Bob123Yang

Not yet. I found it related to Top N value setting. The original Top N max = 30, and I change it to 1000 and rebuild the frontend and the...

But now, once topN>30, the retrieval content is empty (click the light bulb in the Chat) and the LLM will answer "nothing found" in the knowledge base instead of many...

@dosu There is the "message" module names as Response in the agent that is the "final" module, right? Add the json file of my agent, please help review it and...

@dosu Attach my agent JSON for your review and tell me how to resolve the issue - no final answer for each user query. [Multi-Agent Customer Support.json](https://github.com/user-attachments/files/23350212/Multi-Agent.Customer.Support.json)

@dosu Help me review the below picture for the Agent log after the first query "hello" and the current output only has the thinking chain without the final answer. You...

@dosu I created the agent with the template as the below picture with red circle and only change the LLM (=Qwen3-30B-Thinking) and database setting and keep all the default prompt...

Same prompt and setting for all module of the agent: 1. only change LLM to Qwen3-30B-Thinking that will not output the final answer and only output the thinking chain. ----...

@Magicbook1108 For V0.22.1 agent, if I select Qwen3-30B-Thinking, still cannot get the final answer and only get the chain of thinking; if I select Qwen3-30B-Instruct and keep all other setting...

Xinference latest version 14

Pull nightly version and launch it and find a lot of error in the log as below: test@test-test-Product:~/ragflow/ragflow-0.22.1/ragflow$ docker logs -f docker-ragflow-cpu-1 [docling] disabled by USE_DOCLING [mineru] disabled by USE_MINERU...