AI Normalization — Configuration

Config file

ai:
  api_key: "sk-..."
  model: "gpt-4.1-mini"       # default
  # provider: "openai"         # default
  # endpoint: ""               # custom OpenAI-compatible endpoint
  # max_batch: 0               # items per API call (0 = auto)
  # max_concurrency: 0         # parallel API calls (0 = auto)
  # proxy: ""                  # HTTP proxy for AI API calls

Environment variable fallback

If ai.api_key is not set in the config, bbscope falls back to the OPENAI_API_KEY environment variable.

For the web server, use:

OPENAI_API_KEY=sk-...
OPENAI_MODEL=gpt-4.1-mini

Usage

CLI

bbscope poll --db --ai

The --ai flag only works with --db. Without a database, there’s nowhere to store the normalized results.

Web server

If OPENAI_API_KEY is set, AI normalization is automatically enabled for the background poller. Check the /debug page to confirm.

Custom endpoints

For self-hosted or alternative OpenAI-compatible APIs (e.g., Ollama, vLLM, Azure OpenAI), set the endpoint:

ai:
  api_key: "your-key"
  endpoint: "http://localhost:11434/v1"
  model: "llama3"

Cost considerations

• Only new or changed targets are sent to the API. Previously normalized targets are loaded from the database cache.
• The default model (gpt-4.1-mini) is cost-effective for this use case.
• Batch size and concurrency can be tuned with max_batch and max_concurrency if needed.