search

infiniflow / ragflow

RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding.
https://ragflow.io
Apache License 2.0
16.98k stars 1.73k forks source link

[Question]: bad escape issue #2079

Open cotaa956 opened 3 weeks ago

cotaa956 commented 3 weeks ago

Describe your problem

In demo ragflow everytime I use the agent templates and run it, the first response goes well but later responses output is " bad escape" Screenshot_20240823-220135_Chrome

guoyuhao2330 commented 2 weeks ago

image

Sorry I can't reproduce this bug, can you provide more details?

jjohare commented 2 weeks ago

I get this a lot TBH. Seems like it's a ChatGPT 4o mini issue, (for me) in SOME models of accuracy of response, but not others. From the logs. {"0": "bad escape \u at position 7884 (line 105, column 34)"}}}}, "downstream": ["Answer:PoorMapsCover"], "upstream": ["DuckDuckGo:SoftButtonsRefuse", "Wikipedia:WittyRiceLearn",

context

image

cotaa956 commented 2 weeks ago

image

Sorry I can't reproduce this bug, can you provide more details?

Okay I will explain that I used the websearchassistant as a template agent and also the chatbot, the llm I used was deepseekchat and gemini .. whenever I ask a question to be retrieved from the knowledge base the llm response is " bad escape \m" like "according to the knowledge base, what questions can I ask" or something like that "what are the functions of the kidney according to the knowledge base " Here are the images for illustration image_2024-08-27_04-56-43 image_2024-08-27_04-57-22 image_2024-08-27_04-57-09