v0.4.16

LlamaIndex 0.4.16:

• Elastic data loader + Opensearch based vector index! (Huge s/o @jaylmiller )
• Add child branch factor for tree index embedding queries (thanks @kpister )

v0.4.15

• Biggest update is that we’ve fixed the composability over vector stores
• Added @weaviate_io multi-threaded batch importing to decrease load time

v0.4.14

patch fixes:

• patch fix simple vector index in composability
• patch fix weaviate save/load

v0.4.13

LlamaIndex 0.4.13 is out:

• Our KG Index has gotten a big upgrade: now you can query the KG using embeddings instead of exact keyword matching! Huge s/o @logan-markewich
• Our download_loader integration with LlamaHub now allows for more complex loaders that require multiple files! Check out our @github integration (s/o @ahmetkca)
• S/o @Ryanglambert for fixing upset behavior in our pinecone index.

v0.4.12

We’ve gotten user feedback on a key pain point of applying LLM’s to production SQL: too many tables+columns, can’t fit context in one prompt!

LlamaIndex 0.4.12 brings a big new improvement in that direction: you can pass table context to our SQL index…and this table context can be another index 🧠

See in-depth text-to-SQL guide here: https://gpt-index.readthedocs.io/en/latest/guides/sql_guide.html

v0.4.11

LlamaIndex 0.4.11:

• Landed a big refactor to decouple vector storage from the index build/query logic. No user-facing changes atm, but should make it much easier to add new vector stores / maintain the code!
• Added async to vector index construction (s/o @thejessezhang) ⚡)

v0.4.10

LlamaIndex v0.4.10

v0.4.9

LlamaIndex 0.4.9 adds support for a Chroma reader and Chroma vector store! A huge thanks to @atroyn for helping to add this to the LlamaIndex repo. 🎉

LlamaIndex v0.4.8

• Added ability to customize text splitter for a given index!
• PSA: all loaders from LlamaHub https://github.com/emptycrown/llama-hub now use llama_index by default instead of gpt_index. If you do not wish to use the llama_index package yet, we've added a use_gpt_index_import option for download_loader. Set it to True, and you can retain gpt_index imports.

LlamaIndex v0.4.7

LlamaIndex v0.4.7 adds an index playground module:

• If you're not sure which index, model, embedding, etc to use, you can now utilize Playground to easily test them all at once and monitor results.
• Added a documentation page as well.