langchain
langchain copied to clipboard
Azure Content Filtering on Default QA with sources prompt.
System Info
v0.0.176
Who can help?
@hwchase17 @agola11
Information
- [ ] The official example notebooks/scripts
- [X] My own modified scripts
Related Components
- [ ] LLMs/Chat Models
- [ ] Embedding Models
- [ ] Prompts / Prompt Templates / Prompt Selectors
- [ ] Output Parsers
- [ ] Document Loaders
- [ ] Vector Stores / Retrievers
- [ ] Memory
- [ ] Agents / Agent Executors
- [ ] Tools / Toolkits
- [X] Chains
- [ ] Callbacks/Tracing
- [ ] Async
Reproduction
The default QA with Sources Chain: langchain/chains/qa_with_sources/stuff_prompt.py
causes issues with Azure Content Filter, which will return the following:
"error": {
"message": "The response was filtered",
"type": null,
"param": "prompt",
"code": "content_filter",
"status": 400
}
I believe this is due to the sections that mention issues like covid, which is found in the example source content in lines 22-28.
I modified to the following to resolve the issue:
Content: Madam Speaker, Madam Vice President, our First Lady and Second Gentleman. Members of Congress and the Cabinet. Justices of the Supreme Court. My fellow Americans.
Source: 0-pl
Content: And we won’t stop.
Source: 24-pl
Content: And a proud Ukrainian people, I want you to know that we are going to be okay.
Source: 5-pl
Content: More support for patients and families. Our future is forged. \n\nWell I know this nation.
Source: 34-pl
this is probably not the best version of the source content, but it worked for me.
Expected behavior
The QA prompt should not trigger Azure Content Filtering.