[DennyHo0917]

Summary

Adds API Route as an AI model and embeddings provider alongside Gemini, OpenAI, Claude, Grok, Ollama, OpenRouter, and MiniMax.

API Route provides high-speed, reliable OpenAI-compatible API routing:

• Default model: gpt-5.4-mini — fast, high-capability inference for study note generation, quizzes, and chat
• Base URL: https://www.api-route.com/v1 (configurable via API_ROUTE_BASE_URL)

Changes

• New file: backend/src/utils/llm/models/apiroute.ts — API Route provider using ChatOpenAI and OpenAIEmbeddings from @langchain/openai
• Updated: backend/src/utils/llm/models/index.ts — added apiroute (and api_route, api-route aliases) to the model factory pick() switch
• Updated: backend/src/config/env.ts — added API_ROUTE_API_KEY, API_ROUTE_MODEL, and API_ROUTE_BASE_URL configuration variables
• Updated: .env.example — added API Route configuration block and updated LLM_PROVIDER documentation
• Updated: README.md — added API Route to Supported AI Models list
• New tests:
  • backend/src/utils/llm/models/__tests__/apiroute.test.ts — 13 unit tests covering model selection, custom baseURL, temperature clamping, token limits, and API key fallback
  • backend/src/utils/llm/models/__tests__/apiroute-factory.test.ts — 4 integration tests verifying provider selection, alias matching, embeddings fallback, and isolation

How it works

API Route is fully compatible with OpenAI API specifications, allowing seamless reuse of ChatOpenAI and OpenAIEmbeddings from @langchain/openai. Temperature is clamped to [0, 2].

Configuration

LLM_PROVIDER=apiroute
API_ROUTE_API_KEY=your-api-key
API_ROUTE_MODEL=gpt-5.4-mini

Test plan

• 13 unit tests for makeLLM() and makeEmbeddings() in apiroute.test.ts
• 4 factory integration tests in apiroute-factory.test.ts
• Full test suite verification (npx vitest run: 32 passed, 0 failed)
• Backend TypeScript build check (npm run build / tsc -p backend/tsconfig.json: 0 errors)

[nullure]

LGTM