🐢 LLMs can be slow. Working on a recommender solution, beautiful output but oh my, the latency of API batch!

🚀 Step in API response streaming for reducing perceived latency, as the user sees near real-time LLM progress. Just like the #chatgpt you know.

🛠️ Knocked this streaming demo app up. A basic Streamlit front-end, with #FastAPI (backend), LangChain (LLM orchestration), and #Azure OpenAI (LLM model) all configured with callback handlers and streaming enabled to send partial message deltas back to the client via WebSocket.

🥇 Looking forward to baking this into my recommender for a better customer experience!

📝 Demo backend is a boiled-down version of main py in Langchain’s chat-langchain repo.

📦 Demo app code here.