### Running the Server

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Provides the necessary commands to install dependencies and start the FAF server development environment.

```bash
cd examples/server-demo
npm install
npm run dev
```

--------------------------------

### Run RAG Demo Application

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Steps to set up and run the RAG demo application. This involves navigating to the demo directory, installing dependencies, copying the environment configuration, editing it, and starting the development server.

```bash
cd examples/rag-demo
npm install
cp .env.example .env
# Edit .env with your configuration
npm run dev
```

--------------------------------

### Install FAF and LiteLLM

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Installs the Functional Agent Framework (FAF) using npm and sets up the LiteLLM proxy server using pip. Includes commands for starting the LiteLLM server with different configurations and setting API keys.

```bash
npm install functional-agent-framework
```

```bash
pip install litellm
```

```bash
litellm --model gpt-3.5-turbo --port 4000
```

```bash
export OPENAI_API_KEY=your_key_here
litellm --model gpt-4o --port 4000
```

```bash
litellm --config config.yaml --port 4000
```

--------------------------------

### Start PostgreSQL Server

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/test-memory-providers.md

Commands to start a PostgreSQL server, either using Docker or local installation methods for macOS and Ubuntu, and create the necessary database.

```bash
# Using Docker (recommended)
docker run -d --name faf-postgres \
  -e POSTGRES_PASSWORD=testpass \
  -e POSTGRES_DB=faf_memory \
  -p 5432:5432 \
  postgres:15

# Or using local PostgreSQL
brew install postgresql && brew services start postgresql  # macOS
createdb faf_memory  # macOS

sudo apt install postgresql postgresql-contrib  # Ubuntu
sudo systemctl start postgresql  # Ubuntu
sudo -u postgres createdb faf_memory  # Ubuntu
```

--------------------------------

### Development Server Demo Setup

Source: https://github.com/xynehq/faf/blob/main/README.md

Provides instructions for setting up and running the development server demo application. This includes navigating to the example directory, installing dependencies, and starting the development server.

```bash
cd examples/server-demo
npm install
npm run dev
```

--------------------------------

### Start Redis Server

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/test-memory-providers.md

Commands to start a Redis server, either using Docker or local installation methods for macOS and Ubuntu.

```bash
# Using Docker (recommended)
docker run -d --name faf-redis -p 6379:6379 redis:alpine

# Or using local Redis
brew install redis && brew services start redis  # macOS
sudo apt install redis-server && sudo systemctl start redis-server  # Ubuntu
```

--------------------------------

### FAF Hello World Example

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

A TypeScript example demonstrating how to create a simple FAF agent with a greeting tool. It defines context, a tool with Zod schema validation, an agent with instructions, and sets up the execution environment with a LiteLLM provider.

```typescript
import {
  run,
  Tool,
  Agent,
  makeLiteLLMProvider,
  generateRunId,
  generateTraceId
} from 'functional-agent-framework';
import { z } from 'zod';

// 1. Define your context type
type MyContext = {
  userId: string;
  permissions: string[];
};

// 2. Create a simple greeting tool
const greetingTool: Tool<{ name: string }, MyContext> = {
  schema: {
    name: "greet",
    description: "Generate a personalized greeting",
    parameters: z.object({
      name: z.string().describe("Name of the person to greet")
    }),
  },
  execute: async (args, context) => {
    return `Hello, ${args.name}! I'm running on FAF. Your user ID is ${context.userId}.`;
  },
};

// 3. Define an agent
const assistantAgent: Agent<MyContext, string> = {
  name: 'Assistant',
  instructions: () => 'You are a helpful assistant. Use the greeting tool when meeting new people.',
  tools: [greetingTool],
};

// 4. Set up the execution environment
async function runHelloWorld() {
  // Configure model provider
  const modelProvider = makeLiteLLMProvider('http://localhost:4000');
  
  // Create agent registry
  const agentRegistry = new Map([['Assistant', assistantAgent]]);
  
  // Create initial state
  const initialState = {
    runId: generateRunId(),
    traceId: generateTraceId(),
    messages: [{ role: 'user' as const, content: 'Hi, my name is Alice' }],
    currentAgentName: 'Assistant',
    context: { userId: 'user123', permissions: ['user'] },
    turnCount: 0,
  };
  
  // Run the agent
  const result = await run(initialState, {
    agentRegistry,
    modelProvider,
    maxTurns: 5,
    onEvent: (event) => console.log('Event:', event.type),
  });
  
  if (result.outcome.status === 'completed') {
    console.log('Response:', result.outcome.output);
  } else {
    console.error('Error:', result.outcome.error);
  }
}

runHelloWorld().catch(console.error);

```

--------------------------------

### Development Server Setup

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Starts a development server for testing agents via HTTP, configuring it with agents, model provider, memory, and server-specific options like port and CORS.

```typescript
import { runServer, makeLiteLLMProvider, createInMemoryProvider } from 'functional-agent-framework';

async function startDevServer() {
  const modelProvider = makeLiteLLMProvider('http://localhost:4000');
  const memoryProvider = await createInMemoryProvider();
  
  const server = await runServer(
    [assistantAgent, calculatorAgent], // Array of agents
    {
      modelProvider,
      maxTurns: 5,
      onEvent: (event) => console.log(event.type),
      memory: {
        provider: memoryProvider,
        autoStore: true,
        maxMessages: 100
      }
    },
    {
      port: 3000,
      host: '127.0.0.1',
      cors: true
    }
  );
  
  console.log('Server running on http://localhost:3000');
}
```

--------------------------------

### Start LiteLLM Proxy

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/README.md

Instructions to install and start the LiteLLM proxy, a common requirement for using LLM services. This example uses OpenAI.

```bash
# Install LiteLLM
pip install litellm

# Start proxy (example with OpenAI)
litellm --model gpt-3.5-turbo --port 4000
```

--------------------------------

### Install and Start PostgreSQL (Ubuntu)

Source: https://github.com/xynehq/faf/blob/main/docs/memory-system.md

Installs PostgreSQL and related contrib packages, starts the service, and creates a database for FAF memory.

```bash
sudo apt install postgresql postgresql-contrib
sudo systemctl start postgresql
sudo -u postgres createdb faf_memory
```

--------------------------------

### PostgreSQL Memory Provider Setup

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Installs PostgreSQL dependencies and configures the functional-agent-framework to use PostgreSQL for conversation persistence by setting environment variables and providing a PostgreSQL client.

```bash
# Install PostgreSQL dependencies
npm install pg @types/pg
```

```typescript
// Set environment variables
process.env.FAF_MEMORY_TYPE = 'postgres';
process.env.FAF_POSTGRES_HOST = 'localhost';
process.env.FAF_POSTGRES_DB = 'faf_memory';

// Create PostgreSQL client
const { Client } = await import('pg');
const postgresClient = new Client({
  host: 'localhost',
  database: 'faf_memory',
  user: 'postgres',
  password: 'your_password'
});
await postgresClient.connect();

// Create memory provider
const memoryProvider = await createMemoryProviderFromEnv({ postgres: postgresClient });
```

--------------------------------

### Install and Start Redis (Ubuntu)

Source: https://github.com/xynehq/faf/blob/main/docs/memory-system.md

Installs Redis server using apt and starts the Redis service for local use.

```bash
sudo apt install redis-server && sudo systemctl start redis-server
```

--------------------------------

### Redis Memory Provider Setup

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Installs Redis dependencies and configures the functional-agent-framework to use Redis for conversation persistence by setting environment variables and providing a Redis client.

```bash
# Install Redis dependencies
npm install redis
```

```typescript
import { createMemoryProviderFromEnv } from 'functional-agent-framework';

// Set environment variables
process.env.FAF_MEMORY_TYPE = 'redis';
process.env.FAF_REDIS_HOST = 'localhost';
process.env.FAF_REDIS_PORT = '6379';

// Create Redis client
const { createClient } = await import('redis');
const redisClient = createClient({
  url: 'redis://localhost:6379'
});
await redisClient.connect();

// Create memory provider
const memoryProvider = await createMemoryProviderFromEnv({ redis: redisClient });
```

--------------------------------

### Setup and Run RAG Demo

Source: https://github.com/xynehq/faf/blob/main/README.md

Commands to navigate to the RAG demo directory, install dependencies using npm, and start the development server. This is the initial setup for running the demo.

```bash
cd examples/rag-demo
npm install
npm run dev
```

--------------------------------

### Server Setup with Memory Configuration

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Demonstrates setting up the FAF server using environment variables for memory provider selection and configuring server-side options like maxTurns and event handling.

```typescript
const memoryType = process.env.FAF_MEMORY_TYPE || 'memory';
const memoryProvider = await createMemoryProviderFromEnv(externalClients);

const server = await runServer(
  [mathAgent, chatAgent, assistantAgent],
  {
    modelProvider,
    maxTurns: 5,
    onEvent: traceCollector.collect.bind(traceCollector),
    memory: {
      provider: memoryProvider,
      autoStore: true, // Automatically store conversation history
      maxMessages: 100 // Keep last 100 messages per conversation
    }
  },
  {
    port: parseInt(process.env.PORT || '3000'),
    defaultMemoryProvider: memoryProvider
  }
);
```

--------------------------------

### Copy Environment Configuration

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/test-memory-providers.md

Copies the example environment file to be used for configuration. This file will be modified for different provider setups.

```bash
cp .env.example .env
```

--------------------------------

### Install and Start Redis (macOS)

Source: https://github.com/xynehq/faf/blob/main/docs/memory-system.md

Installs Redis using Homebrew and starts the Redis service for local use.

```bash
brew install redis && brew services start redis
```

--------------------------------

### Memory Provider Comparison

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Compares different memory providers (In-Memory, Redis, PostgreSQL) based on their best use cases, persistence, setup complexity, and scalability.

```APIDOC
Memory Provider Comparison:

Provider        | Best For                 | Persistence              | Setup Complexity | Scalability
----------------|--------------------------|--------------------------|------------------|------------
In-Memory       | Development, testing     | None (lost on restart)   | None             | Single instance
Redis           | Production, caching      | Persistent               | Moderate         | High
PostgreSQL      | Production, complex queries | Full persistence         | High             | Very High
```

--------------------------------

### FAF Quick Start Installation and Testing

Source: https://github.com/xynehq/faf/blob/main/README.md

Demonstrates the basic commands to install dependencies, build the project, and run tests for the Functional Agent Framework using npm.

```bash
npm install
npm run build
npm test     # Run tests
```

--------------------------------

### Install and Start PostgreSQL (macOS)

Source: https://github.com/xynehq/faf/blob/main/docs/memory-system.md

Installs PostgreSQL using Homebrew and creates a database for FAF memory.

```bash
brew install postgresql && brew services start postgresql
createdb faf_memory
```

--------------------------------

### LiteLLM Server Setup and Configuration

Source: https://github.com/xynehq/faf/blob/main/docs/model-providers.md

Provides instructions for setting up a LiteLLM server, including installation via pip, creating a configuration file (`litellm.yaml`) to define model mappings and API keys, and starting the server using the `litellm` command.

```bash
pip install litellm[proxy]
```

```yaml
model_list:
  - model_name: gpt-4o
    litellm_params:
      model: openai/gpt-4o
      api_key: os.environ/OPENAI_API_KEY
  
  - model_name: claude-3-sonnet
    litellm_params:
      model: anthropic/claude-3-sonnet-20240229
      api_key: os.environ/ANTHROPIC_API_KEY
  
  - model_name: gemini-pro
    litellm_params:
      model: gemini/gemini-pro
      api_key: os.environ/GOOGLE_API_KEY
```

```bash
litellm --config litellm.yaml --port 4000
```

--------------------------------

### Basic Agent Setup Example

Source: https://github.com/xynehq/faf/blob/main/docs/model-providers.md

Demonstrates the fundamental setup for running an agent using the functional-agent-framework. It includes initializing a model provider (LiteLLM), defining a simple agent configuration, setting up run configuration, and executing the agent with an initial user message.

```typescript
import 'dotenv/config';
import { 
  run, 
  RunConfig, 
  RunState, 
  createTraceId, 
  createRunId,
  makeLiteLLMProvider 
} from 'functional-agent-framework';

// Set up model provider
const modelProvider = makeLiteLLMProvider(
  process.env.LITELLM_URL!,
  process.env.LITELLM_API_KEY!
);

// Define agent
const agent = {
  name: 'Assistant',
  instructions: () => 'You are a helpful assistant.',
  modelConfig: {
    name: 'gpt-4o',
    temperature: 0.7,
    maxTokens: 1000
  }
};

// Run configuration
const config: RunConfig<any> = {
  agentRegistry: new Map([['Assistant', agent]]),
  modelProvider,
  maxTurns: 10
};

// Execute
const result = await run({
  runId: createRunId(crypto.randomUUID()),
  traceId: createTraceId(crypto.randomUUID()),
  messages: [{ role: 'user', content: 'Hello!' }],
  currentAgentName: 'Assistant',
  context: {},
  turnCount: 0
}, config);
```

--------------------------------

### Basic Agent Execution

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Shows how to run an agent using the functional-agent-framework, including setting up a model provider, agent registry, and initial state for the conversation.

```typescript
import { run, makeLiteLLMProvider } from 'functional-agent-framework';

async function runAgent() {
  const modelProvider = makeLiteLLMProvider('http://localhost:4000');
  const agentRegistry = new Map([['Assistant', assistantAgent]]);
  
  const initialState = {
    runId: generateRunId(),
    traceId: generateTraceId(),
    messages: [{ role: 'user', content: 'What is 15 * 7?' }],
    currentAgentName: 'Assistant',
    context: { userId: 'user123', permissions: ['user', 'calculator'] },
    turnCount: 0,
  };
  
  const result = await run(initialState, {
    agentRegistry,
    modelProvider,
    maxTurns: 10,
    modelOverride: 'gpt-4o', // Optional model override
    onEvent: (event) => {
      console.log(`[${event.type}]`, event.data);
    },
  });
  
  return result;
}
```

--------------------------------

### Run Basic FAF Server Setup (TypeScript)

Source: https://github.com/xynehq/faf/blob/main/docs/server-api.md

Demonstrates the basic setup for running a Functional Agent Framework (FAF) server using the `runServer` function. It shows how to import necessary components, define an agent, configure model and memory providers, and start the server.

```TypeScript
import { runServer, makeLiteLLMProvider, createInMemoryProvider } from 'functional-agent-framework';

const myAgent = {
  name: 'MyAgent',
  instructions: () => 'You are a helpful assistant',
  tools: []
};

const modelProvider = makeLiteLLMProvider('http://localhost:4000');
const memoryProvider = createInMemoryProvider();

const server = await runServer(
  [myAgent], 
  { modelProvider },
  { port: 3000, defaultMemoryProvider: memoryProvider }
);
```

--------------------------------

### In-Memory Memory System Setup

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Configures the functional-agent-framework to use an in-memory provider for conversation history, enabling automatic storage and setting a maximum number of messages.

```typescript
import { createInMemoryProvider } from 'functional-agent-framework';

const memoryProvider = await createInMemoryProvider();

const config = {
  agentRegistry,
  modelProvider,
  memory: {
    provider: memoryProvider,
    autoStore: true, // Automatically store conversation history
    maxMessages: 100, // Keep last 100 messages per conversation
  },
  conversationId: 'user-session-123', // Required for memory persistence
};
```

--------------------------------

### Basic Chat Request Example

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Illustrates how to make a basic POST request to the `/chat` endpoint using curl, including a user message and agent specification.

```bash
curl -X POST http://localhost:3000/chat \
  -H "Content-Type: application/json" \
  -d '{ "messages": [{"role": "user", "content": "What is 15 * 7?"}], "agentName": "MathTutor", "context": {"userId": "demo", "permissions": ["user"]} }'
```

--------------------------------

### Conversation with Memory Persistence Example

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Shows how to initiate and continue a conversation using a specific `conversationId`, demonstrating memory persistence across multiple requests.

```bash
# Start conversation
curl -X POST http://localhost:3000/chat \
  -H "Content-Type: application/json" \
  -d '{ "messages": [{"role": "user", "content": "What is 15 * 7?"}], "conversationId": "my-conversation-1", "agentName": "MathTutor", "context": {"userId": "demo", "permissions": ["user"]} }'

# Continue conversation (references previous calculation)
curl -X POST http://localhost:3000/chat \
  -H "Content-Type: application/json" \
  -d '{ "messages": [{"role": "user", "content": "What was my previous calculation?"}], "conversationId": "my-conversation-1", "agentName": "MathTutor", "context": {"userId": "demo", "permissions": ["user"]} }'
```

--------------------------------

### Run FAF Server

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/test-memory-providers.md

Starts the FAF development server using npm. This command is used after configuring the .env file for the desired memory provider.

```bash
npm run dev
```

--------------------------------

### Environment Variables for Configuration

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Details the environment variables required for configuring the LiteLLM client, the FAF server, and different memory providers (Redis, PostgreSQL).

```bash
# Core LiteLLM configuration
LITELLM_URL=http://localhost:4000
LITELLM_API_KEY=sk-your-api-key
LITELLM_MODEL=gpt-3.5-turbo

# Server configuration
PORT=3000

# Memory provider selection
FAF_MEMORY_TYPE=memory  # or 'redis' or 'postgres'

# Redis configuration (if using Redis)
FAF_REDIS_HOST=localhost
FAF_REDIS_PORT=6379
FAF_REDIS_PASSWORD=your-password
FAF_REDIS_DB=0

# PostgreSQL configuration (if using PostgreSQL)
FAF_POSTGRES_HOST=localhost
FAF_POSTGRES_PORT=5432
FAF_POSTGRES_DB=faf_memory
FAF_POSTGRES_USER=postgres
FAF_POSTGRES_PASSWORD=your-password
```

--------------------------------

### Agent Handoffs with `handoffTool`

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Shows how to implement agent handoffs in FAF using the `handoffTool`. This pattern allows an agent to delegate tasks to other specialized agents by listing them in the `handoffs` property. No additional setup is required for the tool itself.

```typescript
import { handoffTool } from 'functional-agent-framework';

const triageAgent: Agent<MyContext, string> = {
  name: 'TriageAgent',
  instructions: () => 'Route requests to specialized agents based on the task.',
  tools: [handoffTool],
  handoffs: ['MathAgent', 'WeatherAgent'], // Allowed handoff targets
};

// The handoff tool automatically transfers control
// No additional setup required
```

--------------------------------

### Redis Provider Setup

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/test-memory-providers.md

Configures the .env file to use the Redis memory provider, specifying connection details like host, port, database, and a key prefix.

```bash
# .env configuration
FAF_MEMORY_TYPE=redis
FAF_REDIS_HOST=localhost
FAF_REDIS_PORT=6379
FAF_REDIS_DB=0
FAF_REDIS_PREFIX=faf:memory:
```

--------------------------------

### Agent Execution with Tracing

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Configures agent execution to include tracing and observability by integrating a ConsoleTraceCollector to log events during the agent's run.

```typescript
import { ConsoleTraceCollector } from 'functional-agent-framework';

const traceCollector = new ConsoleTraceCollector();

const config = {
  agentRegistry,
  modelProvider,
  maxTurns: 10,
  onEvent: traceCollector.collect.bind(traceCollector),
};
```

--------------------------------

### RAG Demo Queries

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

An array of predefined queries used to demonstrate the RAG capabilities, including knowledge retrieval, conversation memory, and source attribution.

```typescript
const demoQueries = [
  "What is return URL?",
  "How do I integrate hypercheckout on android? Remember what we discussed earlier."
];
```

--------------------------------

### Server API Endpoints

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Lists the available RESTful API endpoints for the FAF server, detailing their HTTP methods, descriptions, and common use cases for monitoring and interaction.

```APIDOC
API Endpoints:

/health
  GET: Health check for server monitoring and load balancer checks.

/agents
  GET: List available agents for discovery and frontend integration.

/chat
  POST: General chat endpoint for agent-agnostic conversations.

/agents/{name}/chat
  POST: Agent-specific endpoint for direct agent communication.

/memory/health
  GET: Check the health status of the memory provider system.

/conversations/:id
  GET: Retrieve a specific conversation by its ID.
  DELETE: Delete a specific conversation for cleanup purposes.
```

--------------------------------

### FAF Memory Provider Setup

Source: https://github.com/xynehq/faf/blob/main/docs/README.md

Demonstrates how to create and configure memory providers for conversation persistence in FAF. Includes examples for in-memory, Redis, and PostgreSQL providers.

```typescript
// Development
const memory = await createInMemoryProvider();

// Production
const memory = await createRedisProvider(config, redisClient);
const memory = await createPostgresProvider(config, pgClient);
```

--------------------------------

### Basic Chat Server Setup

Source: https://github.com/xynehq/faf/blob/main/docs/server-api.md

A minimal example of setting up a chat server using the `functional-agent-framework`. It configures a basic chat agent, a LiteLLM model provider, and an in-memory provider for conversation history.

```typescript
import { runServer, makeLiteLLMProvider, createInMemoryProvider } from 'functional-agent-framework';

const chatAgent = {
  name: 'ChatBot',
  instructions: () => 'You are a helpful assistant',
  tools: []
};

async function startChatServer() {
  const server = await runServer(
    [chatAgent],
    {
      modelProvider: makeLiteLLMProvider('http://localhost:4000'),
      maxTurns: 10
    },
    {
      port: 3000,
      cors: true,
      defaultMemoryProvider: createInMemoryProvider()
    }
  );
  
  console.log('Chat server running on http://localhost:3000');
}

startChatServer().catch(console.error);
```

--------------------------------

### Workflow Orchestration Configuration

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Sets up the configuration for running an agent workflow, including registering agents, specifying the model provider, setting maximum turns, and defining event handlers.

```typescript
const workflowConfig: RunConfig<MyContext> = {
  agentRegistry: new Map([
    ['TriageAgent', triageAgent],
    ['MathTutor', mathAgent],
    ['FileManager', fileAgent],
    ['RAGAgent', ragAgent]
  ]),
  modelProvider,
  maxTurns: 10,
  onEvent: (event) => {
    if (event.type === 'handoff') {
      console.log(`🔄 Handoff: ${event.data.from} → ${event.data.to}`);
    }
  }
};
```

--------------------------------

### Build FAF Framework and Server Demo

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/test-memory-providers.md

Commands to build the main FAF framework and the server demo using npm. This is a prerequisite for running the server.

```bash
cd /Users/anurag.sharan/repos/faf && npm run build
cd examples/server-demo && npm run build
```

--------------------------------

### Custom Model Configurations

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Shows how to customize AI model parameters for agents, such as `temperature` for creativity and `maxTokens` for response length. Two examples are provided: a `CreativeAgent` with high temperature and a `FactualAgent` with low temperature for consistency.

```typescript
const creativeAgent: Agent<MyContext, string> = {
  name: 'CreativeAgent',
  instructions: () => 'Be creative and imaginative in your responses.',
  tools: [],
  modelConfig: {
    temperature: 0.9, // High creativity
    maxTokens: 2000
  }
};

const factualAgent: Agent<MyContext, string> = {
  name: 'FactualAgent',
  instructions: () => 'Provide accurate, factual information only.',
  tools: [],
  modelConfig: {
    temperature: 0.1, // Low creativity, high consistency
    maxTokens: 500
  }
};
```

--------------------------------

### RAG Demo: Vertex AI Integration Overview

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Provides an overview of the RAG Demo, highlighting its integration with Google's Vertex AI, key features like streaming responses and source attribution, and its location.

```APIDOC
RAG Demo: Vertex AI Integration

Overview:
Demonstrates real-world integration with Google's Vertex AI RAG system, showcasing streaming responses, source attribution, and performance metrics.

Location: /Users/anurag.sharan/repos/faf/examples/rag-demo/

Key Features:
- Real Vertex AI Integration: Uses Google's @google/genai SDK.
- Streaming Responses: Real-time streaming from Vertex AI.
- Source Attribution: Automatic grounding and citation.
- Performance Metrics: Detailed timing and performance tracking.
- Permission Control: Role-based access to RAG functionality.
- Error Handling: Comprehensive error management.
- Memory Integration: Conversation persistence with FAF memory providers.
```

--------------------------------

### Custom Trace Collector Implementation

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Provides a custom trace collector that logs various events during agent execution, such as run starts, LLM calls, tool executions, handoffs, and run completions, with configurable log levels.

```typescript
import { TraceEvent, ConsoleTraceCollector } from 'functional-agent-framework';

class CustomTraceCollector {
  constructor(private logLevel: 'debug' | 'info' | 'warn' | 'error' = 'info') {}

  collect(event: TraceEvent) {
    switch (event.type) {
      case 'run_start':
        console.log(`🚀 [${event.data.runId}] Starting conversation`);
        break;
      
      case 'llm_call_start':
        console.log(`🤖 [${event.data.agentName}] Calling ${event.data.model}`);
        break;
      
      case 'tool_call_start':
        console.log(`🔧 [${event.data.toolName}] Executing with args:`, event.data.args);
        break;
      
      case 'tool_call_end':
        const status = event.data.status || 'completed';
        console.log(`✅ [${event.data.toolName}] ${status}`);
        break;
      
      case 'handoff':
        console.log(`🔄 Handoff: ${event.data.from} → ${event.data.to}`);
        break;
      
      case 'run_end':
        const outcome = event.data.outcome;
        if (outcome.status === 'completed') {
          console.log(`🎉 Conversation completed`);
        } else {
          console.error(`❌ Conversation failed:`, outcome.error);
        }
        break;
    }
  }
}
```

--------------------------------

### Environment Configuration (Bash)

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Provides essential environment variables for setting up and configuring the RAG system. This includes Google Cloud project details, LiteLLM proxy URL, API key, and the model to be used for RAG operations.

```bash
# Google Cloud configuration
GOOGLE_CLOUD_PROJECT=your-project-id

# LiteLLM configuration  
LITELLM_URL=http://localhost:4000
LITELLM_API_KEY=sk-your-api-key
LITELLM_MODEL=gemini-2.5-flash-lite

```

--------------------------------

### Implement Robust Tool Error Handling

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Shows how to integrate standardized error handling into a tool using the `withErrorHandling` utility. This example covers input validation, permission checks, and general execution errors, returning structured `ToolResponse` objects.

```typescript
import { withErrorHandling, ToolResponse, ToolErrorCodes } from 'functional-agent-framework';

const robustTool: Tool<{ input: string }, MyContext> = {
  schema: {
    name: "robust_tool",
    description: "A tool with proper error handling",
    parameters: z.object({
      input: z.string()
    }),
  },
  execute: withErrorHandling('robust_tool', async (args, context) => {
    // Validation
    if (!args.input.trim()) {
      return ToolResponse.validationError("Input cannot be empty");
    }
    
    // Permission check
    if (!context.permissions.includes('required_permission')) {
      return ToolResponse.permissionDenied(
        "This tool requires special permissions",
        ['required_permission']
      );
    }
    
    // Business logic
    const result = processInput(args.input); // Assuming processInput is defined elsewhere
    
    return ToolResponse.success(result, {
      processingTime: Date.now(),
      inputLength: args.input.length
    });
  }),
};

```

--------------------------------

### Runtime Error Handling in FAF

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Demonstrates how to handle various runtime error types returned by FAF, such as MaxTurnsExceeded, ToolCallError, AgentNotFound, and InputGuardrailTripwire. It shows a TypeScript example using a switch statement to differentiate and process specific error conditions.

```typescript
import { FAFErrorHandler } from 'functional-agent-framework';

if (result.outcome.status === 'error') {
  const error = result.outcome.error;
  
  switch (error._tag) {
    case 'MaxTurnsExceeded':
      console.log(`Exceeded ${error.turns} turns`);
      break;
    case 'ToolCallError':
      console.log(`Tool ${error.tool} failed: ${error.detail}`);
      break;
    case 'AgentNotFound':
      console.log(`Agent ${error.agentName} not found`);
      break;
    case 'InputGuardrailTripwire':
      console.log(`Input blocked: ${error.reason}`);
      break;
    // ... handle other error types
  }
}
```

--------------------------------

### Create Tool Debugging Wrapper (TypeScript)

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

This TypeScript code defines a higher-order function `createDebuggingWrapper` that enhances a `Tool` object by adding logging for execution start, arguments, context, results, and errors. It helps in debugging tool execution flow and performance.

```typescript
const createDebuggingWrapper = <A, Ctx>(tool: Tool<A, Ctx>) => ({
  ...tool,
  execute: async (args: A, context: Ctx) => {
    const startTime = Date.now();
    console.log(`🔧 [${tool.schema.name}] Starting with:`, {
      args: JSON.stringify(args, null, 2),
      context: {
        userId: context.userId,
        permissions: context.permissions
      }
    });
    
    try {
      const result = await tool.execute(args, context);
      const duration = Date.now() - startTime;
      
      console.log(`✅ [${tool.schema.name}] Completed in ${duration}ms`);
      console.log(`📤 [${tool.schema.name}] Result:`, result);
      
      return result;
    } catch (error) {
      const duration = Date.now() - startTime;
      
      console.error(`❌ [${tool.schema.name}] Failed after ${duration}ms:`, error);
      throw error;
    }
  }
});
```

--------------------------------

### Node.js Application Dockerfile (Multi-stage)

Source: https://github.com/xynehq/faf/blob/main/docs/deployment.md

A multi-stage Dockerfile for building a production-ready Node.js application. It uses a builder stage for installing dependencies and compiling code, then copies only necessary artifacts to a minimal production image. Includes non-root user setup, signal handling with dumb-init, and a health check.

```dockerfile
# Multi-stage build for production
FROM node:18-alpine AS builder

WORKDIR /app

# Copy package files
COPY package*.json ./
COPY tsconfig.json .

# Install dependencies
RUN npm ci --only=production && npm cache clean --force

# Copy source code
COPY src/ ./src/

# Build application
RUN npm run build

# Production stage
FROM node:18-alpine AS production

# Install dumb-init for proper signal handling
RUN apk add --no-cache dumb-init

# Create non-root user
RUN addgroup -g 1001 -S faf && \
    adduser -S faf -u 1001

WORKDIR /app

# Copy built application and dependencies
COPY --from=builder /app/node_modules ./node_modules
COPY --from=builder /app/dist ./dist
COPY --from=builder /app/package*.json .

# Change ownership to non-root user
RUN chown -R faf:faf /app
USER faf

# Health check
HEALTHCHECK --interval=30s --timeout=3s --start-period=5s --retries=3 \
    CMD node -e "http.get('http://localhost:' + (process.env.PORT || 3000) + '/health', (res) => process.exit(res.statusCode === 200 ? 0 : 1))"

# Expose port
EXPOSE 3000

# Use dumb-init for proper signal handling
ENTRYPOINT ["dumb-init", "--"]
CMD ["node", "dist/index.js"]
```

--------------------------------

### TypeScript Streaming Chat Tool

Source: https://github.com/xynehq/faf/blob/main/docs/examples.md

Defines a tool for initiating streaming chat conversations. It takes a message and session ID, starts a background streaming process, and returns a stream URL for real-time responses. Dependencies include 'eventsource' and internal utilities like 'generateStreamId', 'startStreamingResponse', 'createResponseStream', 'streamLLMResponse', and 'delay'.

```typescript
import { EventSource } from 'eventsource';

const streamingChatTool: Tool<StreamingArgs, StreamingContext> = {
  schema: {
    name: "streaming_chat",
    description: "Initiate streaming conversation",
    parameters: z.object({
      message: z.string(),
      sessionId: z.string()
    })
  },
  execute: async (args, context) => {
    const streamId = generateStreamId();
    
    // Start streaming in background
    startStreamingResponse(streamId, args.message, context);
    
    return ToolResponse.success('Streaming started', {
      streamId,
      streamUrl: `/stream/${streamId}`,
      message: 'Connect to stream URL for real-time responses'
    });
  }
};

async function startStreamingResponse(streamId: string, message: string, context: StreamingContext) {
  const stream = createResponseStream(streamId);
  
  try {
    stream.write({ type: 'start', data: { streamId } });
    
    // Simulate streaming LLM response
    const response = await streamLLMResponse(message);
    
    for await (const chunk of response) {
      stream.write({ type: 'chunk', data: { content: chunk } });
      await delay(50); // Simulate realistic streaming delay
    }
    
    stream.write({ type: 'end', data: { completed: true } });
  } catch (error) {
    stream.write({ type: 'error', data: { error: error.message } });
  } finally {
    stream.close();
  }
}
```

--------------------------------

### Run Development Server

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/README.md

Starts the FAF development server using npm. The server will be accessible at http://localhost:3000 by default.

```bash
npm run dev
```

--------------------------------

### Run the Vertex AI RAG Demo

Source: https://github.com/xynehq/faf/blob/main/examples/rag-demo/README.md

Provides instructions on how to run the demo application after setting up dependencies and environment variables. It includes commands for direct execution (`npm run dev`) and for building and then starting the application.

```bash
# Make sure you have your .env file configured
cp .env.example .env
# Edit .env with your actual values

# Run the demo
npm run dev

# Or with TypeScript compilation
npm run build
npm start
```

--------------------------------

### Server API Endpoints and Usage

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Defines the available server API endpoints for the FAF framework, including health checks, agent listing, and chat functionalities. It also provides an example of how to interact with the general chat endpoint using cURL, specifying messages, agent name, conversation ID, and context.

```APIDOC
Server API Endpoints:

- GET /health: Health check for the server.
- GET /agents: Lists all available agents.
- POST /chat: General chat endpoint for interacting with agents.
- POST /agents/{name}/chat: Agent-specific chat endpoint.
- GET /memory/health: Health check for the memory system.

Example API Usage (cURL):

```bash
# Chat with an agent
curl -X POST http://localhost:3000/chat \
  -H "Content-Type: json" \
  -d '{ "messages": [{"role": "user", "content": "What is 15 * 7?"}], "agentName": "SmartAssistant", "conversationId": "session-123", "context": {"userId": "user123", "permissions": ["user", "calculator"]} }'
```
```

--------------------------------

### Start PostgreSQL Server

Source: https://github.com/xynehq/faf/blob/main/examples/server-demo/README.md

Commands to start a PostgreSQL server, either via Docker, Homebrew (macOS), or apt (Ubuntu). PostgreSQL is used for persistent conversation storage.

```bash
# Using Docker
docker run -d -p 5432:5432 -e POSTGRES_PASSWORD=password -e POSTGRES_DB=faf_memory postgres:15

# Or install locally (macOS)
brew install postgresql
brew services start postgresql
createdb faf_memory

# Or install locally (Ubuntu)
sudo apt install postgresql postgresql-contrib
sudo systemctl start postgresql
sudo -u postgres createdb faf_memory
```

--------------------------------

### Create a Multi-Tool Agent with TypeScript

Source: https://github.com/xynehq/faf/blob/main/docs/getting-started.md

Demonstrates creating a sophisticated agent capable of using multiple tools, including a calculator and a weather tool. It shows how to define tool schemas, implement execution logic with context, and configure the agent's instructions and model settings.

```typescript
import {
  Tool,
  Agent,
  ToolResponse,
  ToolErrorCodes,
  withErrorHandling
} from 'functional-agent-framework';

type MyContext = {
  userId: string;
  permissions: string[];
};

// Math calculation tool with error handling
const calculatorTool: Tool<{ expression: string }, MyContext> = {
  schema: {
    name: "calculate",
    description: "Perform mathematical calculations",
    parameters: z.object({
      expression: z.string().describe("Math expression to evaluate (e.g., '2 + 2', '10 * 5')")
    }),
  },
  execute: withErrorHandling('calculate', async (args, context) => {
    // Input validation
    const sanitized = args.expression.replace(/[^0-9+\-*/().\s]/g, '');
    if (sanitized !== args.expression) {
      return ToolResponse.validationError(
        "Invalid characters in expression. Only numbers, +, -, *, /, (, ), and spaces are allowed.",
        { originalExpression: args.expression }
      );
    }
    
    try {
      const result = eval(sanitized);
      return ToolResponse.success(`${args.expression} = ${result}`, {
        originalExpression: args.expression,
        result,
        calculationType: 'arithmetic'
      });
    } catch (evalError) {
      return ToolResponse.error(
        ToolErrorCodes.EXECUTION_FAILED,
        `Failed to evaluate expression: ${evalError instanceof Error ? evalError.message : 'Unknown error'}`,
        { expression: args.expression }
      );
    }
  }),
};

// Weather tool (mock implementation)
const weatherTool: Tool<{ location: string }, MyContext> = {
  schema: {
    name: "get_weather",
    description: "Get current weather information for a location",
    parameters: z.object({
      location: z.string().describe("City name or location")
    }),
  },
  execute: async (args, context) => {
    // Check permissions
    if (!context.permissions.includes('weather_access')) {
      return ToolResponse.permissionDenied(
        "Weather access requires 'weather_access' permission",
        ['weather_access']
      );
    }
    
    // Mock weather data
    const weatherData = {
      location: args.location,
      temperature: Math.floor(Math.random() * 30) + 10,
      condition: ['sunny', 'cloudy', 'rainy'][Math.floor(Math.random() * 3)],
      humidity: Math.floor(Math.random() * 100)
    };
    
    return ToolResponse.success(
      `Weather in ${weatherData.location}: ${weatherData.temperature}°C, ${weatherData.condition}, ${weatherData.humidity}% humidity`,
      weatherData
    );
  },
};

// Create a multi-tool agent
const assistantAgent: Agent<MyContext, string> = {
  name: 'SmartAssistant',
  instructions: (state) => `You are a helpful assistant with access to calculation and weather tools.
  
  Current user: ${state.context.userId}
  User permissions: ${state.context.permissions.join(', ')}
  
  You can:
  - Perform mathematical calculations using the calculator tool
  - Get weather information (if user has weather_access permission)
  - Engage in helpful conversation
  
  Always use the appropriate tools when the user asks for calculations or weather information.`,
  
  tools: [calculatorTool, weatherTool],
  
  modelConfig: {
    temperature: 0.1, // Lower temperature for more consistent responses
    maxTokens: 1000
  }
};

```