OpenAI-Compatible Endpoints¶
BrrainzBot talks to an OpenAI-style API for the verification decision. The bot sends a short prompt and expects a strict JSON response.
What You Need¶
- a base URL, for example
https://api.openai.com/v1 - a model name
- an API key
Safety Rules¶
BrrainzBot is intentionally conservative:
- it prefers HTTPS endpoints
- it keeps secrets in a separate local file
- it never checks for new versions in the background
- it only contacts GitHub for updates when you explicitly run
self-update
When to Allow Insecure Local Endpoints¶
Only do this if:
- the endpoint is on your own machine or private network
- you understand that the connection is not using standard HTTPS protection
For normal hosted endpoints, keep HTTPS required.
Keeping Costs and Latency Under Control¶
Use a fast small model first. The bot only needs a short judgement, not a long essay.
Good characteristics:
- fast
- cheap
- reliable JSON output
- good enough classification for obvious humans versus obvious junk
What the Bot Sends¶
The bot sends:
- your guild-specific topic prompt
- a short rules hint
- the user’s three answers
- the attempt number
It does not need a long conversation history for v1.