LlamaIndex

“LlamaIndex - and specifically LlamaParse - is exceptionally good at transforming complex, unstructured business data into clean, machine-readable formats. When managing multi-channel D2C performance metrics, it effortlessly converts heavy spreadsheet exports (like our campaign broadcast CSVs and multi-tab .xlsx files) into structured Markdown. It perfectly preserves the structural context of rows and columns, meaning critical metrics like click rates, conversions, and ROI are kept intact for dow”

“https://github.com/cklapperich/DocumentContextExtractor Anthropic made a blog post about using contextual retrieval to get super-accurate RAG. They also gave a python notebook to demonstrate. llama index ALSO implemented a demo here Motivation: Neither handles the many edge cases trying to replicate this in the real world with over 100s of documents. (rate limits, cost, documents too large for context window, prompt caching doesn't work via llama_index interface, error handling, chunk + context ”

“In my work evaluating AI frameworks and LLM pipelines, LlamaIndex stands out for making Retrieval-Augmented Generation (RAG) implementation fast and seamless. Its data connectors (LlamaHub) make it easy to ingest custom documents such as PDFs and Markdown files. The Python and TypeScript SDKs integrate naturally into typical developer workflows, and onboarding is quick thanks to the excellent documentation. I also appreciate the advanced indexing strategies, which help deliver strong retrieval p”

“I've spent the past 8 months the trenches, I want to share what actually worked vs. wasted our time. We built RAG for Usul AI (9M pages) and an unnamed legal AI enterprise (4M pages). Langchain + Llamaindex We started out with youtube tutorials. First Langchain -> Llamaindex. Got to a working prototype in a couple of days and were optimistic with the progress. We run tests on subset of the data (100 documents) and the results looked great. We spend the next few days running the pipeline on the p”

“hey, I am new to LlamaIndex too. imo, it saves you time and lines of code with boiler plate for what kind of pre-processing, chunking, retrievings you want to use. Regarding speed and accuracy comparision, it depends on how you do it manually as you mentioned.”

“I switched from LlamaIndex to Langchain, primarily because Langchain supports more back ends and LlamaIndex is even more abstracted to the point where it gets to be a real PITA to go beyond the "normal" way to do things with RAG. Langchain abstracts a lot, and in many ways feels like it's in a constant beta. For example Langserve is a great way to quickly throw up a web endpoint for your Langchain app, but it doesn't provide for OpenAI API endpoints which nearly every nice web UI supports. Man”

“The title says it all. This library just is not consistently built. Here's an example - I want to use my own embedding model with VectorStoreIndex that I pass in without using Settings. The only way this actually works is with the following code: vector_index = VectorStoreIndex( nodes=nodes, embed_model=embedding_model, storage_context=storage_context, show_progress=verbose, ) Then VectorStoreIndex has a method called build_index_from_nodes, which makes you think that you could/should initialize”

“The API is also changing fairly often and the documentation is less than clear on many aspects I had to look at the source code many times to understand what is available and how it works...”