Unlock more features

[amotl]

About

Other than the fundamental features about support for LangChain's Vector Store, Document Loader, and Chat History features, there are more features that could be unlocked for CrateDB.

Those have been identified and enumerated within the backlog: Cache, Pipeline, Docstores, Retrievers, Store, ByteStore. Inspirations can be pulled from the full list of supported stores.

Details

This section shares a few relevant pointers to features of LangChain, and providers offering features not unlocked for CrateDB yet.

By feature

• Cache: Cache: Add CrateDBCache and CrateDBSemanticCache #12
• SQLRecordManager: Add support for SQLRecordManager crate-workbench/langchain#8
• A dedicated bm25 module: https://python.langchain.com/docs/integrations/providers/rank_bm25/

By language model

• https://github.com/abetlen/llama-cpp-python
• https://github.com/deepseek-ai/DeepSeek-R1
• https://github.com/huggingface/open-r1
• https://github.com/jacklinquan/portable_deepseek_r1_distill_qwen_7b

By provider

• Apache Spark: https://python.langchain.com/docs/integrations/providers/spark/
• Elasticsearch: https://python.langchain.com/docs/integrations/providers/elasticsearch/
• MongoDB: https://python.langchain.com/docs/integrations/providers/mongodb/
• MongoDB Atlas: https://python.langchain.com/docs/integrations/providers/mongodb_atlas/
• OpenSearch: https://python.langchain.com/docs/integrations/providers/opensearch/
• OracleAI: https://python.langchain.com/docs/integrations/providers/oracleai/