Add baseline DSPy implementation

[yoomlam]

Add baseline DSPy implementation

• Start openhermes LLM using Ollama: ollama run openhermes (Install Ollama from https://ollama.ai/)
• Run:

pip install -r requirements.txt
python dspy_engine.py

Note the DSPy prompt is not yet optimized ("compiled" in DSPY parlance). Waiting on better retrieval results.

[ccheng26]

Adding a clean up function for the embeddings issue with chromadb while testing would be helpful
chromadb.errors.InvalidDimensionException: Embedding dimension 384 does not match collection dimensionality 768

[yoomlam]

Adding a clean up function for the embeddings issue with chromadb while testing would be helpful chromadb.errors.InvalidDimensionException: Embedding dimension 384 does not match collection dimensionality 768

By "clean up", do you mean delete the chromadb folder? If so, I'm hesitant to be deleting any DB automatically in code. I think it's easy enough to run rm -rf chroma_db on the commandline before testing, e.g., rm -rf chroma_db && python dspy_engine.py. With that said, I don't expect to delete the DB at all while testing DSPy -- I only delete the DB when testing different embedding/retrieval parameters.

[ccheng26]

Adding a clean up function for the embeddings issue with chromadb while testing would be helpful chromadb.errors.InvalidDimensionException: Embedding dimension 384 does not match collection dimensionality 768

By "clean up", do you mean delete the chromadb folder? If so, I'm hesitant to be deleting any DB automatically in code. I think it's easy enough to run rm -rf chroma_db on the commandline before testing, e.g., rm -rf chroma_db && python dspy_engine.py. With that said, I don't expect to delete the DB at all while testing DSPy -- I only delete the DB when testing different embedding/retrieval parameters.

I suppose cleaning up the collections won't be as necessary in this file as we're not swapping embeddings often when testing, but if we do plan on integrating this with the chatbot we should have some option delete the files as opposed to doing this everytime.