3
[Feature Request] Preserve conversation context in AI Chat / RAG
Source: paperless-ngx/paperless-ngx#14037 · opened by @JP95Git
Description The current AI chat appears to perform document retrieval based only on the current user question. Previous questions and answers in the conversation are not taken into account when searching for relevant documents. This causes problems with follow-up questions that refer to something mentioned earlier in the conversation. For example: > User: At which company were the carpets cleaned? > Assistant: The carpets were cleaned by BrightClean Carpet Services Ltd. The user then asks: > User: What was the invoice amount? The second question is searched independently. Since the previous conversation context is not considered, the system may retrieve an unrelated document containing an invoice amount. In my example, the correct answer would be 300 €, which was the amount of the carpet-cleaning invoice. Instead, the system returned 201 €, which was actually the amount of an unrelated dentist invoice. Expected behavior The chat should retain the …
No pledges yet. Be the first to back this.
Comments
Similar requests
[Feature Request] Open the AI document chat as a full page/tab instead of a small overlay window
6 votes · 0 comments
[Feature Request] Add detailed debug logging for RAG/LLM requests
2 votes · 0 comments
[Feature Request] Optimize AI suggestion generation
5 votes · 0 comments
[Feature Request] AI chat/RAG: cache the embedding model instead of reloading it on every query
1 vote · 0 comments
[Feature Request] AI chat: size the synthesis prompts by context window, not embedding chunk size
3 votes · 0 comments
No comments yet.