@koi-language/koi 1.0.5 → 1.1.0

package/README.md

@@ -2,9 +2,7 @@
   <img src="koi.png" alt="KOI Language Logo" width="400"/>
 </div>
 
-# Koi
-
-**Agent-first language. Calm orchestration.** 🌊
+# Koi - Agent-first language. Calm orchestration. 🌊
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![GitHub](https://img.shields.io/badge/GitHub-koi--language%2FKoi-6495ED?logo=github)](https://github.com/koi-language/Koi)
@@ -71,58 +69,6 @@ This will install KOI to `~/.koi` and add the `koi` command to your PATH.
 
 **For manual installation, development setup, and troubleshooting:** See the [Installation Guide](doc/00-installation.md).
 
-## Editor Support
-
-### VS Code
-
-```bash
-# From the Koi repository root
-ln -s "$(pwd)/vscode-koi-extension" ~/.vscode/extensions/koi-lang
-# Restart VS Code
-```
-
-### Cursor
-
-```bash
-# From the Koi repository root
-ln -s "$(pwd)/vscode-koi-extension" ~/.cursor/extensions/koi-lang
-# Restart Cursor
-```
-
-**Features:**
-- ✨ Full syntax highlighting
-- 🎨 Custom "Koi Dark" theme
-- 🔤 Auto-closing brackets and quotes
-- 📝 Special playbook highlighting
-- 🔍 Go to Definition support
-
-**For more installation methods and troubleshooting:** See the [Editor Setup Guide](doc/00-editor-setup.md).
-
-## GitHub Syntax Highlighting
-
-KOI code blocks in markdown files (README, Issues, PRs) automatically get syntax highlighting on GitHub!
-
-````markdown
-```koi
-Agent Hello : Worker {
-  on greet(args: Json) {
-    console.log("Hello from KOI!")
-  }
-}
-```
-````
-
-**Status:**
-- ✅ **Current**: Using JavaScript highlighting (very similar syntax)
-- 🚀 **Soon**: Native KOI highlighting (PR submitted to [github/linguist](https://github.com/github/linguist))
-
-Once the PR is merged, GitHub will recognize `.koi` files with:
-- 🎨 Cornflower Blue (#6495ED) in language statistics
-- ✨ Full syntax highlighting for code blocks
-- 📊 Language detection in repositories
-
-See [LINGUIST.md](LINGUIST.md) for technical details and contribution status.
-
 ## Usage
 
 ### With Global Installation
@@ -188,8 +134,6 @@ Agent Analyzer : Worker {
 
     Look for patterns, trends, and anomalies.
     Provide actionable insights.
-
-    Return JSON: { insights: [...], summary: "...", confidence: 0-1 }
     """
   }
 }
@@ -258,7 +202,7 @@ Agent Orchestrator : Worker {
 }
 ```
 
-See [MCP_GUIDE.md](MCP_GUIDE.md) for complete documentation.
+See [MCP Integration Guide](doc/10-mcp-integration.md) for complete documentation.
 
 ### Define Skills
 
@@ -279,13 +223,6 @@ Skill SentimentAnalysis {
       - Determine overall tone (positive, neutral, negative)
       - Rate emotional intensity (0.0 to 1.0)
       - Identify key emotional indicators
-
-      Return JSON: {
-        "sentiment": "positive|neutral|negative",
-        "score": 0.0-1.0,
-        "keywords": ["word1", "word2", ...],
-        "rationale": "brief explanation"
-      }
       """
     }
   }
@@ -351,23 +288,7 @@ Example: "Translate to French and count words"
 - Action 1: Translate → `{ translated: "bonjour monde" }`
 - Action 2: Count words on `${previousResult.translated}` → `{ wordCount: 2 }`
 
-See [TASK_CHAINING_GUIDE.md](TASK_CHAINING_GUIDE.md) for details.
-
-### 🆕 Automatic Routing
-
-Shows how ANY agent can automatically decompose and delegate tasks:
-
-```bash
-koi run examples/auto-routing-demo.koi
-```
-
-Key concepts:
-- Regular agents receiving complex tasks
-- Automatic decomposition by LLM
-- Cascading resolution (own handlers → skills → team routing)
-- Intelligent semantic matching via embeddings
-
-See [AUTO_ROUTING_GUIDE.md](AUTO_ROUTING_GUIDE.md) for explanation.
+See [Task Chaining Guide](doc/08-task-chaining.md) for details.
 
 ### Simple
 
@@ -417,23 +338,7 @@ Example showing MCP (Model Context Protocol) address support:
 koi run examples/mcp-example.koi
 ```
 
-### Automatic Planning
-
-Demonstrates automatic planning system where agents can decompose complex goals:
-
-```bash
-koi run examples/planning-demo.koi
-```
-
-### Planning with Actions
-
-Shows how LLM can plan and generate executable actions:
-
-```bash
-koi run examples/planning-with-actions.koi
-```
-
-See [PLANNING_GUIDE.md](PLANNING_GUIDE.md) for detailed planning system documentation.
+See [MCP Integration Guide](doc/10-mcp-integration.md) and [Planning System Guide](doc/09-planning.md) for more details on advanced features.
 
 ## Source Maps
 
@@ -491,32 +396,6 @@ Comprehensive documentation is available in the [doc/](doc/) directory:
 - **[Advanced Topics](doc/15-advanced.md)** - Debugging, performance, production
 - **[Registry](doc/16-registry.md)** - Shared data store for agent coordination
 
-## Roadmap
-
-- [X] MCP (Model Context Protocol) addresses support
-- [X] Real LLM integration with OpenAI/Anthropic
-- [X] **Full MCP protocol implementation** (WebSocket/HTTP2)
-  - [X] Real protocol support (WebSocket & HTTP/2)
-  - [X] Authentication and authorization
-  - [X] Server discovery
-  - [X] Connection pooling and load balancing
-  - [X] Retry logic and failover
-  - [X] Streaming responses
-  - [X] MCP tools integration
-- [X] Automatic planning system for agents
-  - [X] LLM-based goal decomposition
-  - [X] Sequential step execution
-  - [X] Automatic re-planning on failure
-  - [X] Context passing between steps
-  - [X] Execution tracking and summaries
-- [X] TypeScript/JavaScript import support
-- [X] Unit testing with Jest
-- [X] VSCode/Cursor extension with syntax highlighting
-- [X] GitHub syntax highlighting support
-- [ ] Skills registry and marketplace
-- [ ] Visual debugging
-- [ ] Hot reload in development
-
 ## Contributing
 
 Contributions are welcome! Here are some ways you can help:

package/examples/.build/agent-dialogue.ts

@@ -0,0 +1,138 @@
+// Generated from agent-dialogue.koi
+// Using local runtime from KOI_RUNTIME_PATH: /Users/antonioparraga/Git/Koi-lang/Koi/src/runtime
+import { Agent, Team, Skill, Role, Runtime, SkillRegistry, skillSelector, registry, mcpRegistry } from 'file:///Users/antonioparraga/Git/Koi-lang/Koi/src/runtime/index.js';
+import { createRequire } from 'module';
+const require = createRequire(import.meta.url);
+
+globalThis.SkillRegistry = SkillRegistry;
+globalThis.skillSelector = skillSelector;
+globalThis.registry = registry;
+globalThis.mcpRegistry = mcpRegistry;
+
+// ============================================================
+// Pre-computed Affordances (Build-time Cache)
+// Generated at: 2026-02-16T18:20:41.644Z
+// Total agents: 3
+// Total agent affordances: 3
+// Total skills: 0
+// Total skill affordances: 0
+// This avoids embedding API calls at runtime
+// ============================================================
+
+const CACHED_AFFORDANCES = {
+  "Philosopher": {
+    "ask": {
+      "description": "Asks one brief deep philosophical question and returns it as a JSON question field.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  },
+  "Poet": {
+    "respond": {
+      "description": "Gives a metaphorical poetic answer in up to two sentences and returns it as JSON.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  },
+  "DialogueCoordinator": {
+    "startDialogue": {
+      "description": "Runs three philosopher-question and poet-answer rounds, prints the dialogue, then ends.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  }
+};
+
+const CACHED_SKILL_AFFORDANCES = {};
+
+const Worker = new Role('Worker', ['execute']);
+
+const Coordinator = new Role('Coordinator', ['execute', 'delegate']);
+
+const Philosopher = new Agent({
+  name: 'Philosopher',
+  role: Worker,
+  llm: { provider: "openai", model: "gpt-5.2" },
+  handlers: {
+    ask: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "You are a philosopher who asks deep questions.\n\n    Context: ${args.context}\n\n    Generate ONE brief philosophical question (maximum 2 sentences).\n\n    Return ONLY: { \"question\": \"your question here\" }";
+      handler.__description__ = "Asks one brief deep philosophical question and returns it as a JSON question field.";
+      return handler;
+    })(),
+  }
+});
+
+const Poet = new Agent({
+  name: 'Poet',
+  role: Worker,
+  llm: { provider: "openai", model: "gpt-5.2" },
+  handlers: {
+    respond: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "You are a poet who responds in a metaphorical way.\n\n    Question: \"${args.question}\"\n\n    Respond poetically (maximum 2 sentences).\n\n    Return ONLY: { \"answer\": \"your answer here\" }";
+      handler.__description__ = "Gives a metaphorical poetic answer in up to two sentences and returns it as JSON.";
+      return handler;
+    })(),
+  }
+});
+
+const TestTeam = new Team('TestTeam', {
+  philosopher: Philosopher,
+  poet: Poet,
+});
+
+const DialogueCoordinator = new Agent({
+  name: 'DialogueCoordinator',
+  role: Coordinator,
+  usesTeams: [TestTeam],
+  peers: TestTeam, // Auto-assigned from uses Team
+  llm: { provider: "openai", model: "gpt-5.2" },
+  handlers: {
+    startDialogue: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "Coordinate a dialogue between Philosopher and Poet for 3 turns.\n\n    In each turn:\n    First have the philosopher ask a question and then have the poet respond to that question.\n    Print the dialogue like this:\n    - 🤔 Philosopher: <question here>\n    - 🎭 Poet: <answer here>\n\n    After 3 rounds, finish by showing:\n    ✨ Dialogue completed!";
+      handler.__description__ = "Runs three philosopher-question and poet-answer rounds, prints the dialogue, then ends.";
+      return handler;
+    })(),
+  }
+});
+
+
+// Main execution function
+(async () => {
+  // Register agents with router
+  const { agentRouter } = await import('file:///Users/antonioparraga/Git/Koi-lang/Koi/src/runtime/router.js');
+
+  await agentRouter.register(Philosopher, CACHED_AFFORDANCES['Philosopher']);
+  await agentRouter.register(Poet, CACHED_AFFORDANCES['Poet']);
+  await agentRouter.register(DialogueCoordinator, CACHED_AFFORDANCES['DialogueCoordinator']);
+
+  // Execute
+  const result = await DialogueCoordinator.handle('startDialogue', {});
+  // Result handled by actions
+
+  const _gracefulShutdown = async () => { await mcpRegistry.disconnectAll(); process.exit(0); };
+  process.on('SIGINT', _gracefulShutdown);
+  process.on('SIGTERM', _gracefulShutdown);
+
+  await mcpRegistry.disconnectAll();
+  process.exit(0);
+})().catch(err => {
+  console.error('Error:', err.message);
+  process.exit(1);
+});
+
+//# sourceMappingURL=agent-dialogue.ts.map

package/examples/.build/agent-dialogue.ts.map

@@ -0,0 +1 @@
+{"version":3,"sources":["agent-dialogue.koi"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;;AACA;;AAQA;;;;;;;;;;;;;;;;;;AAiBA;;;;;;;;;;;;;;;;;;AAvBA;;;;;AAwCA","file":"agent-dialogue.koi.js"}

package/examples/.build/chess.ts

@@ -0,0 +1,77 @@
+// Generated from chess.koi
+// Using local runtime from KOI_RUNTIME_PATH: /Users/antonioparraga/Git/Koi-lang/Koi/src/runtime
+import { Agent, Team, Skill, Role, Runtime, SkillRegistry, skillSelector, registry, mcpRegistry } from 'file:///Users/antonioparraga/Git/Koi-lang/Koi/src/runtime/index.js';
+import { createRequire } from 'module';
+const require = createRequire(import.meta.url);
+
+globalThis.SkillRegistry = SkillRegistry;
+globalThis.skillSelector = skillSelector;
+globalThis.registry = registry;
+globalThis.mcpRegistry = mcpRegistry;
+
+// ============================================================
+// Pre-computed Affordances (Build-time Cache)
+// Generated at: 2026-02-16T18:21:54.332Z
+// Total agents: 1
+// Total agent affordances: 1
+// Total skills: 0
+// Total skill affordances: 0
+// This avoids embedding API calls at runtime
+// ============================================================
+
+const CACHED_AFFORDANCES = {
+  "Assistant": {
+    "test_while": {
+      "description": "Play chess turn-by-turn, showing the board, processing user moves, and ending on win/stop.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  }
+};
+
+const CACHED_SKILL_AFFORDANCES = {};
+
+const Worker = new Role('Worker', ['execute']);
+
+const Assistant = new Agent({
+  name: 'Assistant',
+  role: Worker,
+  llm: { provider: "openai", model: "gpt-5.2" },
+  handlers: {
+    test_while: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "Play chess with the user.\n    On each turn you show the board, and the user says where to move with a combination like A2 to A4, B3 to C3\n    Then you make your move and show the board as it looks after your move.\n    End the game when someone wins or if the user expresses their wish to stop.";
+      handler.__description__ = "Play chess turn-by-turn, showing the board, processing user moves, and ending on win/stop.";
+      return handler;
+    })(),
+  }
+});
+
+
+// Main execution function
+(async () => {
+  // Register agents with router
+  const { agentRouter } = await import('file:///Users/antonioparraga/Git/Koi-lang/Koi/src/runtime/router.js');
+
+  await agentRouter.register(Assistant, CACHED_AFFORDANCES['Assistant']);
+
+  // Execute
+  const result = await Assistant.handle('test_while', {});
+  // Result handled by actions
+
+  const _gracefulShutdown = async () => { await mcpRegistry.disconnectAll(); process.exit(0); };
+  process.on('SIGINT', _gracefulShutdown);
+  process.on('SIGTERM', _gracefulShutdown);
+
+  await mcpRegistry.disconnectAll();
+  process.exit(0);
+})().catch(err => {
+  console.error('Error:', err.message);
+  process.exit(1);
+});
+
+//# sourceMappingURL=chess.ts.map

package/examples/.build/chess.ts.map

@@ -0,0 +1 @@
+{"version":3,"sources":["chess.koi"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAGA;;AAEA","file":"chess.koi.js"}

package/examples/.build/delegation-test.ts

@@ -0,0 +1,140 @@
+// Generated from delegation-test.koi
+// Using local runtime from KOI_RUNTIME_PATH: /Users/antonioparraga/Git/M/src/runtime
+import { Agent, Team, Skill, Role, Runtime, SkillRegistry, skillSelector, registry, mcpRegistry } from 'file:///Users/antonioparraga/Git/M/src/runtime/index.js';
+import { createRequire } from 'module';
+const require = createRequire(import.meta.url);
+
+globalThis.SkillRegistry = SkillRegistry;
+globalThis.skillSelector = skillSelector;
+globalThis.registry = registry;
+globalThis.mcpRegistry = mcpRegistry;
+
+// ============================================================
+// Pre-computed Affordances (Build-time Cache)
+// Generated at: 2026-02-16T17:29:50.172Z
+// Total agents: 3
+// Total agent affordances: 3
+// Total skills: 0
+// Total skill affordances: 0
+// This avoids embedding API calls at runtime
+// ============================================================
+
+const CACHED_AFFORDANCES = {
+  "Translator": {
+    "translate": {
+      "description": "Translates text and returns the translated version in JSON format.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  },
+  "WordCounter": {
+    "count": {
+      "description": "Counts the number of words in the provided text and returns the count in JSON.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  },
+  "Orchestrator": {
+    "processText": {
+      "description": "Prompts for a sentence, translates it to Spanish, counts words, and displays results.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  }
+};
+
+const CACHED_SKILL_AFFORDANCES = {};
+
+const Worker = new Role('Worker', ['execute']);
+
+const Coordinator = new Role('Coordinator', ['execute', 'delegate']);
+
+const Translator = new Agent({
+  name: 'Translator',
+  role: Worker,
+  llm: { provider: "openai", model: "gpt-4o-mini" },
+  amnesia: true,
+  handlers: {
+    translate: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "Translate \"${args.text}\" to ${args.language}.\n    Return JSON: { \"translated\": \"translated text here\" }";
+      handler.__description__ = "Translates text and returns the translated version in JSON format.";
+      return handler;
+    })(),
+  }
+});
+
+const WordCounter = new Agent({
+  name: 'WordCounter',
+  role: Worker,
+  llm: { provider: "openai", model: "gpt-4o-mini" },
+  amnesia: true,
+  handlers: {
+    count: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "Count words in: \"${args.text}\"\n    Return JSON: { \"wordCount\": number }";
+      handler.__description__ = "Counts the number of words in the provided text and returns the count in JSON.";
+      return handler;
+    })(),
+  }
+});
+
+const TestTeam = new Team('TestTeam', {
+  translator: Translator,
+  wordCounter: WordCounter,
+});
+
+const Orchestrator = new Agent({
+  name: 'Orchestrator',
+  role: Coordinator,
+  usesTeams: [TestTeam],
+  peers: TestTeam, // Auto-assigned from uses Team
+  llm: { provider: "openai", model: "gpt-4o-mini" },
+  handlers: {
+    processText: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "Prompt the user to enter a sentence.\n    Translate the sentence to Spanish.\n    Count the words in the translation.\n    Print the original sentence, the translation, and the word count.";
+      handler.__description__ = "Prompts for a sentence, translates it to Spanish, counts words, and displays results.";
+      return handler;
+    })(),
+  }
+});
+
+
+// Main execution function
+(async () => {
+  // Register agents with router
+  const { agentRouter } = await import('file:///Users/antonioparraga/Git/M/src/runtime/router.js');
+
+  await agentRouter.register(Translator, CACHED_AFFORDANCES['Translator']);
+  await agentRouter.register(WordCounter, CACHED_AFFORDANCES['WordCounter']);
+  await agentRouter.register(Orchestrator, CACHED_AFFORDANCES['Orchestrator']);
+
+  // Execute
+  const result = await Orchestrator.handle('processText', { text: "hello world" });
+  // Result handled by actions
+
+  const _gracefulShutdown = async () => { await mcpRegistry.disconnectAll(); process.exit(0); };
+  process.on('SIGINT', _gracefulShutdown);
+  process.on('SIGTERM', _gracefulShutdown);
+
+  await mcpRegistry.disconnectAll();
+  process.exit(0);
+})().catch(err => {
+  console.error('Error:', err.message);
+  process.exit(1);
+});
+
+//# sourceMappingURL=delegation-test.ts.map

package/examples/.build/delegation-test.ts.map

@@ -0,0 +1 @@
+{"version":3,"sources":["delegation-test.koi"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;;AACA;;AAQA;;;;;;;;;;;;;;;;;;;AAaA;;;;;;;;;;;;;;;;;;;AAnBA;;;;;AAgCA","file":"delegation-test.koi.js"}

package/examples/.build/dialog-demo.ts

@@ -0,0 +1,77 @@
+// Generated from dialog-demo.koi
+// Using local runtime from KOI_RUNTIME_PATH: /Users/antonioparraga/Git/Koi-lang/Koi/src/runtime
+import { Agent, Team, Skill, Role, Runtime, SkillRegistry, skillSelector, registry, mcpRegistry } from 'file:///Users/antonioparraga/Git/Koi-lang/Koi/src/runtime/index.js';
+import { createRequire } from 'module';
+const require = createRequire(import.meta.url);
+
+globalThis.SkillRegistry = SkillRegistry;
+globalThis.skillSelector = skillSelector;
+globalThis.registry = registry;
+globalThis.mcpRegistry = mcpRegistry;
+
+// ============================================================
+// Pre-computed Affordances (Build-time Cache)
+// Generated at: 2026-02-16T18:35:27.515Z
+// Total agents: 1
+// Total agent affordances: 1
+// Total skills: 0
+// Total skill affordances: 0
+// This avoids embedding API calls at runtime
+// ============================================================
+
+const CACHED_AFFORDANCES = {
+  "Assistant": {
+    "test_while": {
+      "description": "Introduce yourself as Koi, ask their name, chat freely until they say goodbye, then respond cordially.",
+      "confidence": 0.9,
+      "hasPlaybook": true
+    }
+  }
+};
+
+const CACHED_SKILL_AFFORDANCES = {};
+
+const Worker = new Role('Worker', ['execute']);
+
+const Assistant = new Agent({
+  name: 'Assistant',
+  role: Worker,
+  llm: { provider: "openai", model: "gpt-5.2" },
+  handlers: {
+    test_while: (() => {
+      const handler = async function(args) {
+        // This should not be called - playbook will be executed by LLM
+        throw new Error('Playbook-only handler called directly');
+      };
+      handler.__playbookOnly__ = true;
+      handler.__playbook__ = "Talk with the user until they say goodbye to you.\n    Start the conversation by bringing up a topic of general interest, about politics, sports, science, whatever you prefer.\n    Talk with the user until they express their intention to end the conversation, in which case say goodbye cordially and end the conversation.\n    When you ask the user something, let them express themselves. Do not give them a set of options, instead let them express themselves freely.\n    SUPER IMPORTANT, The first thing you must do is introduce yourself. Your name is Koi, so ask for their name and greet them cordially. Then start the conversation.";
+      handler.__description__ = "Introduce yourself as Koi, ask their name, chat freely until they say goodbye, then respond cordially.";
+      return handler;
+    })(),
+  }
+});
+
+
+// Main execution function
+(async () => {
+  // Register agents with router
+  const { agentRouter } = await import('file:///Users/antonioparraga/Git/Koi-lang/Koi/src/runtime/router.js');
+
+  await agentRouter.register(Assistant, CACHED_AFFORDANCES['Assistant']);
+
+  // Execute
+  const result = await Assistant.handle('test_while', {});
+  // Result handled by actions
+
+  const _gracefulShutdown = async () => { await mcpRegistry.disconnectAll(); process.exit(0); };
+  process.on('SIGINT', _gracefulShutdown);
+  process.on('SIGTERM', _gracefulShutdown);
+
+  await mcpRegistry.disconnectAll();
+  process.exit(0);
+})().catch(err => {
+  console.error('Error:', err.message);
+  process.exit(1);
+});
+
+//# sourceMappingURL=dialog-demo.ts.map

package/examples/.build/dialog-demo.ts.map

@@ -0,0 +1 @@
+{"version":3,"sources":["dialog-demo.koi"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAGA;;AAEA","file":"dialog-demo.koi.js"}