Monorepo with our tools to build RAG chatbots from dynamic sources.
CLI tool that leverages the execution of RAG generation workflows using configuration files. This tool will help you with:
- Declare and gather different sources of data, both locally and remotely.
- Parse this data into a LLM friendly format (MarkDown).
- Chunk the data with your desired chunking strategy.
- Embed the chunks with your model and upload them to your vector store.
- Serve an API to make prompts making use of your vector store.
Read more on its usage and examples
Features
- Expandability Every step of the workflow can we tweaked and expanded thanks to its Plugin Architecture, allowing end-customers to add new steps and implementations.
- Declarative workflow orchestration The behaviour of the system is controlled by configuration files, making the executions really flexible and reproducible.
- Traceability We grant that the resulting chunks of information are fully traceable to their sources.