nlproxy/server/main.py

"""FastAPI application factory for NLProxy.

Author: IntelliDeep Labs Team
License: BSL 1.1
"""

from __future__ import annotations

from contextlib import asynccontextmanager

from fastapi import FastAPI
from prometheus_fastapi_instrumentator import Instrumentator, metrics

from nlproxy.server.config import settings
from nlproxy.server.logger import setup_logging
from nlproxy.server.middleware import configure_middlewares, configure_request_id
from nlproxy.server.dependencies import startup, shutdown
from nlproxy.server.apis.health import router as health_router
from nlproxy.server.apis.models import router as models_router
from nlproxy.server.apis.chat import router as chat_router
from nlproxy.server.apis.errors import register_exception_handlers


def create_app() -> FastAPI:
    setup_logging(settings.log_level)

    @asynccontextmanager
    async def lifespan(app: FastAPI):
        await startup()
        try:
            yield
        finally:
            await shutdown()

    app = FastAPI(
        title="nlproxy Enterprise Proxy",
        description="Production-grade prompt compression and safety proxy for LLMs",
        version="1.0.0",
        docs_url="/docs",
        redoc_url="/redoc",
        openapi_url="/openapi.json",
        lifespan=lifespan,
    )

    configure_middlewares(app)
    configure_request_id(app)

    app.include_router(health_router)
    app.include_router(models_router)
    app.include_router(chat_router)

    register_exception_handlers(app)

    if settings.enable_metrics:
        Instrumentator(
            should_group_status_codes=False,
            should_ignore_untemplated=True,
            should_respect_env_var=True,
            env_var_name="nlproxy_ENABLE_METRICS",
        ).add(
            metrics.requests(metric_namespace="nlproxy", metric_subsystem="api"),
            metrics.request_size(metric_namespace="nlproxy", metric_subsystem="api"),
            metrics.response_size(metric_namespace="nlproxy", metric_subsystem="api"),
            metrics.latency(
                buckets=(0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1.0, 2.5, 5.0, 10.0),
                metric_namespace="nlproxy",
                metric_subsystem="api",
            ),
        ).instrument(app).expose(
            app,
            endpoint=settings.metrics_path,
            include_in_schema=False,
        )

    return app


app = create_app()