Agent-C

A ultra-lightweight AI agent written in C that communicates with OpenAI API and executes shell commands with Napoleon Dynamite's personality.

Features

• Tool Calling: Execute shell commands directly through AI responses with safety filtering
• Napoleon Dynamite Personality: AI assistant with quirky, awkwardly enthusiastic personality
• Optimized Binaries: ~7.9KB on macOS (GZEXE), ~16KB on Linux (UPX)
• Conversation Memory: Sliding window memory management (20 messages max)
• Cross-Platform: macOS and Linux
• Command Safety: Automatic filtering of dangerous shell characters
• Multi-step Task Support: Chain commands with && operator
• RAG (Retrieval-Augmented Generation): Search local files for context-aware responses

Quick Start

Prerequisites

• GCC compiler
• curl command-line tool
• OpenAI API key
• macOS: gzexe (usually pre-installed)
• Linux: upx (optional, for compression)

Build

make

The build system auto-detects your platform and applies optimal compression:

• macOS: Uses GZEXE compression → ~7.9KB binary
• Linux: Uses UPX compression → ~16KB binary

Setup

Set your OpenAI API key and configuration:

export OPENAI_KEY=your_openai_api_key_here
export OPENAI_BASE=https://api.openai.com  # Optional, defaults to OpenAI
export OPENAI_MODEL=gpt-3.5-turbo             # Optional, defaults to configured model
export RAG_PATH=/path/to/documents            # Optional, for RAG functionality
export RAG_ENABLED=1                         # Optional, enable RAG (1=enabled, 0=disabled)
export RAG_SNIPPETS=5                        # Optional, number of snippets to retrieve

### Run

```bash
./agent-c

Run with RAG

# Enable RAG with a specific path
./agent-c --rag /path/to/documents

# Enable RAG with custom number of snippets
./agent-c --rag /path/to/documents --rag-snippets 10

Configuration

The agent uses the following environment variables:

• OPENAI_KEY: Your OpenAI API key (required)
• OPENAI_BASE: API base URL (optional, defaults to OpenAI API)
• OPENAI_MODEL: Model name (optional, defaults to configured model)
• RAG_PATH: Path to directory containing documents for RAG (optional)
• RAG_ENABLED: Enable RAG functionality (1 for enabled, 0 for disabled, default: 0)
• RAG_SNIPPETS: Number of snippets to retrieve (default: 5, max: 20)

Usage Examples

Start the agent and interact with it:

Agent> Hello
Gosh! Hello there! Sweet!

The agent can execute shell commands:

Agent> Create a file called test.txt and write "Hello World" to it
$ echo "Hello World" > test.txt
Command output:

Multi-step tasks are supported:

Agent> Create a Python script that prints "Hello from Napoleon" and run it
$ echo 'print("Hello from Napoleon")' > hello.py && python3 hello.py
Command output:
Hello from Napoleon

RAG (Retrieval-Augmented Generation)

Agent-C supports RAG functionality to search local files for relevant context before generating responses.

How RAG Works

1. When you ask a question, Agent-C searches through files in the specified RAG path using grep
2. It finds relevant snippets containing keywords from your query
3. These snippets are included in the context sent to the AI model
4. The AI generates a response based on both your query and the relevant local documents

Supported File Types

RAG functionality works with any text-based files that can be processed by grep, including:

• .txt, .md, .c, .h, .py, .js, .html, .css, .json, .xml, .yaml, .yml
• Any other plain text files

Example Usage with RAG

# With RAG enabled
> What are the main features of this project?

# Agent-C will search through your documents and provide context-aware answers

Architecture

• main.c: Entry point with signal handling and configuration loading
• agent.c: Core agent logic, command execution, and personality handling
• cli.c: Command-line interface with colored prompts
• json.c: JSON parsing and request/response handling
• utils.c: HTTP requests and configuration management
• rag.c: RAG functionality for local file search
• args.c: Command-line argument parsing
• agent-c.h: Header with data structures and function declarations

License

CC0 - "No Rights Reserved"