pagesift_api/
server.rs

1//! HTTP server routing: Firecrawl v1/v2 compatibility layers, Jina Reader adapter,
2//! health checks, and Prometheus metrics.
3
4use std::collections::HashMap;
5use std::future::Future;
6use std::pin::{Pin, pin};
7use std::sync::Arc;
8use std::sync::atomic::{AtomicBool, Ordering};
9use std::task::{Context, Poll};
10use std::time::{Duration, Instant};
11
12use axum::body::{Body, Bytes, HttpBody};
13use axum::extract::rejection::JsonRejection;
14use axum::extract::{Request, State};
15use axum::http::{HeaderMap, Method, StatusCode, header};
16use axum::middleware::{self, Next};
17use axum::response::{IntoResponse, Response};
18use axum::routing::get;
19use axum::{Json, Router};
20use http_body::{Frame, SizeHint};
21use hyper_util::rt::{TokioExecutor, TokioIo, TokioTimer};
22use hyper_util::server::conn::auto;
23use hyper_util::server::graceful::GracefulShutdown;
24use hyper_util::service::TowerToHyperService;
25use pagesift_store::Store;
26use tokio::net::TcpListener;
27use tokio::sync::Mutex;
28use tokio::time::Sleep;
29
30use crate::engine::Engine;
31use crate::error::{ApiError, ErrorCode};
32use crate::jobs::Jobs;
33use crate::limits::{Caller, KeyLimits, Limiter};
34use crate::metrics::Metrics;
35use crate::webhook::{WebhookConfig, WebhookSettings, Webhooks};
36
37/// Maximum documents per page when querying job status.
38pub const PAGE_SIZE: usize = 100;
39pub const DEFAULT_BATCH_CONCURRENCY: usize = 8;
40
41/// 服务端超时，防 slowloris 和挂死的请求。
42#[derive(Debug, Clone, Copy, PartialEq, Eq)]
43pub struct Timeouts {
44    /// 读完请求头的时限（HTTP/1），超时直接断开连接。只在 [`serve_with`] 里生效。
45    pub header_read: Duration,
46    /// 从请求进来到读完请求体的总时限，不是两次读之间的间隔：每 5 秒滴 1 字节也会超时。
47    pub body: Duration,
48    /// 整个请求（到响应头发出为止）的时限。要比同步 scrape 的 `timeout` 上限大，
49    /// crawl/batch 只是创建任务，很快返回，不受影响。
50    pub request: Duration,
51}
52
53impl Default for Timeouts {
54    fn default() -> Self {
55        Self {
56            header_read: Duration::from_secs(30),
57            body: Duration::from_secs(30),
58            request: Duration::from_millis(crate::validate::MAX_SCRAPE_TIMEOUT_MS)
59                + Duration::from_secs(60),
60        }
61    }
62}
63
64#[derive(Clone, Default)]
65pub struct Config {
66    /// When empty, authentication is disabled; otherwise requires `Authorization: Bearer <key>`.
67    pub api_keys: Vec<String>,
68    /// Public URL for building absolute resource links.
69    pub public_url: Option<String>,
70    pub max_batch_concurrency: Option<usize>,
71    /// Persistent store for jobs and resume capability.
72    pub store: Option<Store>,
73    /// Per-key custom limits.
74    pub key_limits: HashMap<String, KeyLimits>,
75    /// Default limits applied to other keys or unauthenticated callers.
76    pub default_limits: KeyLimits,
77    pub webhook: WebhookSettings,
78    /// 请求体和整请求超时；请求头超时由 [`serve_with`] 的参数决定。
79    pub timeouts: Timeouts,
80}
81
82impl std::fmt::Debug for Config {
83    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
84        f.debug_struct("Config")
85            .field("api_keys", &self.api_keys.len())
86            .field("public_url", &self.public_url)
87            .field("max_batch_concurrency", &self.max_batch_concurrency)
88            .field("store", &self.store.is_some())
89            .field("key_limits", &self.key_limits.len())
90            .field("default_limits", &self.default_limits)
91            .field("webhook", &self.webhook)
92            .field("timeouts", &self.timeouts)
93            .finish()
94    }
95}
96
97#[derive(Clone)]
98pub struct AppState {
99    pub engine: Arc<dyn Engine>,
100    pub keys: Arc<HashMap<String, Arc<KeyLimits>>>,
101    pub default_limits: Arc<KeyLimits>,
102    pub public_url: Option<String>,
103    pub batch_concurrency: usize,
104    pub jobs: Jobs,
105    pub limiter: Arc<Limiter>,
106    pub webhooks: Webhooks,
107    pub extracts: Arc<Mutex<HashMap<String, ExtractJob>>>,
108    pub metrics: Arc<Metrics>,
109}
110
111pub struct ExtractJob {
112    pub status: &'static str,
113    pub data: Option<serde_json::Value>,
114    pub error: Option<ApiError>,
115    pub created: std::time::SystemTime,
116}
117
118/// Construct complete HTTP router with Firecrawl v1/v2, Jina Reader, and metrics.
119pub fn router(engine: Arc<dyn Engine>, config: Config) -> Router {
120    let timeouts = config.timeouts;
121    let default_limits = Arc::new(config.default_limits);
122    let mut keys: HashMap<String, Arc<KeyLimits>> = config
123        .api_keys
124        .into_iter()
125        .map(|k| (k, default_limits.clone()))
126        .collect();
127    keys.extend(config.key_limits.into_iter().map(|(k, l)| (k, Arc::new(l))));
128
129    let accounts: HashMap<String, Arc<KeyLimits>> = keys
130        .iter()
131        .map(|(k, l)| {
132            (
133                Caller::new(Some(k), l.clone()).account.to_string(),
134                l.clone(),
135            )
136        })
137        .collect();
138
139    let metrics = Arc::new(Metrics::default());
140    let limiter = Arc::new(Limiter::new(config.store.clone()));
141    let webhooks = Webhooks::new(config.webhook, metrics.clone());
142    let jobs = Jobs::new(
143        config.store,
144        limiter.clone(),
145        Some(webhooks.clone()),
146        metrics.clone(),
147    );
148
149    if tokio::runtime::Handle::try_current().is_ok() {
150        let fallback = default_limits.clone();
151        jobs.restore(&engine, &|account| {
152            accounts
153                .get(account)
154                .cloned()
155                .unwrap_or_else(|| fallback.clone())
156        });
157    } else {
158        tracing::warn!("router built outside a tokio runtime: saved jobs are not resumed");
159    }
160
161    let state = AppState {
162        engine: engine.clone(),
163        keys: Arc::new(keys),
164        default_limits,
165        public_url: config
166            .public_url
167            .map(|u| u.trim_end_matches('/').to_string()),
168        batch_concurrency: config
169            .max_batch_concurrency
170            .unwrap_or(DEFAULT_BATCH_CONCURRENCY),
171        jobs,
172        limiter,
173        webhooks,
174        extracts: Arc::default(),
175        metrics,
176    };
177
178    let firecrawl = crate::firecrawl::firecrawl_routes()
179        .route_layer(middleware::from_fn_with_state(state.clone(), auth));
180    let jina = crate::jina::jina_router(engine);
181
182    Router::new()
183        .route("/healthz", get(|| async { "ok" }))
184        .route("/metrics", get(render_metrics))
185        .merge(firecrawl)
186        .with_state(state.clone())
187        .merge(jina)
188        .fallback(|| async { ApiError::new(ErrorCode::NotFound, "no such endpoint") })
189        .layer(middleware::from_fn(uniform_errors))
190        .layer(middleware::from_fn_with_state(timeouts, deadline))
191        .layer(middleware::from_fn_with_state(state.clone(), observe))
192}
193
194/// 用默认超时在 `listener` 上提供服务，`shutdown` 完成后优雅退出。
195pub async fn serve(
196    listener: TcpListener,
197    app: Router,
198    shutdown: impl Future<Output = ()> + Send,
199) -> std::io::Result<()> {
200    serve_with(listener, app, Timeouts::default(), shutdown).await
201}
202
203/// 和 `axum::serve(..).with_graceful_shutdown(..)` 行为相同，另外给 HTTP/1 设了请求头读取超时
204/// （axum::serve 设不了）：收到 `shutdown` 后停止 accept，通知每个连接优雅关闭，等它们都结束再返回。
205pub async fn serve_with(
206    listener: TcpListener,
207    app: Router,
208    timeouts: Timeouts,
209    shutdown: impl Future<Output = ()> + Send,
210) -> std::io::Result<()> {
211    let mut builder = auto::Builder::new(TokioExecutor::new());
212    builder
213        .http1()
214        .timer(TokioTimer::new())
215        .header_read_timeout(timeouts.header_read);
216    let graceful = GracefulShutdown::new();
217    let mut shutdown = pin!(shutdown);
218    loop {
219        let stream = tokio::select! {
220            r = listener.accept() => match r {
221                Ok((stream, _)) => stream,
222                Err(e) => {
223                    // 和 axum::serve 一样：单个连接的错误忽略，其他错误（比如文件描述符用完）歇 1 秒再接
224                    if !matches!(
225                        e.kind(),
226                        std::io::ErrorKind::ConnectionRefused
227                            | std::io::ErrorKind::ConnectionAborted
228                            | std::io::ErrorKind::ConnectionReset
229                    ) {
230                        tracing::error!(error = %e, "accept failed");
231                        tokio::time::sleep(Duration::from_secs(1)).await;
232                    }
233                    continue;
234                }
235            },
236            () = &mut shutdown => break,
237        };
238        // 响应头和分块正文分两次写时，Nagle 会让第二次写等对端的延迟确认（约 40 ms）
239        let _ = stream.set_nodelay(true);
240        let conn = builder
241            .serve_connection_with_upgrades(
242                TokioIo::new(stream),
243                TowerToHyperService::new(app.clone()),
244            )
245            .into_owned();
246        let conn = graceful.watch(conn);
247        tokio::spawn(async move {
248            if let Err(e) = conn.await {
249                tracing::trace!(error = %e, "connection ended with an error");
250            }
251        });
252    }
253    drop(listener);
254    graceful.shutdown().await;
255    Ok(())
256}
257
258/// 请求体加总时限；整个请求加时限。超时都回统一的 JSON 错误（408 TIMEOUT）。
259async fn deadline(State(t): State<Timeouts>, req: Request, next: Next) -> Response {
260    let fired = Arc::new(AtomicBool::new(false));
261    let req = if req.body().is_end_stream() {
262        req
263    } else {
264        let fired = fired.clone();
265        req.map(|inner| {
266            Body::new(DeadlineBody {
267                inner,
268                sleep: Box::pin(tokio::time::sleep(t.body)),
269                fired,
270            })
271        })
272    };
273    match tokio::time::timeout(t.request, next.run(req)).await {
274        // 读请求体超时时提取器给的是 400，这里换成 408
275        Ok(_) if fired.load(Ordering::Relaxed) => ApiError::new(
276            ErrorCode::Timeout,
277            format!(
278                "request body was not received within {} s",
279                t.body.as_secs_f64()
280            ),
281        )
282        .into_response(),
283        Ok(resp) => resp,
284        Err(_) => ApiError::new(
285            ErrorCode::Timeout,
286            format!(
287                "request did not complete within {} s",
288                t.request.as_secs_f64()
289            ),
290        )
291        .into_response(),
292    }
293}
294
295/// 到点还没读完就报错的请求体。
296struct DeadlineBody {
297    inner: Body,
298    sleep: Pin<Box<Sleep>>,
299    fired: Arc<AtomicBool>,
300}
301
302impl HttpBody for DeadlineBody {
303    type Data = Bytes;
304    type Error = axum::Error;
305
306    fn poll_frame(
307        self: Pin<&mut Self>,
308        cx: &mut Context<'_>,
309    ) -> Poll<Option<Result<Frame<Bytes>, axum::Error>>> {
310        let this = self.get_mut();
311        if let Poll::Ready(frame) = Pin::new(&mut this.inner).poll_frame(cx) {
312            return Poll::Ready(frame);
313        }
314        if this.sleep.as_mut().poll(cx).is_ready() {
315            this.fired.store(true, Ordering::Relaxed);
316            return Poll::Ready(Some(Err(axum::Error::new("request body timed out"))));
317        }
318        Poll::Pending
319    }
320
321    fn is_end_stream(&self) -> bool {
322        self.inner.is_end_stream()
323    }
324
325    fn size_hint(&self) -> SizeHint {
326        self.inner.size_hint()
327    }
328}
329
330/// axum 自带的拒绝（查询参数、路径参数、JSON 解析失败、405 等）是纯文本或空 body，
331/// 统一换成 `{success: false, code, error}`，状态码和 `Allow` 等响应头不变。
332async fn uniform_errors(req: Request, next: Next) -> Response {
333    let resp = next.run(req).await;
334    let status = resp.status();
335    // 自己的错误都带 ErrorCode 扩展，不用动
336    if !status.is_client_error() || resp.extensions().get::<ErrorCode>().is_some() {
337        return resp;
338    }
339    let plain = resp
340        .headers()
341        .get(header::CONTENT_TYPE)
342        .is_none_or(|v| v.as_bytes().starts_with(b"text/plain"));
343    if !plain {
344        return resp;
345    }
346    let (parts, body) = resp.into_parts();
347    let text = axum::body::to_bytes(body, 64 * 1024)
348        .await
349        .map(|b| String::from_utf8_lossy(&b).trim().to_string())
350        .unwrap_or_default();
351    let code = match status {
352        StatusCode::NOT_FOUND => ErrorCode::NotFound,
353        StatusCode::PAYLOAD_TOO_LARGE => ErrorCode::TooLarge,
354        StatusCode::UNAUTHORIZED => ErrorCode::Unauthorized,
355        _ => ErrorCode::InvalidRequest,
356    };
357    let message = if text.is_empty() {
358        status
359            .canonical_reason()
360            .unwrap_or("bad request")
361            .to_lowercase()
362    } else {
363        text
364    };
365    let mut out = ApiError::new(code, message).into_response();
366    *out.status_mut() = status;
367    for (name, value) in &parts.headers {
368        if name != header::CONTENT_TYPE && name != header::CONTENT_LENGTH {
369            out.headers_mut().append(name, value.clone());
370        }
371    }
372    out
373}
374
375/// Authentication and rate limiting middleware.
376async fn auth(
377    State(s): State<AppState>,
378    headers: HeaderMap,
379    mut req: Request,
380    next: Next,
381) -> Response {
382    let caller = if s.keys.is_empty() {
383        Caller::new(None, s.default_limits.clone())
384    } else {
385        let key = headers
386            .get(header::AUTHORIZATION)
387            .and_then(|v| v.to_str().ok())
388            .and_then(|v| v.strip_prefix("Bearer "))
389            .map(str::trim);
390        match key.and_then(|k| s.keys.get(k).map(|l| (k, l))) {
391            Some((k, l)) => Caller::new(Some(k), l.clone()),
392            None => {
393                return ApiError::new(ErrorCode::Unauthorized, "missing or invalid API key")
394                    .into_response();
395            }
396        }
397    };
398    if req.method() == Method::POST
399        && let Err(e) = s.limiter.check_rate(&caller)
400    {
401        let mut resp = e.into_response();
402        resp.extensions_mut().insert(caller);
403        return resp;
404    }
405    req.extensions_mut().insert(caller.clone());
406    let mut resp = next.run(req).await;
407    resp.extensions_mut().insert(caller);
408    resp
409}
410
411/// Metrics and structured request tracing middleware.
412async fn observe(State(s): State<AppState>, req: Request, next: Next) -> Response {
413    let route = req
414        .extensions()
415        .get::<axum::extract::MatchedPath>()
416        .map(|p| p.as_str().to_string())
417        .unwrap_or_else(|| "unmatched".into());
418    let method = req.method().clone();
419    let started = Instant::now();
420    s.metrics.request_started();
421    let resp = next.run(req).await;
422    let elapsed = started.elapsed();
423    let status = resp.status().as_u16();
424    s.metrics.request_finished(&route, status, elapsed);
425    let code = resp.extensions().get::<ErrorCode>().map(|c| c.as_str());
426    if let Some(code) = code {
427        s.metrics.error(code, "http");
428    }
429    let account = resp.extensions().get::<Caller>().map(|c| c.account.clone());
430    let latency_ms = elapsed.as_secs_f64() * 1000.0;
431    if route.starts_with("/v2/") || route.starts_with("/v1/") || route.starts_with("/r") {
432        tracing::info!(
433            target: "pagesift::http",
434            %method, %route, status, latency_ms, account = account.as_deref(), code,
435            "request"
436        );
437    } else {
438        tracing::debug!(target: "pagesift::http", %method, %route, status, latency_ms, "request");
439    }
440    resp
441}
442
443pub fn body<T>(r: Result<Json<T>, JsonRejection>) -> Result<T, ApiError> {
444    r.map(|Json(v)| v)
445        .map_err(|e| ApiError::invalid(format!("invalid request body: {}", e.body_text())))
446}
447
448pub fn base_url(s: &AppState, headers: &HeaderMap) -> String {
449    if let Some(u) = &s.public_url {
450        return u.clone();
451    }
452    let host = headers
453        .get(header::HOST)
454        .and_then(|v| v.to_str().ok())
455        .unwrap_or("localhost:3002");
456    format!("http://{host}")
457}
458
459pub fn webhook(
460    s: &AppState,
461    v: Option<&serde_json::Value>,
462) -> Result<Option<WebhookConfig>, ApiError> {
463    v.filter(|v| !v.is_null())
464        .map(|v| WebhookConfig::parse(v, s.webhooks.ssrf()))
465        .transpose()
466}
467
468async fn render_metrics(State(s): State<AppState>) -> impl IntoResponse {
469    let out = s.metrics.render(&s.jobs.counts().await);
470    (
471        StatusCode::OK,
472        [(header::CONTENT_TYPE, "text/plain; version=0.0.4")],
473        out,
474    )
475}
pagesift_api/server.rs

pagesift_api/
server.rs
