LlamaIndex
Ingestion and retrieval plumbing for document corpora.
- made by
- LlamaIndex
- source
- Official LlamaIndex site
Magic Ship is one shop in Vancouver, BC, working remotely with clients worldwide. We are not a partner, reseller, or certified vendor of LlamaIndex - we just build with this.
What LlamaIndex is
LlamaIndex is a Python and TypeScript framework for getting documents into an index and answers back out: loaders, node parsers, embedding and storage adapters, retrievers, and response synthesis. It has connectors for most vector stores and a document parsing service, LlamaParse, aimed at tables and scanned PDFs.
How we use it
We use the ingestion half more than the query half: parsers that keep section headings attached to their chunks, metadata extraction, and node relationships that let a retrieved chunk pull in its neighbours. The retrieval path itself usually ends up hand-written against the store so the SQL and the filters stay ours to tune. LlamaParse earns its place on PDFs where layout carries meaning.
Where it is the wrong choice
The abstractions move quickly and there is more than one way to do most things, which ages a codebase written against an older release. We keep the framework at the edges of a production retrieval path rather than at the centre of it.
Service lines it turns up in
Related tools
More in Retrieval and vector
Other tools in the same service lines
Building something on LlamaIndex?
Send the problem rather than a job spec. You get an answer on scope, on fit, and on whetherLlamaIndex is even the right call for it.
Start a projectLlamaIndex and LlamaIndex are trademarks of their respective owners, used here to say what we work with.