panopticon-cli 0.4.33 → 0.5.1

package/dist/{chunk-KJ2TRXNK.js → chunk-FTCPTHIJ.js}

@@ -1,7 +1,10 @@
 import {
+  MODEL_CAPABILITIES,
+  getModelCapability,
   init_config_yaml,
+  init_model_capabilities,
   loadConfig
-} from "./chunk-7XNJJBH6.js";
+} from "./chunk-HJSM6E6U.js";
 import {
   AGENTS_DIR,
   PANOPTICON_HOME,
@@ -134,6 +137,43 @@ function capturePane(sessionName, lines = 50) {
     return "";
   }
 }
+async function capturePaneAsync(sessionName, lines = 50) {
+  try {
+    const { stdout } = await execAsync(`tmux capture-pane -t ${sessionName} -p -S -${lines}`, {
+      encoding: "utf-8"
+    });
+    return stdout;
+  } catch {
+    return "";
+  }
+}
+async function waitForClaudePrompt(sessionName, timeoutMs = 15e3) {
+  const start = Date.now();
+  const POLL = 500;
+  while (Date.now() - start < timeoutMs) {
+    const output = await capturePaneAsync(sessionName, 10);
+    const lines = output.split("\n").filter((l) => l.trim());
+    const lastLine = lines[lines.length - 1] || "";
+    if (lastLine.includes("\u276F")) return true;
+    await new Promise((r) => setTimeout(r, POLL));
+  }
+  return false;
+}
+async function confirmDelivery(sessionName, outputBefore, timeoutMs = 1e4) {
+  const start = Date.now();
+  const POLL = 1e3;
+  const beforeLineCount = outputBefore.split("\n").filter((l) => l.trim()).length;
+  while (Date.now() - start < timeoutMs) {
+    await new Promise((r) => setTimeout(r, POLL));
+    const after = await capturePaneAsync(sessionName, 50);
+    const afterLines = after.split("\n").filter((l) => l.trim());
+    const afterLineCount = afterLines.length;
+    if (afterLineCount > beforeLineCount + 1) return true;
+    const newOutput = afterLines.slice(beforeLineCount).join("\n");
+    if (newOutput.includes("\u25CF") || newOutput.includes("\u23BF") || newOutput.includes("Read")) return true;
+  }
+  return false;
+}
 function getAgentSessions() {
   return listSessions().filter((s) => s.name.startsWith("agent-"));
 }
@@ -518,422 +558,6 @@ var init_model_fallback = __esm({
   }
 });
 
-// src/lib/model-capabilities.ts
-function getModelCapability(model) {
-  return MODEL_CAPABILITIES[model];
-}
-var MODEL_CAPABILITIES;
-var init_model_capabilities = __esm({
-  "src/lib/model-capabilities.ts"() {
-    "use strict";
-    init_esm_shims();
-    MODEL_CAPABILITIES = {
-      // ═══════════════════════════════════════════════════════════════════════════
-      // ANTHROPIC MODELS
-      // ═══════════════════════════════════════════════════════════════════════════
-      "claude-opus-4-6": {
-        model: "claude-opus-4-6",
-        provider: "anthropic",
-        displayName: "Claude Opus 4.6",
-        costPer1MTokens: 45,
-        // $5 in / $25 out → same pricing as 4.5
-        contextWindow: 2e5,
-        // 1M available via opt-in beta, but we use 200K
-        skills: {
-          "code-generation": 96,
-          // 80.9% SWE-bench (first >80%), 89.4% Aider Polyglot
-          "code-review": 98,
-          debugging: 97,
-          planning: 99,
-          // User confirms: "Opus 4.6 planning for sure"
-          documentation: 95,
-          testing: 92,
-          security: 98,
-          // Best for security review
-          performance: 90,
-          synthesis: 98,
-          // Best for combining info across domains
-          speed: 40,
-          // Slower but 76% more token efficient
-          "context-length": 95
-        },
-        notes: "Successor to Opus 4.5. Same pricing, 1M context available (opt-in beta). Best for planning, security, complex reasoning."
-      },
-      "claude-sonnet-4-6": {
-        model: "claude-sonnet-4-6",
-        provider: "anthropic",
-        displayName: "Claude Sonnet 4.6",
-        costPer1MTokens: 9,
-        // $3 in / $15 out → avg ~$9
-        contextWindow: 2e5,
-        skills: {
-          "code-generation": 94,
-          "code-review": 94,
-          debugging: 92,
-          planning: 90,
-          documentation: 92,
-          testing: 92,
-          security: 88,
-          performance: 88,
-          synthesis: 90,
-          speed: 70,
-          "context-length": 95
-        },
-        notes: "Successor to Sonnet 4.5. Same pricing tier. Improved coding and reasoning."
-      },
-      "claude-sonnet-4-5": {
-        model: "claude-sonnet-4-5",
-        provider: "anthropic",
-        displayName: "Claude Sonnet 4.5",
-        costPer1MTokens: 9,
-        // $3 in / $15 out → avg ~$9
-        contextWindow: 2e5,
-        skills: {
-          "code-generation": 92,
-          // 77.2% SWE-bench (82% parallel), beats GPT-5 Codex (74.5%)
-          "code-review": 92,
-          debugging: 90,
-          planning: 88,
-          documentation: 90,
-          // 100% AIME with Python
-          testing: 90,
-          // 50% Terminal-Bench, 61.4% OSWorld
-          security: 85,
-          performance: 85,
-          synthesis: 88,
-          speed: 70,
-          "context-length": 95
-        },
-        notes: "Best value: 77.2% SWE-bench at 1/5th Opus cost. Beats GPT-5 Codex."
-      },
-      "claude-haiku-4-5": {
-        model: "claude-haiku-4-5",
-        provider: "anthropic",
-        displayName: "Claude Haiku 4.5",
-        costPer1MTokens: 4,
-        // $0.80 in / $4 out → avg ~$2.4
-        contextWindow: 2e5,
-        skills: {
-          "code-generation": 75,
-          "code-review": 72,
-          debugging: 70,
-          planning: 65,
-          documentation: 75,
-          testing: 70,
-          security: 60,
-          performance: 65,
-          synthesis: 68,
-          speed: 95,
-          // Fastest Anthropic
-          "context-length": 95
-        },
-        notes: "Fast and cheap, good for simple tasks and exploration"
-      },
-      // ═══════════════════════════════════════════════════════════════════════════
-      // OPENAI MODELS
-      // ═══════════════════════════════════════════════════════════════════════════
-      "gpt-5.2-codex": {
-        model: "gpt-5.2-codex",
-        provider: "openai",
-        displayName: "GPT-5.2 Codex",
-        costPer1MTokens: 75,
-        // Premium tier ~$75/M
-        contextWindow: 128e3,
-        skills: {
-          "code-generation": 95,
-          // 80% SWE-bench Verified, 55.6% SWE-bench Pro
-          "code-review": 90,
-          debugging: 92,
-          // 92.4% GPQA Diamond
-          planning: 88,
-          documentation: 85,
-          testing: 90,
-          security: 85,
-          performance: 88,
-          // 52.9% ARC-AGI-2 (best reasoning)
-          synthesis: 88,
-          // 100% AIME 2025 without tools
-          speed: 55,
-          "context-length": 75
-        },
-        notes: "Premium coding: 80% SWE-bench. Best raw reasoning (52.9% ARC-AGI-2). Expensive."
-      },
-      "o3-deep-research": {
-        model: "o3-deep-research",
-        provider: "openai",
-        displayName: "O3 Deep Research",
-        costPer1MTokens: 100,
-        // Expensive reasoning model
-        contextWindow: 2e5,
-        skills: {
-          "code-generation": 85,
-          "code-review": 95,
-          debugging: 98,
-          // Best for debugging
-          planning: 95,
-          documentation: 88,
-          testing: 85,
-          security: 92,
-          performance: 92,
-          synthesis: 95,
-          speed: 20,
-          // Very slow (reasoning chains)
-          "context-length": 95
-        },
-        notes: "Deep reasoning model, excellent for complex debugging and analysis"
-      },
-      "gpt-4o": {
-        model: "gpt-4o",
-        provider: "openai",
-        displayName: "GPT-4o",
-        costPer1MTokens: 15,
-        // $5 in / $15 out
-        contextWindow: 128e3,
-        skills: {
-          "code-generation": 88,
-          "code-review": 85,
-          debugging: 85,
-          planning: 82,
-          documentation: 88,
-          testing: 82,
-          security: 78,
-          performance: 80,
-          synthesis: 85,
-          speed: 75,
-          "context-length": 75
-        },
-        notes: "Good all-rounder, competitive with Sonnet"
-      },
-      "gpt-4o-mini": {
-        model: "gpt-4o-mini",
-        provider: "openai",
-        displayName: "GPT-4o Mini",
-        costPer1MTokens: 1,
-        // Very cheap
-        contextWindow: 128e3,
-        skills: {
-          "code-generation": 72,
-          "code-review": 68,
-          debugging: 65,
-          planning: 60,
-          documentation: 70,
-          testing: 65,
-          security: 55,
-          performance: 60,
-          synthesis: 62,
-          speed: 92,
-          "context-length": 75
-        },
-        notes: "Budget option, good for simple tasks"
-      },
-      // ═══════════════════════════════════════════════════════════════════════════
-      // GOOGLE MODELS
-      // ═══════════════════════════════════════════════════════════════════════════
-      "gemini-3-pro-preview": {
-        model: "gemini-3-pro-preview",
-        provider: "google",
-        displayName: "Gemini 3 Pro",
-        costPer1MTokens: 12,
-        // $4.2 in / $18.9 out
-        contextWindow: 1e6,
-        // 1M context!
-        skills: {
-          "code-generation": 90,
-          // 2439 Elo LiveCodeBench Pro (first >1500 on LMArena)
-          "code-review": 88,
-          debugging: 85,
-          planning: 85,
-          documentation: 88,
-          testing: 85,
-          // ~95% AIME 2025
-          security: 78,
-          performance: 85,
-          // Strong multimodal
-          synthesis: 90,
-          // Best for combining large codebases
-          speed: 80,
-          "context-length": 100
-          // Best context - 1M tokens
-        },
-        notes: "First to exceed 1500 Elo on LMArena. Best for large codebase analysis with 1M context."
-      },
-      "gemini-3-flash-preview": {
-        model: "gemini-3-flash-preview",
-        provider: "google",
-        displayName: "Gemini 3 Flash",
-        costPer1MTokens: 0.5,
-        // Very cheap
-        contextWindow: 1e6,
-        skills: {
-          "code-generation": 75,
-          "code-review": 70,
-          debugging: 68,
-          planning: 62,
-          documentation: 72,
-          testing: 68,
-          security: 55,
-          performance: 65,
-          synthesis: 70,
-          speed: 98,
-          // Fastest overall
-          "context-length": 100
-        },
-        notes: "Extremely fast and cheap, huge context, great for exploration"
-      },
-      "gemini-2.5-pro": {
-        model: "gemini-2.5-pro",
-        provider: "google",
-        displayName: "Gemini 2.5 Pro",
-        costPer1MTokens: 12,
-        contextWindow: 1e6,
-        skills: {
-          "code-generation": 92,
-          "code-review": 90,
-          debugging: 88,
-          planning: 88,
-          documentation: 90,
-          testing: 87,
-          security: 82,
-          performance: 88,
-          synthesis: 92,
-          speed: 75,
-          "context-length": 100
-        },
-        notes: "Advanced reasoning and code capabilities with 1M context"
-      },
-      "gemini-2.5-flash": {
-        model: "gemini-2.5-flash",
-        provider: "google",
-        displayName: "Gemini 2.5 Flash",
-        costPer1MTokens: 0.6,
-        contextWindow: 1e6,
-        skills: {
-          "code-generation": 78,
-          "code-review": 73,
-          debugging: 70,
-          planning: 65,
-          documentation: 75,
-          testing: 70,
-          security: 58,
-          performance: 68,
-          synthesis: 73,
-          speed: 95,
-          "context-length": 100
-        },
-        notes: "Fast and efficient with large context support"
-      },
-      // ═══════════════════════════════════════════════════════════════════════════
-      // Z.AI MODELS
-      // ═══════════════════════════════════════════════════════════════════════════
-      "glm-4.7": {
-        model: "glm-4.7",
-        provider: "zai",
-        displayName: "GLM 4.7",
-        costPer1MTokens: 5,
-        contextWindow: 2e5,
-        // 200K context, 128K output
-        skills: {
-          "code-generation": 88,
-          // 73.8% SWE-bench, 84.9 LiveCodeBench v6 (open-source SOTA)
-          "code-review": 85,
-          debugging: 85,
-          // Strong debugging with Interleaved Thinking
-          planning: 82,
-          // 95.7% AIME 2025 (beats Gemini 3 & GPT-5.1)
-          documentation: 80,
-          testing: 82,
-          // 87.4 τ²-Bench (SOTA for tool use)
-          security: 72,
-          performance: 78,
-          synthesis: 85,
-          // Preserved Thinking retains context across turns
-          speed: 80,
-          "context-length": 95
-          // 200K context
-        },
-        notes: "Top open-source for agentic coding. 73.8% SWE-bench, best tool use. 400B params with Interleaved Thinking."
-      },
-      "glm-4.7-flash": {
-        model: "glm-4.7-flash",
-        provider: "zai",
-        displayName: "GLM 4.7 Flash",
-        costPer1MTokens: 1.5,
-        contextWindow: 128e3,
-        skills: {
-          "code-generation": 72,
-          "code-review": 68,
-          debugging: 65,
-          planning: 62,
-          documentation: 70,
-          testing: 65,
-          security: 55,
-          performance: 62,
-          synthesis: 65,
-          speed: 92,
-          // Fast inference
-          "context-length": 75
-        },
-        notes: "Fast and affordable. Good for quick iterations and exploration."
-      },
-      // ═══════════════════════════════════════════════════════════════════════════
-      // KIMI MODELS
-      // ═══════════════════════════════════════════════════════════════════════════
-      "kimi-k2": {
-        model: "kimi-k2",
-        provider: "kimi",
-        displayName: "Kimi K2",
-        costPer1MTokens: 1.4,
-        // $0.16 in / $2.63 out → very cheap
-        contextWindow: 131e3,
-        skills: {
-          "code-generation": 82,
-          // 65.8% SWE-bench (beats GPT-4.1 at 54.6%)
-          "code-review": 80,
-          debugging: 78,
-          planning: 75,
-          documentation: 80,
-          testing: 75,
-          security: 70,
-          performance: 72,
-          synthesis: 78,
-          speed: 80,
-          "context-length": 75
-        },
-        notes: "Strong value: 65.8% SWE-bench at very low cost. Good for routine tasks."
-      },
-      "kimi-k2.5": {
-        model: "kimi-k2.5",
-        provider: "kimi",
-        displayName: "Kimi K2.5",
-        costPer1MTokens: 8,
-        // ~5.1x cheaper than GPT-5.2
-        contextWindow: 256e3,
-        skills: {
-          "code-generation": 92,
-          // 76.8% SWE-bench, 85 LiveCodeBench v6
-          "code-review": 90,
-          debugging: 90,
-          // Strong analytical capabilities
-          planning: 88,
-          // User confirms "highly capable"
-          documentation: 88,
-          testing: 88,
-          // 92% coding accuracy
-          security: 82,
-          performance: 85,
-          synthesis: 92,
-          // Can coordinate 100 sub-agents, 1500 tool calls
-          speed: 75,
-          // MoE: 1T total params, 32B active
-          "context-length": 98
-          // 256K context
-        },
-        notes: "Best open-source coding model. 5x cheaper than GPT-5.2. Excellent for frontend dev and multi-agent orchestration."
-      }
-    };
-  }
-});
-
 // src/lib/smart-model-selector.ts
 function calculateSkillScore(model, requirements) {
   const cap = getModelCapability(model);
@@ -1237,7 +861,7 @@ var init_work_type_router = __esm({
       config;
       availableModels = null;
       constructor(config) {
-        this.config = config || loadConfig();
+        this.config = config || loadConfig().config;
       }
       /**
        * Get list of available models based on enabled providers
@@ -1334,7 +958,7 @@ var init_work_type_router = __esm({
        * Reload configuration from disk
        */
       reloadConfig() {
-        this.config = loadConfig();
+        this.config = loadConfig().config;
         this.availableModels = null;
       }
       /**
@@ -1365,6 +989,9 @@ export {
   sendKeysAsync,
   sendKeys,
   capturePane,
+  capturePaneAsync,
+  waitForClaudePrompt,
+  confirmDelivery,
   getAgentSessions,
   init_tmux,
   initHook,
@@ -1378,4 +1005,4 @@ export {
   getModelId,
   init_work_type_router
 };
-//# sourceMappingURL=chunk-KJ2TRXNK.js.map
+//# sourceMappingURL=chunk-FTCPTHIJ.js.map