fix: Improve voice noise tolerance and focus on latest message

- Raise VAD thresholds (activation 0.65→0.75, min speech 0.4→0.6s,
  min silence 0.55→0.65s) to reduce false triggers from background noise
- Add "focus on latest message" instruction to all prompts (voice + text)
- Add "greet and wait" behavior for new conversations instead of auto-continuing

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Christian Gick
2026-03-04 13:30:14 +02:00
parent 90cdc7b812
commit 1000891a97
3 changed files with 11 additions and 4 deletions

2
bot.py
View File

@@ -94,6 +94,8 @@ IMPORTANT RULES — FOLLOW THESE STRICTLY:
- If no relevant documents were found, simply say you don't have information on that topic and ask if you can help with something else. Do NOT speculate about why or suggest the user look elsewhere.
- You can see and analyze images that users send. Describe what you see when asked about an image.
- You can read and analyze PDF documents that users send. Summarize content and answer questions about them.
- Always focus on the user's most recent message — whether it was text or voice. Do not automatically continue or summarize previous conversations.
- When a user greets you or starts a new conversation after a pause, respond with a brief greeting and wait for their instructions.
- You can generate images when asked — use the generate_image tool for any image creation, drawing, or illustration requests.
- You can search the web using the web_search tool. Use it when users ask about current events, facts, or anything that needs up-to-date information.
- You can open and read web pages using browse_url. Use it when a user shares a link, or when you need more detail from a search result. Summarize the key content concisely.