-
Install Ollama by following the instructions in the Ollama documentation.
-
Run Ollama with the following commands
ollama run "phi3:latestfor the LLM modelollama run "nomic-embed-text"for the embeddings model -
Install Chromadb by following the instructions in the Chromadb documentation.
-
Once Ollama and Chromadb are installed, you can use the following commands to run the project:
-
To populate the database using Ollama:
python populate_database.py -
To query the database using Ollama:
python query_data.py "What are the latest Credit challenges for Amazon currently?" -
To reset the database
python populate_database.py --reset
-
Additional Notes: - Embedding Model: The current embedding model being used is
nomic-embed-text. You can utilize theget_embedding_function.pyscript to select different embedding models such as AWS Bedrock, Hugging Face, or OpenAI. The quality of the embeddings directly impacts the retrieval accuracy.
- Vector Store: Chroma is being utilized as the vector store for local development due to time constraints. However, any other vector store supported by Langchain community can be used.