Refactor Docker setup and enhance API documentation

[artaasd95]

• Updated docker-compose.yml to rename services and add CLI support.
• Introduced Dockerfiles for API and CLI services.
• Added FastAPI server implementation for the Standard Pipeline.
• Enhanced README.md and api-reference.md with usage instructions for the new REST API.
• Updated requirements.txt and setup.py to include FastAPI and related dependencies.
• Added tests for the API endpoints to ensure functionality.

[molly554]

Thanks for the contribution! Adding a REST API for the Standard Pipeline is a useful feature. However, there are a few issues that need to be addressed before merging.

❌ 2 Test Failures

test_humanize_returns_result and test_humanize_include_steps both return 503 instead of 200.

Root cause: The client fixture sets app.state.config directly, but TestClient(app) triggers the lifespan context manager which reads the real config/config.toml (doesn't exist in CI/test environments) and overwrites app.state.config = None. The fixture assignment happens before TestClient.__enter__, so lifespan clobbers it.

Fix: Either mock _load_config / _validate_config inside the fixture, or use app.dependency_overrides to inject test config, or restructure lifespan to skip validation when config is already set.

⚠️ _validate_config() Hardcodes DeepSeek

if not api_keys.get("deepseek_api_key"):
    missing.append("deepseek_api_key")

This conflicts with PR #16 (OpenRouter support). When a user configures [llm].provider = "openrouter" with openrouter_api_key, this validation will incorrectly report deepseek_api_key as missing.

Fix: Use provider-aware validation — check the key that matches the configured [llm].provider, or reuse resolve_llm_config() from llm_client.py which already handles this correctly.

⚠️ fastapi/uvicorn/pydantic Should Be Optional Dependencies

These 3 packages are added to the base requirements.txt and install_requires, meaning every CLI-only user must install them. The API server is an optional feature.

Suggested fix: Move them to extras_require:

extras_require={
    "api": [
        "fastapi>=0.109.0",
        "uvicorn[standard]>=0.27.0",
        "pydantic>=2.0.0",
    ],
    "legacy": [...],
},

And in api.py / pipeline.py --serve, catch ImportError with a helpful message like "pip install humanize-text[api]".

💡 Minor Suggestions (non-blocking)

1. --serve binds 0.0.0.0 — This exposes the API to the entire local network. Consider defaulting to 127.0.0.1 (localhost only) and adding a --host option for those who need external access.

2. No authentication on /humanize — Anyone who can reach port 8000 can consume the user's DeepSeek/Niutrans API quota. Consider adding at least a simple bearer token check (configurable via env var).

---

To summarize — please fix the 2 failing tests, make config validation provider-aware, and move the API deps to extras. Happy to re-review after those changes. 🙏