[Feature]: Add Fireworks AI as a built-in provider

[willzhqiang]

Feature Request

Please add Fireworks AI as a first-class provider.

Fireworks offers FirePass — a free tier that gives access to latest reasoning models. I tried using the OpenAI-compatible mode, but the latest FirePass models are not in the model list, so there's no way to configure them properly. Native Fireworks provider support would solve this.

Official API info from Fireworks docs:

• Model ID: accounts/fireworks/routers/kimi-k2p5-turbo
• API Key: standard Fireworks API key (generated at app.fireworks.ai/api-keys)
• Base URL (OpenAI-compatible): https://api.fireworks.ai/inference/v1
• Base URL (Anthropic-compatible): https://api.fireworks.ai/inference

[majiayu000]

Hi, I've been looking into this and traced the root cause.

Proposed approach: Follow PR #2687 (Novita AI) pattern exactly: (1) Add FIREWORKS ProviderId const in crates/forge_domain/src/provider.rs and add to built_in_providers() list, (2) Add Fireworks AI entry to crates/forge_repo/src/provider/provider.json with id=fireworks, api_key_vars=FIREWORKS_API_KEY, response_type=OpenAI, url=https://api.fireworks.ai/inference/v1/chat/completions, models endpoint. Include the FirePass models mentioned in the issue (e.g. accounts/fireworks/routers/kimi-k2p5-turbo).

I'll open a draft PR shortly. Happy to adjust based on your preference.