spec-ai

Public library crate for the spec-ai framework.

Overview

This is the main library crate that re-exports all public APIs from the spec-ai workspace crates. It provides a unified interface for building AI agent applications.

Features

Default Features

• openai - OpenAI API integration
• lmstudio - LM Studio local models
• web-scraping - Web scraping capabilities
• vttrs - Video/subtitle processing
• api - HTTP API server
• cli - Command-line interface

Optional Features

LLM Providers:

• anthropic - Anthropic Claude API
• ollama - Ollama local models
• mlx - Apple MLX framework

Database:

• bundled - Bundled DuckDB library (recommended)
• duck-sys - System DuckDB library

Other:

• integration-tests - Enable integration tests
• axum-extra - Additional Axum web framework features

Installation

Add to your Cargo.toml:

[dependencies]
spec-ai = "0.5"

Or install the CLI:

cargo install spec-ai --features bundled

Usage

As a Library

use spec_ai::prelude::*;

// Your agent application code here

As a Binary

This crate also provides the spec-ai binary:

# Start interactive session
spec-ai

# Run a spec file
spec-ai run task.spec

# Use custom config
spec-ai --config custom.toml

Workspace Structure

This crate re-exports functionality from:

• spec-ai-core - Agent runtime, tools, embeddings
• spec-ai-config - Configuration management and persistence
• spec-ai-knowledge-graph - Knowledge graph storage, vector clocks, types
• spec-ai-policy - Policy engine and plugin system
• spec-ai-api - HTTP API server
• spec-ai-cli - Command-line interface

Documentation

For detailed documentation, see:

• Main README - Overview and quick start
• docs/ARCHITECTURE.md - System architecture
• docs/CONFIGURATION.md - Configuration guide
• docs/SETUP.md - Setup instructions

Examples

Example configurations and code can be found in the repository:

• examples/configs/ - Configuration examples
• examples/code/ - Code examples
• specs/ - Example spec files

License

MIT License - see LICENSE or the main README for details.

Contributing

Create an issue or open a PR at the spec-ai repository.