ChickenBenny commented on issue #144: URL: https://github.com/apache/flink-agents/issues/144#issuecomment-6818883657
Hi, is this still wanted? I'd like to pick it up if nobody's on it. Right now you have to write the RAG flow yourself, like the [example](https://github.com/apache/flink-agents/tree/main/python/flink_agents/examples/rag) does: 1. send a `ContextRetrievalRequestEvent` 2. build the prompt from the returned docs and send a `ChatRequestEvent` 3. turn the `ChatResponseEvent` into the output I'm thinking something like: ```python agent = RAGAgent( chat_model=ResourceDescriptor(...), vector_store="knowledge_base", max_results=3, prompt=None, # optional, default template with {context} / {user_query} output_schema=None, # optional, same as ReActAgent ) ``` A few things I'm not sure about: - **Shape**: a standalone `RAGAgent`, or a retrieval option on `ReActAgent`? I lean towards standalone and keeping ReAct as is. - **Languages**: Java and Python together, or Python first and Java in a follow-up PR? - **Empty results**: if no docs come back, should we still call the LLM, or let the user decide? Happy to write up a short design in Discussions first if that's better. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
