Clean code, clear mind, solutions you'll find.
This project aims to re-factor the implementation of the original ReWOO's setting into graph-based design. In addtion, the project evaluate the performance of ReWOO based on the bigbench
and SOTU
dataset.
Why's ReWOO, you might ask? There are several factors that make ReWOO stands out:
- Reduce token consumption.
- Reduce execution time.
- Fine-tuning does not affect the whole system
- Graph-based implementation of ReWOO.
- Evaluation with Ragas.
- Lang-'s ecosystem: LangChain and LangGraph.
- Tavily Search.
- Groq.
- Ollama.
- RAGAs.
Set up environment
Note: For those who want to use Gemini or Ollama, make sure to get those API keys and install the relevant packages.
Clone the repo and install requirements
git clone https://github.com/8Opt/ReWOO-with-Graph.git
cd ReWOO-with-Graph
pip install -r requirements.txt
Ready to use
All informtion of using the code is well-instructed in the /assets/notebooks
.