Skip to content

AI Engines

The AI service supports two engine choices: Pydantic AI and LangChain. You select your engine at project generation time, and both provide the same user-facing API.

Overview

An "engine" is the underlying AI framework that powers your chat functionality. Both engines:

  • Support all providers
  • Provide identical CLI and REST API interfaces
  • Handle conversation management the same way
  • Support streaming responses

The difference is internal implementation, how the engine communicates with the AI models. What matters for you, the builder, comes down to the following: Are you going to extend the service? If so, then pick the framework you're used to.

Pydantic AI

Pydantic AI is built by the Pydantic team, providing a type-safe, validation-first approach to AI.

LangChain

LangChain is a comprehensive framework for building LLM applications with extensive tooling and integrations.

Selecting Your Engine

Engine selection happens at project generation time using bracket notation:

# Generate with Pydantic AI (default)
aegis init my-app --services ai

# Generate with LangChain
aegis init my-app --services ai[langchain]

# LangChain with specific provider
aegis init my-app --services ai[langchain,openai]

# LangChain with SQLite persistence
aegis init my-app --services ai[langchain,sqlite,groq]

Engine is Compile-Time

Unlike providers (which you can switch via environment variables), the engine choice is "baked in" at project generation. To switch engines, you'd need to regenerate the project.

Dependencies

Each engine installs its own set of dependencies (depending on which providers you choose to use):

Pydantic AI: 

pydantic-ai-slim[openai,anthropic,google,groq]

LangChain: 

langchain-openai
langchain-anthropic
langchain-google-genai
langchain-groq
langchain-mistralai
langchain-cohere

Next Steps:

Was this page helpful?