🐛 Fix conversation summary prompt to use user query language

[Copilot]

Problem

When saving conversations, the summary generation prompt was instructing the LLM to "use the same language as the system" which was ambiguous and didn't guarantee the summary would be in the same language as the user's query. This caused conversation summaries to be generated in inconsistent languages, especially when users asked questions in languages other than English.

Solution

Updated the conversation summary prompt in save_conversation() function to explicitly reference the user's query language instead of the ambiguous "system" language.

Changes made:

• Changed "use the same language as the system" to "use the same language as the user query" in both assistant and user role messages
• This ensures the LLM has clear, unambiguous instructions to match the user's input language

Before:

"content": "Summarise following conversation in no more than 3 "
"words, respond ONLY with the summary, use the same "
"language as the system",

After:

"content": "Summarise following conversation in no more than 3 "
"words, respond ONLY with the summary, use the same "
"language as the user query",

Testing

• Verified syntax correctness with Python compilation
• Created validation test confirming prompt correctly references user query language
• Confirmed no other similar prompts exist in the codebase
• Minimal change affecting only 2 lines in 1 file

Fixes #1878.

---

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
nextra-docsgpt	🔄 Building (Inspect)	Visit Preview	💬 Add feedback	Jul 15, 2025 9:35am
oss-docsgpt	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 15, 2025 9:35am