1. Story Gen
入prompt 出response
2. Chat history
用KV store , 可能用Redis
3. Prompt Craft
同OpenAI一樣俾人入Customize Instruction
4. Intelligent Agent
用Agent , 加Maths tools / wikepedia search info
5. Streaming Text
加steaming, 唔洗等哂成個result先出response
6. Retrieval-Augmented Generation (RAG)
用vector database (應該用pinecone)
大概用Python, streamlit, redis, pinecone, langchain, hugging face model (llama2),
localhost 搞掂先再試aws bedrock配lamda 果啲
咁樣睇落有無問題?
