We didn’t have a great UX for building a data chatbot in LlamaIndex 🦙 (its optimized for QA against your data). That’s changed now; you can now build a chatbot over your data in ~4 lines of code! ⚡️🤖 1️⃣ ingest data 2️⃣ index data 3️⃣ chatbot from index... 4️⃣ start using it ?!show more

We offer this chatbot in three main modes, with REPL also included: - ReAct agent with @LangChainAI (we already had a tutorial for this, but now its *way* simpler) - Simple Chat (stores conversational memory) - Condense Question Chat (generates a new question at each step)

Chat engines are *stateful*; our query engines are stateless. The current abstractions we have are light; for instance, the memory abstractions we have are very simple. This gives us the foundation to build out memory modules and + also use other modules etc from langchain

Check out our brand-new chat engine guide here 📗. Contains a bunch of sample notebooks you can try running yourself 🛠️.

@llama_index Wow very nice Jerry and to the Llama team!

@llama_index This is so good

@llama_index What formats can it ingest?

@llama_index Does it support streaming?

@llama_index that should be coming in the next day or two :)

@llama_index so cool

@llama_index it's too hard to use! bug everywhere