Which vector database should be used?

[thompsonmj]

Compile a list of viable options along with relevant pros, cons, and questions to inform an appropriate choice.

[thompsonmj]

Let's make a table with an entry for each option and columns for attributes to consider:

• Open Source
• Able to be self-hosted for free
• Supports Python integration
• Standalone
• Maintained (recently)
• How hosting works
• Licensing
• Clear documentation

[thompsonmj]

Narrowing comparison to:

• Weaviate
• Milvus
• PostgreSQL-pgvector
• FAISS

Comparison table

[thompsonmj]

We'll start by testing Milvus (and Milvus Lite) as a low-complexity option to help inform specs for #2

[thompsonmj]

@smenon8's implementation with chromadb looks like a solid choice, se we can close this and later open a new issue if re-evaluation of the db choice is warranted.