[Question]: Content Length Limit Reached Unexpectedly in RAGFlow

[deulofeu1]

Describe your problem

Hello,

I am experiencing an issue with my dialogue setup involving a RAGFlow dialogue assistant and a local dialogue model. Despite configuring both models to allow for maximum token generation (max_tokens set to 128,000 for RAGFlow and 16,384 for the local model), the models terminate early after generating approximately 185 words or about 1,100 characters.

Here are some additional details of my setup:

RAGFlow Model: max_tokens is configured at 128,000.
Local Dialogue Model: max_tokens is set to 16,384.
Embedding Model: Deployed locally as text-embedding-bge-m3.
Context Length for Embedding Model: Set at 2048 tokens.
The generation stops with the message: "For content length reason, it stopped, continue?" This occurs despite the relatively short output. Could this issue be related to my embedding model configuration or other parameters? I am looking for guidance on how to resolve this premature termination due to supposed length constraints.

Thank you in advance for your support and advice.

[KevinHuSh]

You could disable max tokens in dialog setting.

[deulofeu1]

You could disable max tokens in dialog setting.您可以在对话框设置中禁用 max tokens 。

Yep this resolves the issue. I think the MAX_Token parameter seems to have some problems here.