The post asks if LlamaIndex has plans to develop and provide support for a FastAPI-like tool similar to LangChain's LangServe. The comments discuss the following:
- A community member is new to this space and is trying to figure out the best library to start with long-term.
- Another community member suggests the SECInsights full-stack app template and the privateGPT project (using LlamaIndex) as potential starting points.
- The original poster clarifies that they are interested in the long-term support and feature updates that LangChain provides for its FastAPI-like tool, and they are trying to understand how LlamaIndex compares to LangChain.
- The LlamaIndex community member responds that they are currently more focused on retrieval, queries, and RAG, and they are open to building something similar to LangChain's FastAPI tool based on feedback like this.
- A community member notes that the SECInsights release is nice but not modular, as it requires importing documents from an S3 bucket each time.
- Another community member asks about LlamaIndex's plans for supporting multi-agents and integration with Autogen.
@Andrei two things: 1) we have a full-stack app template in the form of secinsights if you want to get started building an app on that: https://github.com/run-llama/sec-insights
Thanks, good learning to start with. But what I mean is that LangChain will provide continuous updating and support on that FastAPI image with new features. I'm still new to this and try to understand what library is better. I like the multiple query methods from LlamaIndex, but don't have a full picture on long term where LlamaIndex will be compared with LangChain ?
if there's specific features on the fastapi image that you think would be useful let us know, right now my understanding is it's a relatively thin wrapper around their existing stuff.
right now we're a lot deeper into retrieval/queries/RAG. so if you want to build production RAG i'd vouch for our toolkit 🙂
based on feedback like yours we might build something like this too if you think it helps with full-stack prototyping
SECInsights is a very nice release but is not modular for example to have a separate ingestion and use with a database ? instead of importing docs each time from an S3
