webclaw/crates/webclaw-server/src/state.rs

157 lines
6.2 KiB
Rust
Raw Normal View History

//! Shared application state. Cheap to clone via Arc; held by the axum
//! Router for the life of the process.
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
//!
//! Two unrelated keys get carried here:
//!
//! 1. [`AppState::api_key`] — the **bearer token clients must present**
//! to call this server. Set via `WEBCLAW_API_KEY` / `--api-key`.
//! Unset = open mode.
//! 2. The inner [`webclaw_fetch::cloud::CloudClient`] (if any) — our
//! **outbound** credential for api.webclaw.io, used by extractors
//! that escalate on antibot. Set via `WEBCLAW_CLOUD_API_KEY`.
//! Unset = hard-site extractors return a "set WEBCLAW_CLOUD_API_KEY"
//! error with a signup link.
//!
//! Different variables on purpose: conflating the two means operators
//! who want their server behind an auth token can't also enable cloud
//! fallback, and vice versa.
use std::sync::Arc;
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
use tracing::info;
use webclaw_fetch::cloud::CloudClient;
use webclaw_fetch::{BrowserProfile, FetchClient, FetchConfig};
use webclaw_llm::ProviderChain;
/// Single-process state shared across all request handlers.
#[derive(Clone)]
pub struct AppState {
inner: Arc<Inner>,
}
struct Inner {
/// Wrapped in `Arc` because `fetch_and_extract_batch_with_options`
/// (used by the /v1/batch handler) takes `self: &Arc<Self>` so it
/// can clone the client into spawned tasks. The single-call handlers
/// auto-deref `&Arc<FetchClient>` -> `&FetchClient`, so this costs
/// them nothing.
pub fetch: Arc<FetchClient>,
/// The exact [`FetchConfig`] the shared `fetch` client was built from.
/// Endpoints that spin up their own client (e.g. `/v1/crawl`, which
/// builds a `Crawler` with its own internal `FetchClient`) clone this
/// so they inherit the same browser profile / proxy / timeout instead
/// of silently falling back to `FetchConfig::default()` (Chrome).
pub fetch_config: FetchConfig,
/// LLM provider chain (Ollama -> OpenAI -> Anthropic), built once at
/// startup. `/v1/extract` and `/v1/summarize` borrow this instead of
/// rebuilding the chain (and re-probing Ollama) on every request.
pub llm_chain: Arc<ProviderChain>,
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
/// Inbound bearer-auth token for this server's own `/v1/*` surface.
pub api_key: Option<String>,
/// Operator's own Serper.dev API key, read from `SERPER_API_KEY`.
/// Enables `/v1/search`. Unset = `/v1/search` returns 501.
pub serper_api_key: Option<String>,
}
impl AppState {
/// Build the application state. The fetch client is constructed once
/// and shared across requests so connection pools + browser profile
/// state don't churn per request.
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
///
/// `inbound_api_key` is the bearer token clients must present;
/// cloud-fallback credentials come from the env (checked here).
///
/// Async because the LLM provider chain probes Ollama for availability
/// once at startup; doing it here keeps it off the per-request hot path.
pub async fn new(inbound_api_key: Option<String>) -> anyhow::Result<Self> {
let config = FetchConfig {
browser: BrowserProfile::Firefox,
..FetchConfig::default()
};
let mut fetch = FetchClient::new(config.clone())
.map_err(|e| anyhow::anyhow!("failed to build fetch client: {e}"))?;
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
// Cloud fallback: only activates when the operator has provided
// an api.webclaw.io key. Supports both WEBCLAW_CLOUD_API_KEY
// (preferred, disambiguates from the inbound-auth key) and
// WEBCLAW_API_KEY as a fallback when there's no inbound key
// configured (backwards compat with MCP / CLI conventions).
if let Some(cloud) = build_cloud_client(inbound_api_key.as_deref()) {
info!(
base = cloud.base_url(),
"cloud fallback enabled — antibot-protected sites will escalate via api.webclaw.io"
);
fetch = fetch.with_cloud(cloud);
}
let llm_chain = Arc::new(ProviderChain::default().await);
// Operator's own Serper.dev key enables /v1/search. Empty/unset
// leaves search returning 501 with a setup hint.
let serper_api_key = std::env::var("SERPER_API_KEY")
.ok()
.filter(|k| !k.trim().is_empty());
if serper_api_key.is_some() {
info!("search enabled — using SERPER_API_KEY for /v1/search");
}
Ok(Self {
inner: Arc::new(Inner {
fetch: Arc::new(fetch),
fetch_config: config,
llm_chain,
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
api_key: inbound_api_key,
serper_api_key,
}),
})
}
pub fn fetch(&self) -> &Arc<FetchClient> {
&self.inner.fetch
}
/// The [`FetchConfig`] the shared client was built from. Cloned by
/// endpoints that need to construct their own client with identical
/// settings (currently `/v1/crawl`).
pub fn fetch_config(&self) -> &FetchConfig {
&self.inner.fetch_config
}
/// The shared LLM provider chain. Borrowed by `/v1/extract` and
/// `/v1/summarize`; `&ProviderChain` coerces to `&dyn LlmProvider`.
pub fn llm_chain(&self) -> &ProviderChain {
&self.inner.llm_chain
}
pub fn api_key(&self) -> Option<&str> {
self.inner.api_key.as_deref()
}
/// Operator's Serper.dev key for `/v1/search`, if configured.
pub fn serper_api_key(&self) -> Option<&str> {
self.inner.serper_api_key.as_deref()
}
}
feat(extractors): wave 5 \u2014 Amazon, eBay, Trustpilot via cloud fallback Three hard-site extractors that all require antibot bypass to ever return usable data. They ship in OSS so the parsers + schema live with the rest of the vertical extractors, but the fetch path routes through cloud::smart_fetch_html \u2014 meaning: - With WEBCLAW_CLOUD_API_KEY configured on webclaw-server (or WEBCLAW_API_KEY in MCP / CLI), local fetch is tried first; on challenge-page detection we escalate to api.webclaw.io/v1/scrape with formats=['html'] and parse the antibot-bypassed HTML locally. - Without a cloud key, callers get a typed CloudError::NotConfigured whose Display message points at https://webclaw.io/signup. Self-hosters without a webclaw.io account know exactly what to do. ## New extractors (all auto-dispatched \u2014 unique hosts) - amazon_product: ASIN extraction from /dp/, /gp/product/, /product/, /exec/obidos/ASIN/ URL shapes across every amazon.* locale. Parses the Product JSON-LD Amazon ships for SEO; falls back to #productTitle and #landingImage DOM selectors when JSON-LD is absent. Returns price, currency, availability, condition, brand, image, aggregate rating, SKU / MPN. - ebay_listing: item-id extraction from /itm/{id} and /itm/{slug}/{id} URLs across ebay.com / .co.uk / .de / .fr / .it. Parses both bare Offer (Buy It Now) and AggregateOffer (used-copies / auctions) from the Product JSON-LD. Returns price or low/high-price range, currency, condition, seller, offer_count, aggregate rating. - trustpilot_reviews: reactivated from the `trustpilot_reviews` file that was previously dead-code'd. Parser already worked; it just needed the smart_fetch_html path to get past AWS WAF's 'Verifying Connection' interstitial. Organisation / LocalBusiness JSON-LD block gives aggregate rating + up to 20 recent reviews. ## FetchClient change - Added optional `cloud: Option<Arc<CloudClient>>` field with `FetchClient::with_cloud(cloud) -> Self` builder + `cloud(&self)` accessor. Extractors call client.cloud() to decide whether they can escalate. Cheap clones (Arc-wrapped). ## webclaw-server wiring AppState::new() now reads the cloud credential from env: 1. WEBCLAW_CLOUD_API_KEY \u2014 preferred, disambiguates from the server's own inbound bearer token. 2. WEBCLAW_API_KEY \u2014 fallback only when the server is in open mode (no inbound-auth key set), matching the MCP / CLI convention of that env var. When present, state.rs builds a CloudClient and attaches it to the FetchClient via with_cloud(). Log line at startup so operators see when cloud fallback is active. ## Catalog + dispatch All three extractors registered in list() and in dispatch_by_url. /v1/extractors catalog now exposes 22 verticals. Explicit /v1/scrape/{vertical} routes work per the existing pattern. ## Tests - 7 new unit tests (parse_asin multi-shape + parse from JSON-LD fixture + DOM-fallback on missing JSON-LD for Amazon; ebay URL-matching + slugged-URL parsing + both Offer and AggregateOffer fixtures). - Full extractors suite: 68 passing (was 59, +9 from the new files). - fmt + clippy clean. - No live-test story for these three inside CI \u2014 verifying them means having WEBCLAW_CLOUD_API_KEY set against a real cloud backend. Integration-test harness is a separate follow-up. Catalog summary: 22 verticals total across wave 1-5. Hard-site three are gated behind an actionable cloud-fallback upgrade path rather than silently returning nothing or 403-ing the caller.
2026-04-22 16:16:11 +02:00
/// Resolve the outbound cloud key. Prefers `WEBCLAW_CLOUD_API_KEY`;
/// falls back to `WEBCLAW_API_KEY` *only* when no inbound key is
/// configured (i.e. open mode — the same env var can't mean two
/// things to one process).
fn build_cloud_client(inbound_api_key: Option<&str>) -> Option<CloudClient> {
let cloud_key = std::env::var("WEBCLAW_CLOUD_API_KEY").ok();
if let Some(k) = cloud_key.as_deref()
&& !k.trim().is_empty()
{
return Some(CloudClient::with_key(k));
}
// Reuse WEBCLAW_API_KEY only when not also acting as our own
// inbound-auth token — otherwise we'd be telling the operator
// they can't have both.
if inbound_api_key.is_none()
&& let Ok(k) = std::env::var("WEBCLAW_API_KEY")
&& !k.trim().is_empty()
{
return Some(CloudClient::with_key(k));
}
None
}