AI provider
Connect any OpenAI-compatible endpoint to score names and write summaries.
Radar uses an AI model for four things: scoring whether a term is the name of one specific product (0–1), summarising Reddit asks, writing a one-line description for high-scoring terms, and summarising search results. Everything else works without it.
Setup
- Open Settings → AI.
- Enter the Base URL, API key and Model.
- Click Save and test connection. Chrome asks for permission to reach that host the first time.
Examples
| Provider | Base URL |
|---|---|
| DeepSeek (default) | https://api.deepseek.com |
| OpenAI | https://api.openai.com/v1 |
| Local (Ollama, LM Studio…) | http://localhost:11434/v1 |
Any endpoint that speaks the OpenAI Chat Completions format works.
Minimum score
Terms scored below the minimum (0.5 by default) go to the Low list. Terms scored 0.7 or higher land in Worth.
Cost and failures
- Only terms that have never been scored are sent, 25 per request; scores are cached.
- A failed batch is retried once and then skipped. A 429 waits for
retry-after. - If the provider is down the report is still produced, just without scores.
Your key stays local
The key is stored in
chrome.storage.local, never synced, and only sent to the endpoint you entered.