Vector-cache

Support to add context along with the query, useful when similar queries are fired across different contexts.

Major Update

- Update vector store implementations to use cosine similarity instead of cosine distances, as its a more semantically relevant metric (and is always positive).
- Add redis_vector store example
- Added feature for adaptive threshold adjustment
- Update example function signature to new vector cache

- Cleaned requirement.txt, now only ChromaDB and openai are required installs.
- Added default value to inmemory LRU and LFU cache storage (only for testing).
- Updated README
- Fixed module init files, so not to throw error of optinal imports.
- Added LRU example.

Major Release
Add new cache stores support. We now support:

memcache
simple in_memory lru cache (only for development and testing)
simple in_memory lfu cache (only for development and testing)
Add new vector stores support. We now support:

pgvector
redis vector store
qdrant
pinecone
Update setup.py, so that 3rd party libraries are now extra_requirements

Add support for defining custom keys used as index for the documents.:

This can be useful when namespacing for different things of the same type. (different properties on a hotel booking site).

You can define a prefix (string)

You can pass a function or lambda (callable)

Defaults to UUID4

Major bugfix release.

1. Fixed persistent mode ChromaDB issues.
2. Added example using OpenAI embeddings & openai chat completion.
3. Updated openai embedding models available.