chore: reuse reqwest::Client across requests in forage embedder; minor forage updates

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

applications/forage/embedder/src/main.rs

@@ -53,7 +53,12 @@ struct Args {
 #[derive(Clone)]
 enum Mode {
     Mock,
-    OpenAi { api_key: String },
+    /// `client` is created once at startup and reused across requests.
+    /// `reqwest::Client` is cheaply cloneable (`Arc`-backed connection pool).
+    OpenAi {
+        api_key: String,
+        client: reqwest::Client,
+    },
 }
 
 #[derive(Deserialize)]
@@ -70,7 +75,7 @@ struct EmbedResp {
 async fn post_embed(State(mode): State<Arc<Mode>>, Json(req): Json<EmbedReq>) -> impl IntoResponse {
     let vector = match mode.as_ref() {
         Mode::Mock => mock_embed(&req.text),
-        Mode::OpenAi { api_key } => match openai_embed(api_key, &req.text).await {
+        Mode::OpenAi { api_key, client } => match openai_embed(client, api_key, &req.text).await {
             Ok(v) => v,
             Err(e) => {
                 return (
@@ -119,8 +124,11 @@ fn mock_embed(text: &str) -> Vec<f32> {
 }
 
 /// OpenAI text-embedding-3-small call.
-async fn openai_embed(api_key: &str, text: &str) -> Result<Vec<f32>, String> {
-    let client = reqwest::Client::new();
+async fn openai_embed(
+    client: &reqwest::Client,
+    api_key: &str,
+    text: &str,
+) -> Result<Vec<f32>, String> {
     let resp = client
         .post("https://api.openai.com/v1/embeddings")
         .bearer_auth(api_key)
@@ -178,7 +186,10 @@ async fn main() {
             Mode::Mock
         } else {
             println!("forage-embedder: OpenAI mode (text-embedding-3-small)");
-            Mode::OpenAi { api_key: key }
+            Mode::OpenAi {
+                api_key: key,
+                client: reqwest::Client::new(),
+            }
         }
     };
 

applications/forage/engine/src/lib.rs

@@ -1175,10 +1175,10 @@ fn call_embedder(
 /// 3. Strip a trailing slash from the path unless the path is the root `/`.
 fn canonicalize_url(url: &str) -> String {
     // 1. Strip amp. subdomain
-    let s = if url.starts_with("https://amp.") {
-        format!("https://{}", &url["https://amp.".len()..])
-    } else if url.starts_with("http://amp.") {
-        format!("http://{}", &url["http://amp.".len()..])
+    let s = if let Some(rest) = url.strip_prefix("https://amp.") {
+        format!("https://{rest}")
+    } else if let Some(rest) = url.strip_prefix("http://amp.") {
+        format!("http://{rest}")
     } else {
         url.to_owned()
     };
@@ -1199,7 +1199,7 @@ fn canonicalize_url(url: &str) -> String {
         // Find position of the first slash after "://"
         let after_scheme = base.find("://").map_or(0, |i| i + 3);
         let first_path_slash = base[after_scheme..].find('/');
-        let has_real_path = first_path_slash.map_or(false, |j| base.len() > after_scheme + j + 1);
+        let has_real_path = first_path_slash.is_some_and(|j| base.len() > after_scheme + j + 1);
         if has_real_path && base.ends_with('/') {
             base[..base.len() - 1].to_owned()
         } else {
@@ -1229,6 +1229,73 @@ fn canonicalize_url(url: &str) -> String {
     }
 }
 
+#[cfg(test)]
+mod canon_tests {
+    use super::canonicalize_url;
+
+    #[test]
+    fn strips_amp_subdomain_https() {
+        assert_eq!(
+            canonicalize_url("https://amp.example.com/article/123"),
+            "https://example.com/article/123"
+        );
+    }
+
+    #[test]
+    fn strips_amp_subdomain_http() {
+        assert_eq!(
+            canonicalize_url("http://amp.cnn.com/story"),
+            "http://cnn.com/story"
+        );
+    }
+
+    #[test]
+    fn removes_amp_query_param_standalone() {
+        assert_eq!(
+            canonicalize_url("https://example.com/article?amp=1"),
+            "https://example.com/article"
+        );
+    }
+
+    #[test]
+    fn removes_amp_tf_query_param_among_others() {
+        assert_eq!(
+            canonicalize_url("https://example.com/a?q=rust&amp_tf=1&page=2"),
+            "https://example.com/a?q=rust&page=2"
+        );
+    }
+
+    #[test]
+    fn strips_trailing_slash_from_real_path() {
+        assert_eq!(
+            canonicalize_url("https://example.com/article/"),
+            "https://example.com/article"
+        );
+    }
+
+    #[test]
+    fn preserves_root_slash() {
+        assert_eq!(
+            canonicalize_url("https://example.com/"),
+            "https://example.com/"
+        );
+    }
+
+    #[test]
+    fn preserves_meaningful_query_params() {
+        let url = "https://example.com/search?q=rust&lang=en";
+        assert_eq!(canonicalize_url(url), url);
+    }
+
+    #[test]
+    fn amp_subdomain_plus_amp_param_both_stripped() {
+        assert_eq!(
+            canonicalize_url("https://amp.example.com/post/?amp=1"),
+            "https://example.com/post"
+        );
+    }
+}
+
 /// Round-robin interleave items by category to ensure the cold-start exploit
 /// pool spans ≥3 categories. Preserves score ordering within each category.
 ///

applications/forage/extension/manifest.json

@@ -3,7 +3,7 @@
   "name": "Forage",
   "version": "0.1.0",
   "description": "Automatically capture browsing signals for your Forage personalized feed",
-  "permissions": ["storage"],
+  "permissions": ["storage", "tabs"],
   "host_permissions": ["http://*/*", "https://*/*"],
   "content_scripts": [
     {

applications/forage/js/capture.js

@@ -15,10 +15,13 @@
  * Configuration:
  *   Change USER_ID to match the Forage user you are browsing as (1, 2, or 3
  *   for the seed users; any positive integer for a new user).
+ *   Set TOKEN to the value passed with --token when starting the server,
+ *   or leave empty ('') if the server was started without --token.
  */
 (function forageCapture() {
   const SERVER = 'http://localhost:4242';
   const USER_ID = 1;
+  const TOKEN = '';
   const DWELL_MS = 30_000;
 
   const url = location.href;
@@ -46,9 +49,11 @@
 
   let itemId = null;
 
+  const authHeaders = TOKEN ? { Authorization: `Bearer ${TOKEN}` } : {};
+
   fetch(`${SERVER}/capture`, {
     method: 'POST',
-    headers: { 'Content-Type': 'application/json' },
+    headers: { 'Content-Type': 'application/json', ...authHeaders },
     body: JSON.stringify({
       url,
       canonical_url: canonicalUrl,
@@ -73,7 +78,7 @@
     if (itemId == null) return;
     fetch(`${SERVER}/signal`, {
       method: 'POST',
-      headers: { 'Content-Type': 'application/json' },
+      headers: { 'Content-Type': 'application/json', ...authHeaders },
       body: JSON.stringify({
         user_id: USER_ID,
         item_id: itemId,