fix: add fireworks-ai-firepass provider with kimi-k2p6-turbo model

[kassieclaire]

Summary

Add a dedicated fireworks-ai-firepass provider to separate Firepass router models from the standard fireworks-ai provider. This replaces the previous approach of embedding firepass models directly within the fireworks-ai provider.

Changes

New Provider: fireworks-ai-firepass

• Added FIREWORKS_AI_FIREPASS ProviderId constant ("fireworks-ai-firepass") with display name "FireworksAIFirepass"
• Registered in built_in_providers() list and FromStr mappings
• New provider entry in provider.json sharing the same API URL (https://api.fireworks.ai/inference/v1/chat/completions) and FIREWORKS_AI_API_KEY env var as the existing fireworks-ai provider

New Model: Kimi K2.6 Turbo (firepass)

• Model ID: accounts/fireworks/routers/kimi-k2p6-turbo
• Display name: Kimi K2.6 Turbo (firepass)
• 262K context length
• Supports: tool use, parallel tool calls, reasoning, text + image input modalities
• Based on the existing k2p6-turbo configuration

Removed Deprecated Models

• Removed kimi-k2p5-turbo (Kimi K2.5 Turbo Firepass) from the fireworks-ai provider
• Removed kimi-k2p6-turbo (Kimi K2.6 Turbo Fire Pass) from the fireworks-ai provider

Pipeline Transformers

• Extended ReasoningContent transformer to apply to FIREWORKS_AI_FIREPASS
• Extended EnforceStrictToolSchema + EnforceStrictResponseFormatSchema pipeline to apply to FIREWORKS_AI_FIREPASS

Tests

• Added 3 unit tests: from_str, display_name, in_built_in_providers

Files Changed

File	Change
crates/forge_domain/src/provider.rs	New provider constant, mappings, and tests (+21)
crates/forge_repo/src/provider/provider.json	Removed firepass models from fireworks-ai, added new provider entry (+20/-20)
crates/forge_app/src/dto/openai/transformers/pipeline.rs	Extended transformers to new provider (+2)

Verification

• cargo check passes
• All 590 tests pass in forge_domain + forge_app
• All 317 tests pass in forge_repo

[Copilot]

Pull request overview

This PR introduces a dedicated fireworks-ai-firepass provider to separate Firepass router-backed models from the existing fireworks-ai provider, and wires the new provider into the OpenAI-request transformation pipeline.

Changes:

• Added ProviderId::FIREWORKS_AI_FIREPASS with FromStr, display name, and built-in provider registration (plus unit tests).
• Updated provider.json to remove Firepass router models from fireworks-ai and add a new fireworks-ai-firepass provider containing accounts/fireworks/routers/kimi-k2p6-turbo.
• Extended request transformers (ReasoningContent + strict schema enforcement) to also apply to FIREWORKS_AI_FIREPASS.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
crates/forge_domain/src/provider.rs	Adds new provider ID constant, display mapping, parsing, and tests.
crates/forge_repo/src/provider/provider.json	Moves Firepass router model(s) under a new provider entry.
crates/forge_app/src/dto/openai/transformers/pipeline.rs	Applies existing transformer behaviors to the new provider ID.

Comments suppressed due to low confidence (1)

crates/forge_app/src/dto/openai/transformers/pipeline.rs:107

• ProviderId::FIREWORKS_AI_FIREPASS is now included in the strict schema enforcement condition, but there’s no unit test verifying this provider gets the same tool/response-format schema normalization guarantees as fireworks-ai. Add a test similar to test_fireworks_provider_enforces_strict_tool_and_response_format_schemas for fireworks-ai-firepass.

        let strict_schema = EnforceStrictToolSchema
            .pipe(EnforceStrictResponseFormatSchema)
            .when(move |_| {
                provider.id == ProviderId::FIREWORKS_AI
                    || provider.id == ProviderId::FIREWORKS_AI_FIREPASS

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CLAassistant]

All committers have signed the CLA.