolean> {\n    const checkStale = opts?.checkStale ?? true;\n\n    if (checkStale && await this.isStale()) {\n      this.killDaemon();\n      // Brief pause for port release\n      await new Promise((r) => setTimeout(r, 200));\n    } else if (await this.isHealthy()) {\n      return true;\n    }\n\n    this.spawnDaemon();\n\n    for (const delay of DAEMON_HEALTH_RETRY_DELAYS) {\n      await new Promise((r) => setTimeout(r, delay));\n      if (await this.isHealthy()) return true;\n    }\n    return false;\n  }\n\n  spawnDaemon(): void {\n    const daemonScript = this.resolveDaemonScript();\n    if (!daemonScript || !fs.existsSync(daemonScript)) return;\n\n    const child = spawn('node', [daemonScript, '--vault', this.vaultDir], {\n      detached: true,\n      stdio: 'ignore',\n    });\n    child.unref();\n  }\n\n  /**\n   * Resolve the daemon entry script path.\n   * Priority:\n   * 1. Plugin root env var (set by the agent host) → dist/src/daemon/main.js\n   * 2. Walk up from the current file to find the dist/ directory containing\n   *    the daemon entry. This handles both chunk files (dist/chunk-*.js) and\n   *    thin entry points (dist/src/hooks/*.js) after bundling.\n   */\n  private resolveDaemonScript(): string | undefined {\n    const pluginRoot = new AgentRegistry().resolvePluginRoot();\n    if (pluginRoot) {\n      return path.join(pluginRoot, 'dist', 'src', 'daemon', 'main.js');\n    }\n\n    // Walk up from import.meta.dirname looking for the daemon entry\n    let dir = import.meta.dirname;\n    for (let i = 0; i < 5; i++) {\n      const candidate = path.join(dir, 'dist', 'src', 'daemon', 'main.js');\n      if (fs.existsSync(candidate)) return candidate;\n      // Also check if we're already inside dist/\n      const inDist = path.join(dir, 'src', 'daemon', 'main.js');\n      if (fs.existsSync(inDist)) return inDist;\n      dir = path.dirname(dir);\n    }\n    return undefined;\n  }\n\n  private readDaemonJson(): DaemonInfo | null {\n    try {\n      const jsonPath = path.join(this.vaultDir, 'daemon.json');\n      const content = fs.readFileSync(jsonPath, 'utf-8');\n      const info = JSON.parse(content);\n      if (typeof info.port !== 'number') return null;\n      return info as DaemonInfo;\n    } catch {\n      return null;\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;AAAA,OAAO,QAAQ;AACf,OAAO,UAAU;AACjB,SAAS,aAAa;AAoBf,IAAM,eAAN,MAAmB;AAAA,EAChB;AAAA,EAER,YAAY,UAAkB;AAC5B,SAAK,WAAW;AAAA,EAClB;AAAA,EAEA,MAAM,KAAK,UAAkB,MAAsC;AACjE,QAAI;AACF,YAAM,OAAO,KAAK,eAAe;AACjC,UAAI,CAAC,KAAM,QAAO,EAAE,IAAI,MAAM;AAE9B,YAAM,MAAM,MAAM,MAAM,oBAAoB,KAAK,IAAI,GAAG,QAAQ,IAAI;AAAA,QAClE,QAAQ;AAAA,QACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,QAC9C,MAAM,KAAK,UAAU,IAAI;AAAA,QACzB,QAAQ,YAAY,QAAQ,wBAAwB;AAAA,MACtD,CAAC;AAED,UAAI,CAAC,IAAI,GAAI,QAAO,EAAE,IAAI,MAAM;AAChC,YAAM,OAAO,MAAM,IAAI,KAAK;AAC5B,aAAO,EAAE,IAAI,MAAM,KAAK;AAAA,IAC1B,QAAQ;AACN,aAAO,EAAE,IAAI,MAAM;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAM,IAAI,UAAyC;AACjD,QAAI;AACF,YAAM,OAAO,KAAK,eAAe;AACjC,UAAI,CAAC,KAAM,QAAO,EAAE,IAAI,MAAM;AAE9B,YAAM,MAAM,MAAM,MAAM,oBAAoB,KAAK,IAAI,GAAG,QAAQ,IAAI;AAAA,QAClE,QAAQ,YAAY,QAAQ,wBAAwB;AAAA,MACtD,CAAC;AAED,UAAI,CAAC,IAAI,GAAI,QAAO,EAAE,IAAI,MAAM;AAChC,YAAM,OAAO,MAAM,IAAI,KAAK;AAC5B,aAAO,EAAE,IAAI,MAAM,KAAK;AAAA,IAC1B,QAAQ;AACN,aAAO,EAAE,IAAI,MAAM;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAM,YAA8B;AAClC,QAAI;AACF,YAAM,OAAO,KAAK,eAAe;AACjC,UAAI,CAAC,KAAM,QAAO;AAElB,YAAM,MAAM,MAAM,MAAM,oBAAoB,KAAK,IAAI,WAAW;AAAA,QAC9D,QAAQ,YAAY,QAAQ,8BAA8B;AAAA,MAC5D,CAAC;AACD,UAAI,CAAC,IAAI,GAAI,QAAO;AACpB,YAAM,OAAO,MAAM,IAAI,KAAK;AAC5B,aAAO,KAAK,SAAS;AAAA,IACvB,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQA,MAAc,UAA4B;AACxC,QAAI;AACF,YAAM,WAAW,KAAK,KAAK,KAAK,UAAU,aAAa;AACvD,YAAM,OAAO,GAAG,SAAS,QAAQ;AACjC,UAAI,KAAK,IAAI,IAAI,KAAK,UAAU,8BAA8B;AAC5D,eAAO;AAAA,MACT;AAEA,YAAM,OAAO,KAAK,eAAe;AACjC,UAAI,CAAC,KAAM,QAAO;AAElB,YAAM,MAAM,MAAM,MAAM,oBAAoB,KAAK,IAAI,WAAW;AAAA,QAC9D,QAAQ,YAAY,QAAQ,8BAA8B;AAAA,MAC5D,CAAC;AACD,UAAI,CAAC,IAAI,GAAI,QAAO;AACpB,YAAM,OAAO,MAAM,IAAI,KAAK;AAC5B,UAAI,CAAC,KAAK,KAAM,QAAO;AAGvB,UAAI,CAAC,KAAK,QAAS,QAAO;AAE1B,aAAO,KAAK,YAAY,iBAAiB;AAAA,IAC3C,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKQ,aAAmB;AACzB,QAAI;AACF,YAAM,OAAO,KAAK,eAAe;AACjC,UAAI,CAAC,KAAM;AACX,cAAQ,KAAK,KAAK,KAAK,SAAS;AAAA,IAClC,QAAQ;AAAA,IAAqB;AAC7B,QAAI;AACF,SAAG,WAAW,KAAK,KAAK,KAAK,UAAU,aAAa,CAAC;AAAA,IACvD,QAAQ;AAAA,IAAqB;AAAA,EAC/B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,cAAc,MAAmD;AACrE,UAAM,aAAa,MAAM,cAAc;AAEvC,QAAI,cAAc,MAAM,KAAK,QAAQ,GAAG;AACtC,WAAK,WAAW;AAEhB,YAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,GAAG,CAAC;AAAA,IAC7C,WAAW,MAAM,KAAK,UAAU,GAAG;AACjC,aAAO;AAAA,IACT;AAEA,SAAK,YAAY;AAEjB,eAAW,SAAS,4BAA4B;AAC9C,YAAM,IAAI,QAAQ,CAAC,MAAM,WAAW,GAAG,KAAK,CAAC;AAC7C,UAAI,MAAM,KAAK,UAAU,EAAG,QAAO;AAAA,IACrC;AACA,WAAO;AAAA,EACT;AAAA,EAEA,cAAoB;AAClB,UAAM,eAAe,KAAK,oBAAoB;AAC9C,QAAI,CAAC,gBAAgB,CAAC,GAAG,WAAW,YAAY,EAAG;AAEnD,UAAM,QAAQ,MAAM,QAAQ,CAAC,cAAc,WAAW,KAAK,QAAQ,GAAG;AAAA,MACpE,UAAU;AAAA,MACV,OAAO;AAAA,IACT,CAAC;AACD,UAAM,MAAM;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUQ,sBAA0C;AAChD,UAAM,aAAa,IAAI,cAAc,EAAE,kBAAkB;AACzD,QAAI,YAAY;AACd,aAAO,KAAK,KAAK,YAAY,QAAQ,OAAO,UAAU,SAAS;AAAA,IACjE;AAGA,QAAI,MAAM,YAAY;AACtB,aAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,YAAM,YAAY,KAAK,KAAK,KAAK,QAAQ,OAAO,UAAU,SAAS;AACnE,UAAI,GAAG,WAAW,SAAS,EAAG,QAAO;AAErC,YAAM,SAAS,KAAK,KAAK,KAAK,OAAO,UAAU,SAAS;AACxD,UAAI,GAAG,WAAW,MAAM,EAAG,QAAO;AAClC,YAAM,KAAK,QAAQ,GAAG;AAAA,IACxB;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,iBAAoC;AAC1C,QAAI;AACF,YAAM,WAAW,KAAK,KAAK,KAAK,UAAU,aAAa;AACvD,YAAM,UAAU,GAAG,aAAa,UAAU,OAAO;AACjD,YAAM,OAAO,KAAK,MAAM,OAAO;AAC/B,UAAI,OAAO,KAAK,SAAS,SAAU,QAAO;AAC1C,aAAO;AAAA,IACT,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}

package/dist/{chunk-67R6EMYD.js → chunk-OPO47BVS.js}

@@ -2,9 +2,8 @@ import { createRequire as __cr } from 'node:module'; const require = __cr(import
 import {
   DAEMON_CLIENT_TIMEOUT_MS,
   EMBEDDING_REQUEST_TIMEOUT_MS,
-  LLM_REQUEST_TIMEOUT_MS,
-  estimateTokens
-} from "./chunk-JBD5KP5G.js";
+  LLM_REQUEST_TIMEOUT_MS
+} from "./chunk-TDLQBGKA.js";
 
 // src/intelligence/ollama.ts
 var ENDPOINT_GENERATE = "/api/generate";
@@ -15,27 +14,23 @@ var OllamaBackend = class _OllamaBackend {
   name = "ollama";
   baseUrl;
   model;
-  contextWindow;
   defaultMaxTokens;
   constructor(config) {
     this.baseUrl = config?.base_url ?? _OllamaBackend.DEFAULT_BASE_URL;
     this.model = config?.model ?? config?.summary_model ?? "llama3.2";
-    this.contextWindow = config?.context_window ?? 8192;
     this.defaultMaxTokens = config?.max_tokens ?? 1024;
   }
   async summarize(prompt, opts) {
     const maxTokens = opts?.maxTokens ?? this.defaultMaxTokens;
-    const contextLength = opts?.contextLength ?? this.contextWindow;
-    const promptTokens = estimateTokens(prompt);
-    const numCtx = Math.max(promptTokens + maxTokens, contextLength);
+    const options = { num_predict: maxTokens };
+    if (opts?.contextLength) {
+      options.num_ctx = opts.contextLength;
+    }
     const body = {
       model: this.model,
       prompt,
       stream: false,
-      options: {
-        num_ctx: numCtx,
-        num_predict: maxTokens
-      }
+      options
     };
     if (opts?.systemPrompt) {
       body.system = opts.systemPrompt;
@@ -103,7 +98,6 @@ var OllamaBackend = class _OllamaBackend {
 // src/intelligence/lm-studio.ts
 var ENDPOINT_CHAT = "/api/v1/chat";
 var ENDPOINT_MODELS_LOAD = "/api/v1/models/load";
-var ENDPOINT_MODELS_UNLOAD = "/api/v1/models/unload";
 var ENDPOINT_MODELS_LIST = "/v1/models";
 var ENDPOINT_MODELS_NATIVE = "/api/v1/models";
 var ENDPOINT_EMBEDDINGS = "/v1/embeddings";
@@ -112,7 +106,7 @@ var LmStudioBackend = class _LmStudioBackend {
   name = "lm-studio";
   baseUrl;
   model;
-  loadedInstanceId = null;
+  instanceId = null;
   contextWindow;
   defaultMaxTokens;
   constructor(config) {
@@ -123,21 +117,22 @@ var LmStudioBackend = class _LmStudioBackend {
   }
   /**
    * Generate text using LM Studio's native REST API (/api/v1/chat).
-   * Supports per-request context_length, reasoning control, and system_prompt.
+   * Routes to our specific instance by ID when available, with model name +
+   * context_length as fallback. This ensures correct routing when multiple
+   * daemons share the same LM Studio, and graceful degradation when our
+   * instance is evicted by idle TTL.
    */
   async summarize(prompt, opts) {
     const maxTokens = opts?.maxTokens ?? this.defaultMaxTokens;
+    const contextLength = opts?.contextLength ?? this.contextWindow;
     const body = {
-      model: this.loadedInstanceId ?? this.model,
+      model: this.instanceId ?? this.model,
       input: prompt,
       max_output_tokens: maxTokens,
       store: false
     };
-    if (!this.loadedInstanceId) {
-      const contextLength = opts?.contextLength ?? this.contextWindow;
-      if (contextLength) {
-        body.context_length = contextLength;
-      }
+    if (contextLength) {
+      body.context_length = contextLength;
     }
     if (opts?.systemPrompt) {
       body.system_prompt = opts.systemPrompt;
@@ -153,6 +148,9 @@ var LmStudioBackend = class _LmStudioBackend {
     });
     if (!response.ok) {
       const errorBody = await response.text().catch(() => "");
+      if (response.status === 404 && this.instanceId) {
+        this.instanceId = null;
+      }
       throw new Error(`LM Studio summarize failed: ${response.status} ${errorBody.slice(0, 500)}`);
     }
     const data = await response.json();
@@ -183,9 +181,13 @@ var LmStudioBackend = class _LmStudioBackend {
   }
   /**
    * Ensure a model instance is loaded with the desired settings.
-   * First checks for an existing compatible instance to reuse (prevents
-   * accumulation across daemon restarts), then loads a new one only if needed.
-   * Unloads incompatible instances of the same model to prevent resource exhaustion.
+   * Called every digest cycle (not cached) so it recovers from idle TTL eviction.
+   *
+   * The load API is necessary to control offload_kv_cache_to_gpu — a load-time
+   * setting that cannot be set per-request via the chat API.
+   *
+   * Multi-daemon safe: finds or loads our own compatible instance without
+   * touching instances from other daemons/projects. Routes by instance ID.
    */
   async ensureLoaded(contextLength, gpuKvCache) {
     const ctx = contextLength ?? this.contextWindow;
@@ -195,12 +197,10 @@ var LmStudioBackend = class _LmStudioBackend {
       const matchesContext = !ctx || instance.config.context_length === ctx;
       const matchesKvCache = instance.config.offload_kv_cache_to_gpu === kvCache;
       if (matchesContext && matchesKvCache) {
-        this.loadedInstanceId = instance.id;
-        await this.unloadIncompatibleInstances(instances, ctx, kvCache);
+        this.instanceId = instance.id;
         return;
       }
     }
-    await this.unloadIncompatibleInstances(instances, ctx, kvCache);
     const body = {
       model: this.model,
       flash_attention: true,
@@ -220,9 +220,9 @@ var LmStudioBackend = class _LmStudioBackend {
       throw new Error(`LM Studio model load failed: ${response.status} ${errorBody.slice(0, 200)}`);
     }
     const loadResult = await response.json();
-    const instanceId = loadResult.id ?? loadResult.instance_id ?? loadResult.model_instance_id;
-    if (instanceId) {
-      this.loadedInstanceId = instanceId;
+    const id = loadResult.instance_id ?? loadResult.id ?? loadResult.model_instance_id;
+    if (id) {
+      this.instanceId = id;
     }
   }
   /**
@@ -242,27 +242,6 @@ var LmStudioBackend = class _LmStudioBackend {
       return [];
     }
   }
-  /**
-   * Unload instances of this model that don't match the desired settings.
-   * Best-effort — failures are silently ignored to avoid blocking the load path.
-   */
-  async unloadIncompatibleInstances(instances, contextLength, gpuKvCache) {
-    for (const instance of instances) {
-      const matchesContext = !contextLength || instance.config.context_length === contextLength;
-      const matchesKvCache = instance.config.offload_kv_cache_to_gpu === gpuKvCache;
-      if (!matchesContext || !matchesKvCache) {
-        try {
-          await fetch(`${this.baseUrl}${ENDPOINT_MODELS_UNLOAD}`, {
-            method: "POST",
-            headers: { "Content-Type": "application/json" },
-            body: JSON.stringify({ model: instance.id }),
-            signal: AbortSignal.timeout(DAEMON_CLIENT_TIMEOUT_MS)
-          });
-        } catch {
-        }
-      }
-    }
-  }
   async isAvailable() {
     try {
       const response = await fetch(`${this.baseUrl}${ENDPOINT_MODELS_LIST}`, {
@@ -291,4 +270,4 @@ export {
   OllamaBackend,
   LmStudioBackend
 };
-//# sourceMappingURL=chunk-67R6EMYD.js.map
+//# sourceMappingURL=chunk-OPO47BVS.js.map

package/dist/chunk-OPO47BVS.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/intelligence/ollama.ts","../src/intelligence/lm-studio.ts"],"sourcesContent":["import type { LlmProvider, EmbeddingProvider, LlmResponse, EmbeddingResponse, LlmRequestOptions } from './llm.js';\nimport { LLM_REQUEST_TIMEOUT_MS, EMBEDDING_REQUEST_TIMEOUT_MS, DAEMON_CLIENT_TIMEOUT_MS } from '../constants.js';\n\ninterface OllamaConfig {\n  model?: string;\n  base_url?: string;\n  context_window?: number;\n  max_tokens?: number;\n  // Legacy fields (ignored, kept for backward compat during migration)\n  embedding_model?: string;\n  summary_model?: string;\n}\n\n// Ollama API endpoints\nconst ENDPOINT_GENERATE = '/api/generate';\nconst ENDPOINT_EMBED = '/api/embed';\nconst ENDPOINT_TAGS = '/api/tags';\n\nexport class OllamaBackend implements LlmProvider, EmbeddingProvider {\n  static readonly DEFAULT_BASE_URL = 'http://localhost:11434';\n  readonly name = 'ollama';\n  private baseUrl: string;\n  private model: string;\n  private defaultMaxTokens: number;\n\n  constructor(config?: OllamaConfig) {\n    this.baseUrl = config?.base_url ?? OllamaBackend.DEFAULT_BASE_URL;\n    this.model = config?.model ?? config?.summary_model ?? 'llama3.2';\n    this.defaultMaxTokens = config?.max_tokens ?? 1024;\n  }\n\n  async summarize(prompt: string, opts?: LlmRequestOptions): Promise<LlmResponse> {\n    const maxTokens = opts?.maxTokens ?? this.defaultMaxTokens;\n\n    // Only send num_ctx when explicitly requested (e.g., digest needing 65K).\n    // Ollama reloads the model on ANY num_ctx change, so omitting it lets\n    // requests use whatever context is already loaded — no reload thrashing.\n    const options: Record<string, unknown> = { num_predict: maxTokens };\n    if (opts?.contextLength) {\n      options.num_ctx = opts.contextLength;\n    }\n\n    const body: Record<string, unknown> = {\n      model: this.model,\n      prompt,\n      stream: false,\n      options,\n    };\n\n    // System prompt — sent as a separate field instead of concatenated into prompt\n    if (opts?.systemPrompt) {\n      body.system = opts.systemPrompt;\n    }\n\n    // Thinking control — false suppresses chain-of-thought for reasoning models\n    if (opts?.reasoning) {\n      body.think = opts.reasoning === 'off' ? false : opts.reasoning;\n    }\n\n    // Keep model loaded between requests (useful for digest cycles)\n    if (opts?.keepAlive) {\n      body.keep_alive = opts.keepAlive;\n    }\n\n    const response = await fetch(`${this.baseUrl}${ENDPOINT_GENERATE}`, {\n      method: 'POST',\n      headers: { 'Content-Type': 'application/json' },\n      body: JSON.stringify(body),\n      signal: AbortSignal.timeout(opts?.timeoutMs ?? LLM_REQUEST_TIMEOUT_MS),\n    });\n\n    if (!response.ok) {\n      const errorBody = await response.text().catch(() => '');\n      throw new Error(`Ollama summarize failed: ${response.status} ${errorBody.slice(0, 500)}`);\n    }\n\n    const data = await response.json() as { response: string; model: string };\n    return { text: data.response, model: data.model };\n  }\n\n  async embed(text: string): Promise<EmbeddingResponse> {\n    const response = await fetch(`${this.baseUrl}${ENDPOINT_EMBED}`, {\n      method: 'POST',\n      headers: { 'Content-Type': 'application/json' },\n      body: JSON.stringify({\n        model: this.model,\n        input: text,\n      }),\n      signal: AbortSignal.timeout(EMBEDDING_REQUEST_TIMEOUT_MS),\n    });\n\n    if (!response.ok) {\n      throw new Error(`Ollama embed failed: ${response.status} ${response.statusText}`);\n    }\n\n    const data = await response.json() as { embeddings: number[][]; model: string };\n    const embedding = data.embeddings[0];\n    return { embedding, model: data.model, dimensions: embedding.length };\n  }\n\n  async isAvailable(): Promise<boolean> {\n    try {\n      const response = await fetch(`${this.baseUrl}${ENDPOINT_TAGS}`, {\n        signal: AbortSignal.timeout(DAEMON_CLIENT_TIMEOUT_MS),\n      });\n      return response.ok;\n    } catch {\n      return false;\n    }\n  }\n\n  /** List available models on this Ollama instance. */\n  async listModels(timeoutMs?: number): Promise<string[]> {\n    try {\n      const response = await fetch(`${this.baseUrl}${ENDPOINT_TAGS}`, {\n        signal: AbortSignal.timeout(timeoutMs ?? DAEMON_CLIENT_TIMEOUT_MS),\n      });\n      const data = await response.json() as { models: Array<{ name: string }> };\n      return data.models.map((m) => m.name);\n    } catch {\n      return [];\n    }\n  }\n}\n","import type { LlmProvider, EmbeddingProvider, LlmResponse, EmbeddingResponse, LlmRequestOptions } from './llm.js';\nimport { LLM_REQUEST_TIMEOUT_MS, EMBEDDING_REQUEST_TIMEOUT_MS, DAEMON_CLIENT_TIMEOUT_MS } from '../constants.js';\n\ninterface LmStudioConfig {\n  model?: string;\n  base_url?: string;\n  context_window?: number;\n  max_tokens?: number;\n  // Legacy fields\n  embedding_model?: string;\n  summary_model?: string;\n}\n\n// LM Studio API endpoints\nconst ENDPOINT_CHAT = '/api/v1/chat';\nconst ENDPOINT_MODELS_LOAD = '/api/v1/models/load';\nconst ENDPOINT_MODELS_LIST = '/v1/models';\nconst ENDPOINT_MODELS_NATIVE = '/api/v1/models';\nconst ENDPOINT_EMBEDDINGS = '/v1/embeddings';\n\n/** Shape of a loaded instance from the LM Studio native models API. */\ninterface NativeLoadedInstance {\n  id: string;\n  config: {\n    context_length: number;\n    flash_attention: boolean;\n    offload_kv_cache_to_gpu: boolean;\n  };\n}\n\n/** Shape of a model entry from the LM Studio native models API. */\ninterface NativeModelEntry {\n  type: string;\n  key: string;\n  loaded_instances: NativeLoadedInstance[];\n}\n\nexport class LmStudioBackend implements LlmProvider, EmbeddingProvider {\n  static readonly DEFAULT_BASE_URL = 'http://localhost:1234';\n  readonly name = 'lm-studio';\n  private baseUrl: string;\n  private model: string;\n  private instanceId: string | null = null;\n  private contextWindow: number | undefined;\n  private defaultMaxTokens: number;\n\n  constructor(config?: LmStudioConfig) {\n    this.baseUrl = config?.base_url ?? LmStudioBackend.DEFAULT_BASE_URL;\n    this.model = config?.model ?? config?.summary_model ?? 'llama3.2';\n    this.contextWindow = config?.context_window;\n    this.defaultMaxTokens = config?.max_tokens ?? 1024;\n  }\n\n  /**\n   * Generate text using LM Studio's native REST API (/api/v1/chat).\n   * Routes to our specific instance by ID when available, with model name +\n   * context_length as fallback. This ensures correct routing when multiple\n   * daemons share the same LM Studio, and graceful degradation when our\n   * instance is evicted by idle TTL.\n   */\n  async summarize(prompt: string, opts?: LlmRequestOptions): Promise<LlmResponse> {\n    const maxTokens = opts?.maxTokens ?? this.defaultMaxTokens;\n    const contextLength = opts?.contextLength ?? this.contextWindow;\n\n    const body: Record<string, unknown> = {\n      model: this.instanceId ?? this.model,\n      input: prompt,\n      max_output_tokens: maxTokens,\n      store: false,\n    };\n\n    // Always send context_length — even when routing by instance ID.\n    // If our instance was evicted and LM Studio auto-loads, this ensures\n    // the replacement gets the correct context window.\n    if (contextLength) {\n      body.context_length = contextLength;\n    }\n\n    // System prompt — sent separately from user content\n    if (opts?.systemPrompt) {\n      body.system_prompt = opts.systemPrompt;\n    }\n\n    // Reasoning control — 'off' suppresses chain-of-thought for reasoning models\n    if (opts?.reasoning) {\n      body.reasoning = opts.reasoning;\n    }\n\n    const response = await fetch(`${this.baseUrl}${ENDPOINT_CHAT}`, {\n      method: 'POST',\n      headers: { 'Content-Type': 'application/json' },\n      body: JSON.stringify(body),\n      signal: AbortSignal.timeout(opts?.timeoutMs ?? LLM_REQUEST_TIMEOUT_MS),\n    });\n\n    if (!response.ok) {\n      const errorBody = await response.text().catch(() => '');\n      // If our instance was evicted, clear the ID so ensureLoaded\n      // reloads on the next cycle instead of hitting a stale ID repeatedly\n      if (response.status === 404 && this.instanceId) {\n        this.instanceId = null;\n      }\n      throw new Error(`LM Studio summarize failed: ${response.status} ${errorBody.slice(0, 500)}`);\n    }\n\n    const data = await response.json() as {\n      model_instance_id: string;\n      output: Array<{ type: string; content: string }>;\n    };\n    const messageOutput = data.output.find((o) => o.type === 'message');\n    const text = messageOutput?.content ?? '';\n    return { text, model: data.model_instance_id };\n  }\n\n  /**\n   * Generate embeddings using LM Studio's OpenAI-compatible endpoint.\n   * (The native API doesn't have an embedding endpoint — OpenAI-compat is fine here.)\n   */\n  async embed(text: string): Promise<EmbeddingResponse> {\n    const response = await fetch(`${this.baseUrl}${ENDPOINT_EMBEDDINGS}`, {\n      method: 'POST',\n      headers: { 'Content-Type': 'application/json' },\n      body: JSON.stringify({\n        model: this.model,\n        input: text,\n      }),\n      signal: AbortSignal.timeout(EMBEDDING_REQUEST_TIMEOUT_MS),\n    });\n\n    if (!response.ok) {\n      throw new Error(`LM Studio embed failed: ${response.status}`);\n    }\n\n    const data = await response.json() as {\n      data: Array<{ embedding: number[] }>;\n      model: string;\n    };\n    const embedding = data.data[0].embedding;\n    return { embedding, model: data.model, dimensions: embedding.length };\n  }\n\n  /**\n   * Ensure a model instance is loaded with the desired settings.\n   * Called every digest cycle (not cached) so it recovers from idle TTL eviction.\n   *\n   * The load API is necessary to control offload_kv_cache_to_gpu — a load-time\n   * setting that cannot be set per-request via the chat API.\n   *\n   * Multi-daemon safe: finds or loads our own compatible instance without\n   * touching instances from other daemons/projects. Routes by instance ID.\n   */\n  async ensureLoaded(contextLength?: number, gpuKvCache?: boolean): Promise<void> {\n    const ctx = contextLength ?? this.contextWindow;\n    const kvCache = gpuKvCache ?? false;\n\n    // Query native API for existing loaded instances of this model\n    const instances = await this.getLoadedInstances();\n\n    // Look for a compatible instance we can reuse (ours or anyone's)\n    for (const instance of instances) {\n      const matchesContext = !ctx || instance.config.context_length === ctx;\n      const matchesKvCache = instance.config.offload_kv_cache_to_gpu === kvCache;\n      if (matchesContext && matchesKvCache) {\n        this.instanceId = instance.id;\n        return;\n      }\n    }\n\n    // No compatible instance — load our own (don't touch others)\n    const body: Record<string, unknown> = {\n      model: this.model,\n      flash_attention: true,\n      offload_kv_cache_to_gpu: kvCache,\n    };\n    if (ctx) {\n      body.context_length = ctx;\n    }\n\n    const response = await fetch(`${this.baseUrl}${ENDPOINT_MODELS_LOAD}`, {\n      method: 'POST',\n      headers: { 'Content-Type': 'application/json' },\n      body: JSON.stringify(body),\n      signal: AbortSignal.timeout(LLM_REQUEST_TIMEOUT_MS),\n    });\n\n    if (!response.ok) {\n      const errorBody = await response.text().catch(() => '');\n      throw new Error(`LM Studio model load failed: ${response.status} ${errorBody.slice(0, 200)}`);\n    }\n\n    const loadResult = await response.json() as Record<string, unknown>;\n    const id = (loadResult.instance_id ?? loadResult.id ?? loadResult.model_instance_id) as string | undefined;\n    if (id) {\n      this.instanceId = id;\n    }\n  }\n\n  /**\n   * Query the LM Studio native API for loaded instances of this model.\n   * Returns an empty array if the API is unavailable or the model has no loaded instances.\n   */\n  private async getLoadedInstances(): Promise<NativeLoadedInstance[]> {\n    try {\n      const response = await fetch(`${this.baseUrl}${ENDPOINT_MODELS_NATIVE}`, {\n        signal: AbortSignal.timeout(DAEMON_CLIENT_TIMEOUT_MS),\n      });\n      if (!response.ok) return [];\n\n      const data = await response.json() as { models: NativeModelEntry[] };\n      const entry = data.models.find((m) => m.key === this.model);\n      return entry?.loaded_instances ?? [];\n    } catch {\n      return [];\n    }\n  }\n\n  async isAvailable(): Promise<boolean> {\n    try {\n      const response = await fetch(`${this.baseUrl}${ENDPOINT_MODELS_LIST}`, {\n        signal: AbortSignal.timeout(DAEMON_CLIENT_TIMEOUT_MS),\n      });\n      return response.ok;\n    } catch {\n      return false;\n    }\n  }\n\n  /** List available models on this LM Studio instance. */\n  async listModels(timeoutMs?: number): Promise<string[]> {\n    try {\n      const response = await fetch(`${this.baseUrl}${ENDPOINT_MODELS_LIST}`, {\n        signal: AbortSignal.timeout(timeoutMs ?? DAEMON_CLIENT_TIMEOUT_MS),\n      });\n      const data = await response.json() as { data: Array<{ id: string }> };\n      return data.data.map((m) => m.id);\n    } catch {\n      return [];\n    }\n  }\n}\n"],"mappings":";;;;;;;;AAcA,IAAM,oBAAoB;AAC1B,IAAM,iBAAiB;AACvB,IAAM,gBAAgB;AAEf,IAAM,gBAAN,MAAM,eAAwD;AAAA,EACnE,OAAgB,mBAAmB;AAAA,EAC1B,OAAO;AAAA,EACR;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,QAAuB;AACjC,SAAK,UAAU,QAAQ,YAAY,eAAc;AACjD,SAAK,QAAQ,QAAQ,SAAS,QAAQ,iBAAiB;AACvD,SAAK,mBAAmB,QAAQ,cAAc;AAAA,EAChD;AAAA,EAEA,MAAM,UAAU,QAAgB,MAAgD;AAC9E,UAAM,YAAY,MAAM,aAAa,KAAK;AAK1C,UAAM,UAAmC,EAAE,aAAa,UAAU;AAClE,QAAI,MAAM,eAAe;AACvB,cAAQ,UAAU,KAAK;AAAA,IACzB;AAEA,UAAM,OAAgC;AAAA,MACpC,OAAO,KAAK;AAAA,MACZ;AAAA,MACA,QAAQ;AAAA,MACR;AAAA,IACF;AAGA,QAAI,MAAM,cAAc;AACtB,WAAK,SAAS,KAAK;AAAA,IACrB;AAGA,QAAI,MAAM,WAAW;AACnB,WAAK,QAAQ,KAAK,cAAc,QAAQ,QAAQ,KAAK;AAAA,IACvD;AAGA,QAAI,MAAM,WAAW;AACnB,WAAK,aAAa,KAAK;AAAA,IACzB;AAEA,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,iBAAiB,IAAI;AAAA,MAClE,QAAQ;AAAA,MACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,MAC9C,MAAM,KAAK,UAAU,IAAI;AAAA,MACzB,QAAQ,YAAY,QAAQ,MAAM,aAAa,sBAAsB;AAAA,IACvE,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,MAAM,EAAE;AACtD,YAAM,IAAI,MAAM,4BAA4B,SAAS,MAAM,IAAI,UAAU,MAAM,GAAG,GAAG,CAAC,EAAE;AAAA,IAC1F;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,WAAO,EAAE,MAAM,KAAK,UAAU,OAAO,KAAK,MAAM;AAAA,EAClD;AAAA,EAEA,MAAM,MAAM,MAA0C;AACpD,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,cAAc,IAAI;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,MAC9C,MAAM,KAAK,UAAU;AAAA,QACnB,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,MACT,CAAC;AAAA,MACD,QAAQ,YAAY,QAAQ,4BAA4B;AAAA,IAC1D,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,IAAI,MAAM,wBAAwB,SAAS,MAAM,IAAI,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,YAAY,KAAK,WAAW,CAAC;AACnC,WAAO,EAAE,WAAW,OAAO,KAAK,OAAO,YAAY,UAAU,OAAO;AAAA,EACtE;AAAA,EAEA,MAAM,cAAgC;AACpC,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,aAAa,IAAI;AAAA,QAC9D,QAAQ,YAAY,QAAQ,wBAAwB;AAAA,MACtD,CAAC;AACD,aAAO,SAAS;AAAA,IAClB,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,WAAuC;AACtD,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,aAAa,IAAI;AAAA,QAC9D,QAAQ,YAAY,QAAQ,aAAa,wBAAwB;AAAA,MACnE,CAAC;AACD,YAAM,OAAO,MAAM,SAAS,KAAK;AACjC,aAAO,KAAK,OAAO,IAAI,CAAC,MAAM,EAAE,IAAI;AAAA,IACtC,QAAQ;AACN,aAAO,CAAC;AAAA,IACV;AAAA,EACF;AACF;;;AC7GA,IAAM,gBAAgB;AACtB,IAAM,uBAAuB;AAC7B,IAAM,uBAAuB;AAC7B,IAAM,yBAAyB;AAC/B,IAAM,sBAAsB;AAmBrB,IAAM,kBAAN,MAAM,iBAA0D;AAAA,EACrE,OAAgB,mBAAmB;AAAA,EAC1B,OAAO;AAAA,EACR;AAAA,EACA;AAAA,EACA,aAA4B;AAAA,EAC5B;AAAA,EACA;AAAA,EAER,YAAY,QAAyB;AACnC,SAAK,UAAU,QAAQ,YAAY,iBAAgB;AACnD,SAAK,QAAQ,QAAQ,SAAS,QAAQ,iBAAiB;AACvD,SAAK,gBAAgB,QAAQ;AAC7B,SAAK,mBAAmB,QAAQ,cAAc;AAAA,EAChD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAM,UAAU,QAAgB,MAAgD;AAC9E,UAAM,YAAY,MAAM,aAAa,KAAK;AAC1C,UAAM,gBAAgB,MAAM,iBAAiB,KAAK;AAElD,UAAM,OAAgC;AAAA,MACpC,OAAO,KAAK,cAAc,KAAK;AAAA,MAC/B,OAAO;AAAA,MACP,mBAAmB;AAAA,MACnB,OAAO;AAAA,IACT;AAKA,QAAI,eAAe;AACjB,WAAK,iBAAiB;AAAA,IACxB;AAGA,QAAI,MAAM,cAAc;AACtB,WAAK,gBAAgB,KAAK;AAAA,IAC5B;AAGA,QAAI,MAAM,WAAW;AACnB,WAAK,YAAY,KAAK;AAAA,IACxB;AAEA,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,aAAa,IAAI;AAAA,MAC9D,QAAQ;AAAA,MACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,MAC9C,MAAM,KAAK,UAAU,IAAI;AAAA,MACzB,QAAQ,YAAY,QAAQ,MAAM,aAAa,sBAAsB;AAAA,IACvE,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,MAAM,EAAE;AAGtD,UAAI,SAAS,WAAW,OAAO,KAAK,YAAY;AAC9C,aAAK,aAAa;AAAA,MACpB;AACA,YAAM,IAAI,MAAM,+BAA+B,SAAS,MAAM,IAAI,UAAU,MAAM,GAAG,GAAG,CAAC,EAAE;AAAA,IAC7F;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAIjC,UAAM,gBAAgB,KAAK,OAAO,KAAK,CAAC,MAAM,EAAE,SAAS,SAAS;AAClE,UAAM,OAAO,eAAe,WAAW;AACvC,WAAO,EAAE,MAAM,OAAO,KAAK,kBAAkB;AAAA,EAC/C;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,MAAM,MAA0C;AACpD,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,mBAAmB,IAAI;AAAA,MACpE,QAAQ;AAAA,MACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,MAC9C,MAAM,KAAK,UAAU;AAAA,QACnB,OAAO,KAAK;AAAA,QACZ,OAAO;AAAA,MACT,CAAC;AAAA,MACD,QAAQ,YAAY,QAAQ,4BAA4B;AAAA,IAC1D,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,IAAI,MAAM,2BAA2B,SAAS,MAAM,EAAE;AAAA,IAC9D;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAIjC,UAAM,YAAY,KAAK,KAAK,CAAC,EAAE;AAC/B,WAAO,EAAE,WAAW,OAAO,KAAK,OAAO,YAAY,UAAU,OAAO;AAAA,EACtE;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAM,aAAa,eAAwB,YAAqC;AAC9E,UAAM,MAAM,iBAAiB,KAAK;AAClC,UAAM,UAAU,cAAc;AAG9B,UAAM,YAAY,MAAM,KAAK,mBAAmB;AAGhD,eAAW,YAAY,WAAW;AAChC,YAAM,iBAAiB,CAAC,OAAO,SAAS,OAAO,mBAAmB;AAClE,YAAM,iBAAiB,SAAS,OAAO,4BAA4B;AACnE,UAAI,kBAAkB,gBAAgB;AACpC,aAAK,aAAa,SAAS;AAC3B;AAAA,MACF;AAAA,IACF;AAGA,UAAM,OAAgC;AAAA,MACpC,OAAO,KAAK;AAAA,MACZ,iBAAiB;AAAA,MACjB,yBAAyB;AAAA,IAC3B;AACA,QAAI,KAAK;AACP,WAAK,iBAAiB;AAAA,IACxB;AAEA,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,oBAAoB,IAAI;AAAA,MACrE,QAAQ;AAAA,MACR,SAAS,EAAE,gBAAgB,mBAAmB;AAAA,MAC9C,MAAM,KAAK,UAAU,IAAI;AAAA,MACzB,QAAQ,YAAY,QAAQ,sBAAsB;AAAA,IACpD,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,MAAM,EAAE;AACtD,YAAM,IAAI,MAAM,gCAAgC,SAAS,MAAM,IAAI,UAAU,MAAM,GAAG,GAAG,CAAC,EAAE;AAAA,IAC9F;AAEA,UAAM,aAAa,MAAM,SAAS,KAAK;AACvC,UAAM,KAAM,WAAW,eAAe,WAAW,MAAM,WAAW;AAClE,QAAI,IAAI;AACN,WAAK,aAAa;AAAA,IACpB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAc,qBAAsD;AAClE,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,sBAAsB,IAAI;AAAA,QACvE,QAAQ,YAAY,QAAQ,wBAAwB;AAAA,MACtD,CAAC;AACD,UAAI,CAAC,SAAS,GAAI,QAAO,CAAC;AAE1B,YAAM,OAAO,MAAM,SAAS,KAAK;AACjC,YAAM,QAAQ,KAAK,OAAO,KAAK,CAAC,MAAM,EAAE,QAAQ,KAAK,KAAK;AAC1D,aAAO,OAAO,oBAAoB,CAAC;AAAA,IACrC,QAAQ;AACN,aAAO,CAAC;AAAA,IACV;AAAA,EACF;AAAA,EAEA,MAAM,cAAgC;AACpC,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,oBAAoB,IAAI;AAAA,QACrE,QAAQ,YAAY,QAAQ,wBAAwB;AAAA,MACtD,CAAC;AACD,aAAO,SAAS;AAAA,IAClB,QAAQ;AACN,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA,EAGA,MAAM,WAAW,WAAuC;AACtD,QAAI;AACF,YAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,GAAG,oBAAoB,IAAI;AAAA,QACrE,QAAQ,YAAY,QAAQ,aAAa,wBAAwB;AAAA,MACnE,CAAC;AACD,YAAM,OAAO,MAAM,SAAS,KAAK;AACjC,aAAO,KAAK,KAAK,IAAI,CAAC,MAAM,EAAE,EAAE;AAAA,IAClC,QAAQ;AACN,aAAO,CAAC;AAAA,IACV;AAAA,EACF;AACF;","names":[]}

package/dist/{chunk-IYFKPSRP.js → chunk-OSZRLHIJ.js}

@@ -2,10 +2,10 @@ import { createRequire as __cr } from 'node:module'; const require = __cr(import
 import {
   LmStudioBackend,
   OllamaBackend
-} from "./chunk-67R6EMYD.js";
+} from "./chunk-OPO47BVS.js";
 import {
   LLM_REQUEST_TIMEOUT_MS
-} from "./chunk-JBD5KP5G.js";
+} from "./chunk-TDLQBGKA.js";
 
 // node_modules/@anthropic-ai/sdk/internal/tslib.mjs
 function __classPrivateFieldSet(receiver, state, value, kind, f) {
@@ -4911,4 +4911,4 @@ export {
   createLlmProvider,
   createEmbeddingProvider
 };
-//# sourceMappingURL=chunk-IYFKPSRP.js.map
+//# sourceMappingURL=chunk-OSZRLHIJ.js.map

package/dist/chunk-PD7LV22R.js

@@ -0,0 +1,150 @@
+import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
+import {
+  ARTIFACT_TYPES
+} from "./chunk-2AMAOSRF.js";
+import {
+  CANDIDATE_CONTENT_PREVIEW
+} from "./chunk-TDLQBGKA.js";
+
+// src/prompts/index.ts
+import fs from "fs";
+import path from "path";
+import { fileURLToPath } from "url";
+function resolvePromptsDir() {
+  let dir = path.dirname(fileURLToPath(import.meta.url));
+  for (let i = 0; i < 5; i++) {
+    if (fs.existsSync(path.join(dir, "package.json"))) {
+      return path.join(dir, "dist", "src", "prompts");
+    }
+    if (fs.existsSync(path.join(dir, "extraction.md"))) {
+      return dir;
+    }
+    dir = path.dirname(dir);
+  }
+  return path.dirname(fileURLToPath(import.meta.url));
+}
+var PROMPTS_DIR = resolvePromptsDir();
+var promptCache = /* @__PURE__ */ new Map();
+function loadPrompt(name) {
+  let cached = promptCache.get(name);
+  if (!cached) {
+    cached = fs.readFileSync(path.join(PROMPTS_DIR, `${name}.md`), "utf-8").trim();
+    promptCache.set(name, cached);
+  }
+  return cached;
+}
+function interpolate(template, vars) {
+  let result = template;
+  for (const [key, value] of Object.entries(vars)) {
+    result = result.replaceAll(`{{${key}}}`, value);
+  }
+  return result;
+}
+function buildExtractionPrompt(sessionId, eventCount, toolSummary, maxTokens) {
+  return interpolate(loadPrompt("extraction"), {
+    sessionId,
+    eventCount: String(eventCount),
+    toolSummary,
+    maxTokens: String(maxTokens ?? 2048)
+  });
+}
+function buildSummaryPrompt(sessionId, user, content, maxTokens) {
+  return interpolate(loadPrompt("summary"), {
+    sessionId,
+    user,
+    content,
+    maxTokens: String(maxTokens ?? 1024)
+  });
+}
+function buildTitlePrompt(summary, sessionId) {
+  return interpolate(loadPrompt("title"), {
+    summary,
+    sessionId
+  });
+}
+var ARTIFACT_TYPE_DESCRIPTIONS = [
+  '"spec" \u2014 Design specifications, architecture documents',
+  '"plan" \u2014 Implementation plans, roadmaps',
+  '"rfc" \u2014 Requests for comment, proposals',
+  '"doc" \u2014 Documentation, guides, READMEs',
+  '"other" \u2014 Other substantive documents'
+];
+function buildSimilarityPrompt(currentSummary, candidateSummary) {
+  return interpolate(loadPrompt("session-similarity"), {
+    currentSummary,
+    candidateSummary
+  });
+}
+function buildClassificationPrompt(sessionId, candidates, maxTokens) {
+  const fileList = candidates.map((c) => {
+    const truncated = c.content.slice(0, CANDIDATE_CONTENT_PREVIEW);
+    return `### ${c.path}
+\`\`\`
+${truncated}
+\`\`\``;
+  }).join("\n\n");
+  return interpolate(loadPrompt("classification"), {
+    sessionId,
+    fileList,
+    artifactTypes: ARTIFACT_TYPE_DESCRIPTIONS.map((d) => `- ${d}`).join("\n"),
+    validTypes: ARTIFACT_TYPES.join("|"),
+    maxTokens: String(maxTokens ?? 1024)
+  });
+}
+
+// src/intelligence/response.ts
+var REASONING_PATTERNS = [
+  // <think>...</think>answer  (DeepSeek, Qwen, GLM, many others)
+  /<think>[\s\S]*?<\/think>\s*/gi,
+  // Implicit opening: reasoning...</think>answer  (GLM-4.7 observed)
+  /^[\s\S]*?<\/think>\s*/i,
+  // <reasoning>...</reasoning>answer
+  /<reasoning>[\s\S]*?<\/reasoning>\s*/gi,
+  // <|thinking|>...<|/thinking|>answer
+  /<\|thinking\|>[\s\S]*?<\|\/thinking\|>\s*/gi,
+  // Plain-text "Thinking Process:" block followed by actual content
+  // (Qwen 3.5 via LM Studio without native thinking mode)
+  // Matches from "Thinking Process:" up to the last numbered step, then the synthesis follows
+  /^Thinking Process:[\s\S]*?(?=\n(?:## |# |\*\*[A-Z]))/i
+];
+function stripReasoningTokens(text) {
+  if (!text) return text;
+  for (const pattern of REASONING_PATTERNS) {
+    const stripped = text.replace(pattern, "").trim();
+    if (stripped && stripped !== text.trim()) {
+      return stripped;
+    }
+  }
+  return text;
+}
+function extractJson(text) {
+  const cleaned = stripReasoningTokens(text);
+  const fenceMatch = cleaned.match(/```(?:json)?\s*\n?([\s\S]*?)\n?```/);
+  if (fenceMatch) {
+    return JSON.parse(fenceMatch[1].trim());
+  }
+  const objectMatch = cleaned.match(/\{[\s\S]*\}/);
+  if (objectMatch) {
+    return JSON.parse(objectMatch[0]);
+  }
+  return JSON.parse(cleaned);
+}
+function extractNumber(text) {
+  const cleaned = stripReasoningTokens(text).trim();
+  const match = cleaned.match(/(\d+\.?\d*)/);
+  if (match) return parseFloat(match[1]);
+  return parseFloat(cleaned);
+}
+
+export {
+  loadPrompt,
+  buildExtractionPrompt,
+  buildSummaryPrompt,
+  buildTitlePrompt,
+  buildSimilarityPrompt,
+  buildClassificationPrompt,
+  stripReasoningTokens,
+  extractJson,
+  extractNumber
+};
+//# sourceMappingURL=chunk-PD7LV22R.js.map

package/dist/chunk-PD7LV22R.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/prompts/index.ts","../src/intelligence/response.ts"],"sourcesContent":["/**\n * Prompt loader — reads .md templates from disk and interpolates variables.\n * Prompts are markdown files in this directory, not TypeScript strings.\n */\n\nimport fs from 'node:fs';\nimport path from 'node:path';\nimport { fileURLToPath } from 'node:url';\nimport { ARTIFACT_TYPES } from '../vault/types.js';\nimport { CANDIDATE_CONTENT_PREVIEW } from '../constants.js';\n\n/**\n * Resolve the prompts directory. With tsup code-splitting, import.meta.url\n * points to a chunk file (dist/chunk-XXXX.js), not dist/src/prompts/.\n * Walk up from the current file to find package.json, then use dist/src/prompts/.\n */\nfunction resolvePromptsDir(): string {\n  let dir = path.dirname(fileURLToPath(import.meta.url));\n  for (let i = 0; i < 5; i++) {\n    if (fs.existsSync(path.join(dir, 'package.json'))) {\n      return path.join(dir, 'dist', 'src', 'prompts');\n    }\n    // Also check if we're already in the right place (tsc output or dev mode)\n    if (fs.existsSync(path.join(dir, 'extraction.md'))) {\n      return dir;\n    }\n    dir = path.dirname(dir);\n  }\n  // Final fallback: adjacent to current file (works with tsc)\n  return path.dirname(fileURLToPath(import.meta.url));\n}\n\nconst PROMPTS_DIR = resolvePromptsDir();\n\nconst promptCache = new Map<string, string>();\n\nexport function loadPrompt(name: string): string {\n  let cached = promptCache.get(name);\n  if (!cached) {\n    cached = fs.readFileSync(path.join(PROMPTS_DIR, `${name}.md`), 'utf-8').trim();\n    promptCache.set(name, cached);\n  }\n  return cached;\n}\n\nfunction interpolate(template: string, vars: Record<string, string>): string {\n  let result = template;\n  for (const [key, value] of Object.entries(vars)) {\n    result = result.replaceAll(`{{${key}}}`, value);\n  }\n  return result;\n}\n\n// --- Prompt builders ---\n\nexport function buildExtractionPrompt(\n  sessionId: string,\n  eventCount: number,\n  toolSummary: string,\n  maxTokens?: number,\n): string {\n  return interpolate(loadPrompt('extraction'), {\n    sessionId,\n    eventCount: String(eventCount),\n    toolSummary,\n    maxTokens: String(maxTokens ?? 2048),\n  });\n}\n\nexport function buildSummaryPrompt(\n  sessionId: string,\n  user: string,\n  content: string,\n  maxTokens?: number,\n): string {\n  return interpolate(loadPrompt('summary'), {\n    sessionId,\n    user,\n    content,\n    maxTokens: String(maxTokens ?? 1024),\n  });\n}\n\nexport function buildTitlePrompt(\n  summary: string,\n  sessionId: string,\n): string {\n  return interpolate(loadPrompt('title'), {\n    summary,\n    sessionId,\n  });\n}\n\nconst ARTIFACT_TYPE_DESCRIPTIONS = [\n  '\"spec\" — Design specifications, architecture documents',\n  '\"plan\" — Implementation plans, roadmaps',\n  '\"rfc\" — Requests for comment, proposals',\n  '\"doc\" — Documentation, guides, READMEs',\n  '\"other\" — Other substantive documents',\n];\n\nexport function buildSimilarityPrompt(\n  currentSummary: string,\n  candidateSummary: string,\n): string {\n  return interpolate(loadPrompt('session-similarity'), {\n    currentSummary,\n    candidateSummary,\n  });\n}\n\nexport function buildClassificationPrompt(\n  sessionId: string,\n  candidates: Array<{ path: string; content: string }>,\n  maxTokens?: number,\n): string {\n  const fileList = candidates\n    .map((c) => {\n      const truncated = c.content.slice(0, CANDIDATE_CONTENT_PREVIEW);\n      return `### ${c.path}\\n\\`\\`\\`\\n${truncated}\\n\\`\\`\\``;\n    })\n    .join('\\n\\n');\n\n  return interpolate(loadPrompt('classification'), {\n    sessionId,\n    fileList,\n    artifactTypes: ARTIFACT_TYPE_DESCRIPTIONS.map((d) => `- ${d}`).join('\\n'),\n    validTypes: ARTIFACT_TYPES.join('|'),\n    maxTokens: String(maxTokens ?? 1024),\n  });\n}\n","/**\n * Clean LLM response text before parsing.\n *\n * Reasoning models (DeepSeek, Qwen, GLM, etc.) embed chain-of-thought\n * in the response using special tags. These must be stripped before\n * JSON parsing or value extraction.\n */\n\n// Patterns for reasoning model chain-of-thought tokens.\n// Order matters: most specific patterns first.\nconst REASONING_PATTERNS = [\n  // <think>...</think>answer  (DeepSeek, Qwen, GLM, many others)\n  /<think>[\\s\\S]*?<\\/think>\\s*/gi,\n  // Implicit opening: reasoning...</think>answer  (GLM-4.7 observed)\n  /^[\\s\\S]*?<\\/think>\\s*/i,\n  // <reasoning>...</reasoning>answer\n  /<reasoning>[\\s\\S]*?<\\/reasoning>\\s*/gi,\n  // <|thinking|>...<|/thinking|>answer\n  /<\\|thinking\\|>[\\s\\S]*?<\\|\\/thinking\\|>\\s*/gi,\n  // Plain-text \"Thinking Process:\" block followed by actual content\n  // (Qwen 3.5 via LM Studio without native thinking mode)\n  // Matches from \"Thinking Process:\" up to the last numbered step, then the synthesis follows\n  /^Thinking Process:[\\s\\S]*?(?=\\n(?:## |# |\\*\\*[A-Z]))/i,\n];\n\n/**\n * Strip reasoning/chain-of-thought tokens from LLM response text.\n * Returns the final answer without the thinking process.\n */\nexport function stripReasoningTokens(text: string): string {\n  if (!text) return text;\n\n  for (const pattern of REASONING_PATTERNS) {\n    const stripped = text.replace(pattern, '').trim();\n    if (stripped && stripped !== text.trim()) {\n      return stripped;\n    }\n  }\n\n  return text;\n}\n\n/**\n * Extract JSON from an LLM response that may contain markdown fences,\n * reasoning tokens, or other wrapper text.\n *\n * Tries in order:\n * 1. Strip reasoning tokens\n * 2. Extract from ```json ... ``` code fences\n * 3. Find bare {...} JSON object\n * 4. Parse the cleaned text directly\n */\nexport function extractJson(text: string): unknown {\n  const cleaned = stripReasoningTokens(text);\n\n  // Try code fence extraction\n  const fenceMatch = cleaned.match(/```(?:json)?\\s*\\n?([\\s\\S]*?)\\n?```/);\n  if (fenceMatch) {\n    return JSON.parse(fenceMatch[1].trim());\n  }\n\n  // Try bare JSON object\n  const objectMatch = cleaned.match(/\\{[\\s\\S]*\\}/);\n  if (objectMatch) {\n    return JSON.parse(objectMatch[0]);\n  }\n\n  // Try direct parse\n  return JSON.parse(cleaned);\n}\n\n/**\n * Extract a numeric value from an LLM response that may contain\n * reasoning tokens or extra text around the number.\n */\nexport function extractNumber(text: string): number {\n  const cleaned = stripReasoningTokens(text).trim();\n  const match = cleaned.match(/(\\d+\\.?\\d*)/);\n  if (match) return parseFloat(match[1]);\n  return parseFloat(cleaned);\n}\n"],"mappings":";;;;;;;;;AAKA,OAAO,QAAQ;AACf,OAAO,UAAU;AACjB,SAAS,qBAAqB;AAS9B,SAAS,oBAA4B;AACnC,MAAI,MAAM,KAAK,QAAQ,cAAc,YAAY,GAAG,CAAC;AACrD,WAAS,IAAI,GAAG,IAAI,GAAG,KAAK;AAC1B,QAAI,GAAG,WAAW,KAAK,KAAK,KAAK,cAAc,CAAC,GAAG;AACjD,aAAO,KAAK,KAAK,KAAK,QAAQ,OAAO,SAAS;AAAA,IAChD;AAEA,QAAI,GAAG,WAAW,KAAK,KAAK,KAAK,eAAe,CAAC,GAAG;AAClD,aAAO;AAAA,IACT;AACA,UAAM,KAAK,QAAQ,GAAG;AAAA,EACxB;AAEA,SAAO,KAAK,QAAQ,cAAc,YAAY,GAAG,CAAC;AACpD;AAEA,IAAM,cAAc,kBAAkB;AAEtC,IAAM,cAAc,oBAAI,IAAoB;AAErC,SAAS,WAAW,MAAsB;AAC/C,MAAI,SAAS,YAAY,IAAI,IAAI;AACjC,MAAI,CAAC,QAAQ;AACX,aAAS,GAAG,aAAa,KAAK,KAAK,aAAa,GAAG,IAAI,KAAK,GAAG,OAAO,EAAE,KAAK;AAC7E,gBAAY,IAAI,MAAM,MAAM;AAAA,EAC9B;AACA,SAAO;AACT;AAEA,SAAS,YAAY,UAAkB,MAAsC;AAC3E,MAAI,SAAS;AACb,aAAW,CAAC,KAAK,KAAK,KAAK,OAAO,QAAQ,IAAI,GAAG;AAC/C,aAAS,OAAO,WAAW,KAAK,GAAG,MAAM,KAAK;AAAA,EAChD;AACA,SAAO;AACT;AAIO,SAAS,sBACd,WACA,YACA,aACA,WACQ;AACR,SAAO,YAAY,WAAW,YAAY,GAAG;AAAA,IAC3C;AAAA,IACA,YAAY,OAAO,UAAU;AAAA,IAC7B;AAAA,IACA,WAAW,OAAO,aAAa,IAAI;AAAA,EACrC,CAAC;AACH;AAEO,SAAS,mBACd,WACA,MACA,SACA,WACQ;AACR,SAAO,YAAY,WAAW,SAAS,GAAG;AAAA,IACxC;AAAA,IACA;AAAA,IACA;AAAA,IACA,WAAW,OAAO,aAAa,IAAI;AAAA,EACrC,CAAC;AACH;AAEO,SAAS,iBACd,SACA,WACQ;AACR,SAAO,YAAY,WAAW,OAAO,GAAG;AAAA,IACtC;AAAA,IACA;AAAA,EACF,CAAC;AACH;AAEA,IAAM,6BAA6B;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAEO,SAAS,sBACd,gBACA,kBACQ;AACR,SAAO,YAAY,WAAW,oBAAoB,GAAG;AAAA,IACnD;AAAA,IACA;AAAA,EACF,CAAC;AACH;AAEO,SAAS,0BACd,WACA,YACA,WACQ;AACR,QAAM,WAAW,WACd,IAAI,CAAC,MAAM;AACV,UAAM,YAAY,EAAE,QAAQ,MAAM,GAAG,yBAAyB;AAC9D,WAAO,OAAO,EAAE,IAAI;AAAA;AAAA,EAAa,SAAS;AAAA;AAAA,EAC5C,CAAC,EACA,KAAK,MAAM;AAEd,SAAO,YAAY,WAAW,gBAAgB,GAAG;AAAA,IAC/C;AAAA,IACA;AAAA,IACA,eAAe,2BAA2B,IAAI,CAAC,MAAM,KAAK,CAAC,EAAE,EAAE,KAAK,IAAI;AAAA,IACxE,YAAY,eAAe,KAAK,GAAG;AAAA,IACnC,WAAW,OAAO,aAAa,IAAI;AAAA,EACrC,CAAC;AACH;;;ACxHA,IAAM,qBAAqB;AAAA;AAAA,EAEzB;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA;AAAA;AAAA;AAAA,EAIA;AACF;AAMO,SAAS,qBAAqB,MAAsB;AACzD,MAAI,CAAC,KAAM,QAAO;AAElB,aAAW,WAAW,oBAAoB;AACxC,UAAM,WAAW,KAAK,QAAQ,SAAS,EAAE,EAAE,KAAK;AAChD,QAAI,YAAY,aAAa,KAAK,KAAK,GAAG;AACxC,aAAO;AAAA,IACT;AAAA,EACF;AAEA,SAAO;AACT;AAYO,SAAS,YAAY,MAAuB;AACjD,QAAM,UAAU,qBAAqB,IAAI;AAGzC,QAAM,aAAa,QAAQ,MAAM,oCAAoC;AACrE,MAAI,YAAY;AACd,WAAO,KAAK,MAAM,WAAW,CAAC,EAAE,KAAK,CAAC;AAAA,EACxC;AAGA,QAAM,cAAc,QAAQ,MAAM,aAAa;AAC/C,MAAI,aAAa;AACf,WAAO,KAAK,MAAM,YAAY,CAAC,CAAC;AAAA,EAClC;AAGA,SAAO,KAAK,MAAM,OAAO;AAC3B;AAMO,SAAS,cAAc,MAAsB;AAClD,QAAM,UAAU,qBAAqB,IAAI,EAAE,KAAK;AAChD,QAAM,QAAQ,QAAQ,MAAM,aAAa;AACzC,MAAI,MAAO,QAAO,WAAW,MAAM,CAAC,CAAC;AACrC,SAAO,WAAW,OAAO;AAC3B;","names":[]}

package/dist/{chunk-JBD5KP5G.js → chunk-TDLQBGKA.js}

@@ -26,6 +26,7 @@ var FILE_WATCH_STABILITY_MS = 1e3;
 var PROVIDER_DETECT_TIMEOUT_MS = 3e3;
 var STALE_BUFFER_MAX_AGE_MS = 24 * 60 * 60 * 1e3;
 var DAEMON_HEALTH_RETRY_DELAYS = [100, 200, 400, 800, 1500];
+var DAEMON_STALE_GRACE_PERIOD_MS = 6e4;
 var MAX_SLUG_LENGTH = 100;
 var CANDIDATE_CONTENT_PREVIEW = 2e3;
 var LINEAGE_RECENT_SESSIONS_LIMIT = 5;
@@ -51,6 +52,7 @@ var DIGEST_SUBSTRATE_TYPE_WEIGHTS = {
   artifact: 1,
   team: 1
 };
+var LLM_REASONING_MODE = "off";
 
 export {
   CHARS_PER_TOKEN,
@@ -76,6 +78,7 @@ export {
   PROVIDER_DETECT_TIMEOUT_MS,
   STALE_BUFFER_MAX_AGE_MS,
   DAEMON_HEALTH_RETRY_DELAYS,
+  DAEMON_STALE_GRACE_PERIOD_MS,
   MAX_SLUG_LENGTH,
   CANDIDATE_CONTENT_PREVIEW,
   LINEAGE_RECENT_SESSIONS_LIMIT,
@@ -89,6 +92,7 @@ export {
   MCP_LOGS_DEFAULT_LIMIT,
   DIGEST_TIERS,
   DIGEST_TIER_MIN_CONTEXT,
-  DIGEST_SUBSTRATE_TYPE_WEIGHTS
+  DIGEST_SUBSTRATE_TYPE_WEIGHTS,
+  LLM_REASONING_MODE
 };
-//# sourceMappingURL=chunk-JBD5KP5G.js.map
+//# sourceMappingURL=chunk-TDLQBGKA.js.map

package/dist/chunk-TDLQBGKA.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/constants.ts"],"sourcesContent":["/**\n * Shared constants for the Myco codebase.\n * Per CLAUDE.md: \"No Magic Literals — Numeric and string constants\n * MUST NOT appear inline in logic.\"\n */\n\n// --- Token estimation ---\n/** Approximate characters per token for the chars/4 heuristic. */\nexport const CHARS_PER_TOKEN = 4;\n\n/** Estimate token count from character length using the CHARS_PER_TOKEN heuristic. */\nexport function estimateTokens(text: string): number {\n  return Math.ceil(text.length / CHARS_PER_TOKEN);\n}\n\n// --- Embedding ---\n/** Max characters of text sent to the embedding model. */\nexport const EMBEDDING_INPUT_LIMIT = 8000;\n\n// --- Truncation limits (display/preview) ---\n/** Max chars for a user prompt preview in event summaries. */\nexport const PROMPT_PREVIEW_CHARS = 300;\n/** Max chars for an AI response preview in event summaries. */\nexport const AI_RESPONSE_PREVIEW_CHARS = 500;\n/** Max chars for a command string preview. */\nexport const COMMAND_PREVIEW_CHARS = 80;\n/** Max chars for a content snippet in search results. */\nexport const CONTENT_SNIPPET_CHARS = 120;\n/** Max chars for a tool output preview in hooks. */\nexport const TOOL_OUTPUT_PREVIEW_CHARS = 200;\n/** Max chars for a session summary preview in MCP tools. */\nexport const SESSION_SUMMARY_PREVIEW_CHARS = 300;\n/** Max chars for a recall summary preview. */\nexport const RECALL_SUMMARY_PREVIEW_CHARS = 200;\n\n// --- Context injection layer budgets (chars, not tokens — used with .slice()) ---\nexport const CONTEXT_PLAN_PREVIEW_CHARS = 100;\nexport const CONTEXT_SESSION_PREVIEW_CHARS = 80;\nexport const CONTEXT_SPORE_PREVIEW_CHARS = 80;\n\n// --- Processor maxTokens budgets ---\n/** Response token budget for observation extraction. */\nexport const EXTRACTION_MAX_TOKENS = 2048;\n/** Response token budget for session summary. */\nexport const SUMMARY_MAX_TOKENS = 512;\n/** Response token budget for session title generation. */\nexport const TITLE_MAX_TOKENS = 32;\n/** Response token budget for artifact classification. */\nexport const CLASSIFICATION_MAX_TOKENS = 1024;\n\n// --- Timeouts ---\n/** Daemon client HTTP request timeout (ms). */\nexport const DAEMON_CLIENT_TIMEOUT_MS = 2000;\n/** Health check timeout (ms) — fail fast if daemon isn't responding. */\nexport const DAEMON_HEALTH_CHECK_TIMEOUT_MS = 500;\n/** LLM request timeout (ms). All LLM calls are background daemon work — no need to be aggressive. */\nexport const LLM_REQUEST_TIMEOUT_MS = 180_000;\n/** Embedding request timeout (ms). Embeddings run in background batch processing — generous timeout. */\nexport const EMBEDDING_REQUEST_TIMEOUT_MS = 60_000;\n/** Digest LLM request timeout (ms). Digest cycles use large context windows and may need model loading time. */\nexport const DIGEST_LLM_REQUEST_TIMEOUT_MS = 600_000;\n/** Stdin read timeout for hooks (ms). */\nexport const STDIN_TIMEOUT_MS = 100;\n/** Chokidar write stability threshold (ms). */\nexport const FILE_WATCH_STABILITY_MS = 1000;\n/** Provider detection timeout for detect-providers CLI command (ms). */\nexport const PROVIDER_DETECT_TIMEOUT_MS = 3000;\n\n// --- Buffer cleanup ---\n/** Max age for stale buffer files before cleanup (ms). */\nexport const STALE_BUFFER_MAX_AGE_MS = 24 * 60 * 60 * 1000;\n\n// --- Retry backoff ---\n/** Retry delays for daemon health check (ms). */\nexport const DAEMON_HEALTH_RETRY_DELAYS = [100, 200, 400, 800, 1500];\n\n/** Grace period after daemon.json is written before stale checks can trigger a restart (ms).\n *  Prevents rapid restart loops from concurrent hooks or session reloads. */\nexport const DAEMON_STALE_GRACE_PERIOD_MS = 60_000;\n\n// --- Slug limits ---\n/** Max length for slugified artifact IDs. */\nexport const MAX_SLUG_LENGTH = 100;\n\n// --- Content preview for classification prompt ---\n/** Max chars of file content per candidate in classification prompt. */\nexport const CANDIDATE_CONTENT_PREVIEW = 2000;\n\n// --- Transcript mining ---\n/** Minimum content length to consider a transcript entry meaningful. */\nexport const MIN_TRANSCRIPT_CONTENT_LENGTH = 10;\n\n// --- Query limits ---\n/** Max recent sessions to check for lineage heuristics. */\nexport const LINEAGE_RECENT_SESSIONS_LIMIT = 5;\n/** Max related spores to query for session notes. */\nexport const RELATED_SPORES_LIMIT = 50;\n\n// --- Context injection ---\n/** Max active plans to inject at session start. */\nexport const SESSION_CONTEXT_MAX_PLANS = 3;\n/** Max spores to inject per prompt. */\nexport const PROMPT_CONTEXT_MAX_SPORES = 3;\n/** Minimum similarity score for prompt context injection (0-1). */\nexport const PROMPT_CONTEXT_MIN_SIMILARITY = 0.3;\n/** Max token budget for session-start context injection. */\nexport const SESSION_CONTEXT_MAX_TOKENS = 500;\n/** Max token budget for per-prompt context injection. */\nexport const PROMPT_CONTEXT_MAX_TOKENS = 300;\n/** Minimum prompt length to trigger context search. */\nexport const PROMPT_CONTEXT_MIN_LENGTH = 10;\n\n// --- MCP tool defaults ---\n/** Default result limit for myco_search. */\nexport const MCP_SEARCH_DEFAULT_LIMIT = 10;\n/** Default result limit for myco_sessions. */\nexport const MCP_SESSIONS_DEFAULT_LIMIT = 20;\n/** Default result limit for myco_logs. */\nexport const MCP_LOGS_DEFAULT_LIMIT = 50;\n\n// --- Digest — Tiers ---\n/** Available token-budget tiers for digest synthesis. */\nexport const DIGEST_TIERS = [1500, 3000, 5000, 10000] as const;\nexport type DigestTier = (typeof DIGEST_TIERS)[number];\n\n// --- Digest — Context window minimums per tier ---\n/** Minimum context window (tokens) required to run a digest at a given tier. */\nexport const DIGEST_TIER_MIN_CONTEXT: Record<number, number> = {\n  1500: 6500,\n  3000: 11500,\n  5000: 18500,\n  10000: 30500,\n};\n\n// --- Digest — Substrate ---\n/** Scoring weights by note type when selecting substrate for synthesis. */\nexport const DIGEST_SUBSTRATE_TYPE_WEIGHTS: Record<string, number> = {\n  session: 3,\n  spore: 3,\n  plan: 2,\n  artifact: 1,\n  team: 1,\n};\n\n// --- LLM reasoning control ---\n/** Reasoning mode for all Myco LLM calls. Suppresses chain-of-thought tokens from reasoning models. */\nexport const LLM_REASONING_MODE = 'off' as const;\n\n// --- Digest — System prompt overhead estimate ---\n"],"mappings":";;;AAQO,IAAM,kBAAkB;AAGxB,SAAS,eAAe,MAAsB;AACnD,SAAO,KAAK,KAAK,KAAK,SAAS,eAAe;AAChD;AAIO,IAAM,wBAAwB;AAI9B,IAAM,uBAAuB;AAE7B,IAAM,4BAA4B;AAElC,IAAM,wBAAwB;AAE9B,IAAM,wBAAwB;AAE9B,IAAM,4BAA4B;AAElC,IAAM,gCAAgC;AAEtC,IAAM,+BAA+B;AAGrC,IAAM,6BAA6B;AACnC,IAAM,gCAAgC;AACtC,IAAM,8BAA8B;AAcpC,IAAM,2BAA2B;AAEjC,IAAM,iCAAiC;AAEvC,IAAM,yBAAyB;AAE/B,IAAM,+BAA+B;AAErC,IAAM,gCAAgC;AAEtC,IAAM,mBAAmB;AAEzB,IAAM,0BAA0B;AAEhC,IAAM,6BAA6B;AAInC,IAAM,0BAA0B,KAAK,KAAK,KAAK;AAI/C,IAAM,6BAA6B,CAAC,KAAK,KAAK,KAAK,KAAK,IAAI;AAI5D,IAAM,+BAA+B;AAIrC,IAAM,kBAAkB;AAIxB,IAAM,4BAA4B;AAQlC,IAAM,gCAAgC;AAEtC,IAAM,uBAAuB;AAI7B,IAAM,4BAA4B;AAElC,IAAM,4BAA4B;AAElC,IAAM,gCAAgC;AAMtC,IAAM,4BAA4B;AAIlC,IAAM,2BAA2B;AAEjC,IAAM,6BAA6B;AAEnC,IAAM,yBAAyB;AAI/B,IAAM,eAAe,CAAC,MAAM,KAAM,KAAM,GAAK;AAK7C,IAAM,0BAAkD;AAAA,EAC7D,MAAM;AAAA,EACN,KAAM;AAAA,EACN,KAAM;AAAA,EACN,KAAO;AACT;AAIO,IAAM,gCAAwD;AAAA,EACnE,SAAS;AAAA,EACT,OAAO;AAAA,EACP,MAAM;AAAA,EACN,UAAU;AAAA,EACV,MAAM;AACR;AAIO,IAAM,qBAAqB;","names":[]}

package/dist/{chunk-2GJFTIWX.js → chunk-TK2ZYIAL.js}

@@ -1,7 +1,7 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 import {
   AgentRegistry
-} from "./chunk-BNIYWCST.js";
+} from "./chunk-EQVQEFOA.js";
 
 // src/version.ts
 import fs from "fs";
@@ -30,4 +30,4 @@ function readVersionFrom(dir) {
 export {
   getPluginVersion
 };
-//# sourceMappingURL=chunk-2GJFTIWX.js.map
+//# sourceMappingURL=chunk-TK2ZYIAL.js.map

package/dist/{chunk-ZCBL5HER.js → chunk-XIIVIMFC.js}

@@ -1,7 +1,7 @@
 import { createRequire as __cr } from 'node:module'; const require = __cr(import.meta.url);
 import {
   STDIN_TIMEOUT_MS
-} from "./chunk-JBD5KP5G.js";
+} from "./chunk-TDLQBGKA.js";
 
 // src/hooks/read-stdin.ts
 function readStdin() {
@@ -18,4 +18,4 @@ function readStdin() {
 export {
   readStdin
 };
-//# sourceMappingURL=chunk-ZCBL5HER.js.map
+//# sourceMappingURL=chunk-XIIVIMFC.js.map