This directory contains the documentation for Marvin, a Python framework for building AI applications with LLMs.
The documentation is organized into the following sections:
- Tasks - The fundamental building blocks of AI workflows
- Agents - Specialized AI workers with different roles and capabilities
- Threads - Maintaining conversation context across multiple interactions
- Teams - Coordinating multiple AI agents to solve complex problems
- Tools and Context - Extending AI capabilities with custom functions and additional information
- Memory - Enabling agents to remember information across conversations
- run - Execute a task with an LLM
- classify - Categorize content into predefined classes
- extract - Pull structured data from unstructured text
- cast - Convert content to a specified structure
- generate - Create structured data or content
- summarize - Create concise summaries of content
- say - Have conversational interactions with an LLM
- plan - Create structured plans for complex tasks
- fn - Create AI-powered functions with a decorator
- Installation - Install Marvin and set up your environment
- Quickstart - Build your first AI application in minutes
- Configure LLMs - Use different LLM providers with Marvin
- Configuration - Configure Marvin using environment variables and settings
- Building a Multi-step Workflow - Create a complete AI application with multiple connected steps
- Building a Conversational Assistant - Create a personalized chatbot with memory
- Migration Guide - Upgrade to Marvin 3.0 from previous versions
- Memory - Patterns for using memory in your applications
- Tools - Patterns for creating and using tools
- Task Results - Working with structured task results
- Running Tasks - Different ways to run tasks
- Instructions - Crafting effective instructions
- Interactivity - Building interactive applications
If you'd like to contribute to the documentation:
- Make your changes or additions following the existing format and style
- Use clear, concise language and provide practical examples
- Submit a pull request with your changes
The documentation uses Mintlify for rendering. The configuration is in the mint.json file.
- Add more examples and use cases
- Expand multi-modal capabilities documentation as they are implemented
- Add more integration guides with other frameworks and services