🗃️ All relevant documents from the database are stuffed into the prompt
🥳 This week, we published a demo on combining @LangChainAI and #Weaviate
📚 https://t.co/sWAGGbvUIN
📹 Or watch @ecardenas300's demo: https://t.co/GVRcPhwyQU
🧵 Check out the 4 CombineDocument techniques implemented in #LangChain that solve the limited input token lengths👇
🗃️ All relevant documents from the database are stuffed into the prompt
✨ Another prompt is created to combine all of the initial outputs into one
🧠 It then uses its local memory with the summaries generated so far to influence the next output
🔄 This is repeated until all documents have been processed
💯 The language model then assigns a score based on the certainty of the answer and ranks them
💬 The top two documents are stuffed into the model to generate a single response