TaskWeaver
FreeA code-first agent framework for seamlessly planning and executing data analytics tasks. 
About TaskWeaver
TaskWeaver is a code-first agent framework developed by Microsoft for seamlessly planning and executing data analytics tasks. It interprets user requests through code snippets and efficiently coordinates plugins in the form of functions to execute tasks in a stateful manner. Unlike traditional chat-based agents, TaskWeaver preserves both chat history and code execution history, including in-memory data like DataFrames, enabling rich data structure handling. Key capabilities include task decomposition and progress tracking for complex tasks, reflective execution for self-correction, support for various LLMs including local models, vision input for the Planner role, a reasoning role called Recepta, shared memory for inter-role communication, and dynamic experience selection. It can be run via Docker with a containerized code execution environment and integrates with AgentOps for monitoring.
Key Features
Pros & Cons
- Code-first approach provides expressiveness and flexibility for data tasks
- Stateful execution retains in-memory data, enabling iterative analysis
- Reflective execution improves robustness and accuracy
- Supports both cloud-based and local LLMs, reducing dependency on external APIs
- Modular plugin design allows easy extension with custom functions
- Active development with frequent updates and community contributions
- Free and open-source (MIT license)
- Docker support simplifies setup and deployment
- Repository archived as read-only (future date noted), potentially limited future updates
- Primarily focused on data analytics; may not suit general agent tasks
- Requires Python environment and dependency management
- Steeper learning curve for users unfamiliar with code-based agents
- Documentation may still be evolving