Provider and Model Configuration

Core fields

API schema

When adding a provider, choose the right API schema:

• OpenAI — For OpenAI and OpenAI-compatible providers (most providers support this)
• Anthropic — For Claude models
• Google — For Gemini models; requires project ID and region
• Others — Choose based on provider docs

OpenAI schema also lets you pick between standard mode (Chat Completions) and the newer Responses API.

Advanced options

• Custom Headers (JSON) — Some providers require extra headers
• Custom Body (JSON) — Add extra fields to the request body

Recommended setup

1. Configure at least two models (primary + backup).
2. Save models separately even under one provider for easy switching.
3. Name models by use case, like "writing-longform" or "coding-fast".

Troubleshooting order

1. Is the API key still valid? (Check in the provider console.)
2. Does the Model ID match exactly? (Copy it directly.)
3. Is the Base URL correct? (No extra path or wrong protocol.)
4. Is a proxy or firewall blocking requests?

Parameter tips

• Temperature: Lower for factual answers, higher for creative writing
• Max Tokens: Too small truncates output, too large increases cost and latency
• Top P: Usually leave at default; don't adjust both Temperature and Top P at once

Security tips

• Don't commit API keys to public repos
• Rotate keys regularly
• If a key leaks, revoke it immediately in the provider console