feat(insights): read location context from Apollo (#177)

The local location_history table is a one-time Takeout import: it ends
in 2019, still has the ghost-device and coarse cell-fix samples Apollo
has since excluded, and was read loosely. The per-photo line took the
nearest sample within +/-3 h at any distance (17% of photo moments got a
position >2 km away), and get_location_history listed the first 20 rows
of a +/-14-day window sorted ascending (always ~14 days before the date).

With APOLLO_API_BASE_URL set:
- per-photo context comes from /api/location/at (30 min max gap; no
  sample -> no line). A Place stay renders as "at Home (14:02-18:30)" in
  the display zone; an unnamed sample is reverse-geocoded.
- get_location_history returns Apollo's /api/location/day timeline for
  the date (+/-1 day at most), and is offered even with an empty local
  table.

Both fall back to the local table when Apollo is unset or a call fails;
the local tool path now covers the requested local day(s) and samples
evenly instead of listing a window's first rows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

migrations/2026-10-18-000000_add_clip_failures/down.sql

@@ -1 +0,0 @@
-DROP TABLE IF EXISTS clip_failures;

migrations/2026-10-18-000000_add_clip_failures/up.sql

@@ -1,17 +0,0 @@
--- "Don't retry" markers for the CLIP drain. A permanent encode failure
--- (bytes Apollo can't decode — corrupt or non-image files with an image
--- extension) used to leave `clip_embedding` NULL, so the drain re-sent
--- the file and logged the same 422 every pass.
---
--- Keyed on the path, scoped to the bytes that failed: the drain skips a
--- row only while its `content_hash` still matches, so replacing the file
--- makes it eligible again with no cleanup. To retry everything (e.g.
--- after an Apollo decoder upgrade): `DELETE FROM clip_failures;`
-CREATE TABLE clip_failures (
-    library_id INTEGER NOT NULL REFERENCES libraries(id) ON DELETE CASCADE,
-    rel_path TEXT NOT NULL,
-    content_hash TEXT NOT NULL,
-    error TEXT NOT NULL,
-    failed_at BIGINT NOT NULL,
-    PRIMARY KEY (library_id, rel_path)
-);

src/ai/clip_client.rs

@@ -27,10 +27,8 @@
 //! L2-normalized so the rerank reduces to a plain dot product.
 //!
 //! Error mapping (reflected in [`ClipError`]):
-//! - 422 `decode_failed` / `empty_text` → rejected input: ImageApi marks
-//!   the row failed or surfaces the empty-query error to the search caller.
-//! - Any other 4xx → permanent, but not the input's fault (auth, routing),
-//!   so nothing is marked.
+//! - 422 `decode_failed` / `empty_text` → permanent: ImageApi marks the
+//!   row failed or surfaces the empty-query error to the search caller.
 //! - 503 `cuda_oom` / `engine_unavailable` → defer-and-retry: no marker.
 //! - Any other 5xx / network error → defer.
 
@@ -93,12 +91,8 @@ pub struct ClipHealth {
 
 #[derive(Debug)]
 pub enum ClipError {
-    /// Apollo refused the input itself (422: decode failure on
-    /// /encode_image, empty text on /encode_text). Retrying the same
-    /// bytes will fail the same way.
-    Rejected(anyhow::Error),
-    /// Any other 4xx: won't fix itself on retry, but it's a deployment
-    /// problem (auth, routing), not this input's.
+    /// Apollo refused for a reason that won't change on retry (decode
+    /// failure on /encode_image, empty text on /encode_text).
     Permanent(anyhow::Error),
     /// Apollo couldn't process this turn but might next time (CUDA OOM,
     /// engine not loaded, network hiccup).
@@ -111,7 +105,6 @@ pub enum ClipError {
 impl std::fmt::Display for ClipError {
     fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
         match self {
-            ClipError::Rejected(e) => write!(f, "rejected: {e}"),
             ClipError::Permanent(e) => write!(f, "permanent: {e}"),
             ClipError::Transient(e) => write!(f, "transient: {e}"),
             ClipError::Disabled => write!(f, "clip client disabled"),
@@ -285,7 +278,7 @@ fn classify_error_response(status: u16, body_text: &str) -> ClipError {
         .unwrap_or_default();
 
     if status == 422 {
-        return ClipError::Rejected(anyhow::anyhow!(
+        return ClipError::Permanent(anyhow::anyhow!(
             "clip {} {}: {}",
             status,
             detail_code,
@@ -330,9 +323,6 @@ fn classify_error_response(status: u16, body_text: &str) -> ClipError {
 mod tests {
     use super::*;
 
-    fn is_rejected(e: &ClipError) -> bool {
-        matches!(e, ClipError::Rejected(_))
-    }
     fn is_permanent(e: &ClipError) -> bool {
         matches!(e, ClipError::Permanent(_))
     }
@@ -341,16 +331,16 @@ mod tests {
     }
 
     #[test]
-    fn classify_422_decode_failed_is_rejected() {
-        assert!(is_rejected(&classify_error_response(
+    fn classify_422_decode_failed_is_permanent() {
+        assert!(is_permanent(&classify_error_response(
             422,
             r#"{"detail":"decode_failed: bad bytes"}"#
         )));
     }
 
     #[test]
-    fn classify_422_empty_text_is_rejected() {
-        assert!(is_rejected(&classify_error_response(
+    fn classify_422_empty_text_is_permanent() {
+        assert!(is_permanent(&classify_error_response(
             422,
             r#"{"detail":"empty_text"}"#
         )));

src/bin/probe_clip_search.rs

@@ -227,7 +227,7 @@ async fn main() -> anyhow::Result<()> {
                     }
                     let _ = call_start;
                 }
-                Err(ClipError::Rejected(e) | ClipError::Permanent(e)) => {
+                Err(ClipError::Permanent(e)) => {
                     perm_fail += 1;
                     warn!("permanent encode failure for {rel_path}: {e}");
                 }

src/clip_search.rs

@@ -172,9 +172,7 @@ pub async fn score_photos(
     // 1. Encode the query text. Fast — Apollo's text encoder is ~50ms on CPU.
     let query_resp = match state.clip_client.encode_text(q_text).await {
         Ok(r) => r,
-        Err(ClipError::Rejected(e) | ClipError::Permanent(e)) => {
-            return Err(ScoreError::Rejected(e.to_string()));
-        }
+        Err(ClipError::Permanent(e)) => return Err(ScoreError::Rejected(e.to_string())),
         Err(ClipError::Transient(e)) => return Err(ScoreError::Unavailable(e.to_string())),
         Err(ClipError::Disabled) => return Err(ScoreError::Disabled),
     };

src/clip_watch.rs

@@ -6,12 +6,12 @@
 //! parallel encode calls to Apollo, and persist the resulting embeddings
 //! into `image_exif.clip_embedding` / `clip_model_version`.
 //!
-//! A permanent failure (un-decodable bytes) records a `clip_failures`
-//! marker scoped to the row's `content_hash`, so the drain stops
-//! re-sending the file (and re-logging the same 422) every pass until
-//! its bytes change. `file_types::is_image_file` filters out videos /
-//! non-media client-side, so most permanent failures are corrupt files
-//! or non-images with an image extension.
+//! Unlike the face pipeline, CLIP has no marker rows — a permanent
+//! failure (un-decodable bytes) leaves the row's `clip_embedding` NULL
+//! and the drain will retry on the next tick. For personal-library
+//! scale this is fine; the per-tick cap bounds the wasted work, and
+//! `file_types::is_image_file` filters out videos / non-media client-
+//! side so most permanent failures are decoded-but-corrupt files (rare).
 //!
 //! The watcher thread isn't in any pre-existing async context, so we
 //! build a short-lived tokio runtime per pass and `block_on` the join
@@ -172,29 +172,15 @@ async fn process_one(
                 cand.rel_path, resp.embedding_dim, resp.duration_ms, resp.model_version
             );
         }
-        Err(ClipError::Rejected(e)) => {
-            warn!(
-                "clip_watch: permanent failure on {}: {} — marking failed",
-                cand.rel_path, e
-            );
-            let mut dao = exif_dao.lock().expect("exif dao");
-            if let Err(me) = dao.mark_clip_failed(
-                library_id,
-                &cand.rel_path,
-                &cand.content_hash,
-                &e.to_string(),
-            ) {
-                warn!(
-                    "clip_watch: mark failed errored for {}: {:?}",
-                    cand.rel_path, me
-                );
-            }
-        }
         Err(ClipError::Permanent(e)) => {
-            // Not this file's fault (auth / routing) — don't mark it, or
-            // a misconfigured Apollo would blacklist the whole library.
+            // No marker — the row sits with NULL embedding and the drain
+            // retries next pass. For personal-library scale the cost of
+            // re-attempting permanently-broken files is bounded by the
+            // per-tick cap. If this becomes a recurring noise source,
+            // add a `clip_status` column with `failed` semantics like
+            // face_detections has.
             warn!(
-                "clip_watch: Apollo refused {} (will retry next pass): {}",
+                "clip_watch: permanent failure on {} (will retry next pass): {}",
                 cand.rel_path, e
             );
         }

src/database/mod.rs

@@ -551,17 +551,6 @@ pub trait ExifDao: Sync + Send {
         model_version: &str,
     ) -> Result<(), DbError>;
 
-    /// Record that Apollo permanently rejected these bytes, so
-    /// `list_clip_unencoded_candidates` stops offering the row until its
-    /// `content_hash` changes. Upserts — one marker per path.
-    fn mark_clip_failed(
-        &mut self,
-        library_id: i32,
-        rel_path: &str,
-        content_hash: &str,
-        error: &str,
-    ) -> Result<(), DbError>;
-
     /// Load every `(content_hash, clip_embedding)` pair from the live
     /// image_exif rows for the given libraries, optionally filtered to a
     /// single `model_version` (cosine sim across mixed geometries is
@@ -1530,21 +1519,10 @@ impl ExifDao for SqliteExifDao {
             // (clip_embedding IS NULL AND content_hash IS NOT NULL)
             // filter; the planner hits it directly. ORDER BY id ASC
             // keeps drain progress monotone across ticks.
-            // Skip rows whose current bytes already failed permanently
-            // (`mark_clip_failed`); a changed hash makes them eligible.
-            let failed = schema::clip_failures::table
-                .filter(schema::clip_failures::library_id.eq(library_id))
-                .filter(schema::clip_failures::rel_path.eq(rel_path))
-                .filter(
-                    schema::clip_failures::content_hash
-                        .nullable()
-                        .eq(content_hash),
-                );
             image_exif
                 .filter(library_id.eq(library_id_val))
                 .filter(clip_embedding.is_null())
                 .filter(content_hash.is_not_null())
-                .filter(diesel::dsl::not(diesel::dsl::exists(failed)))
                 .select((rel_path, content_hash.assume_not_null()))
                 .order(id.asc())
                 .limit(limit)
@@ -1607,41 +1585,6 @@ impl ExifDao for SqliteExifDao {
         })
     }
 
-    fn mark_clip_failed(
-        &mut self,
-        library_id_val: i32,
-        rel_path_val: &str,
-        content_hash_val: &str,
-        error_val: &str,
-    ) -> Result<(), DbError> {
-        trace_db_call("insert", "mark_clip_failed", |_span| {
-            use schema::clip_failures::dsl::*;
-
-            let mut connection = self.connection.lock().expect("Unable to get ExifDao");
-
-            let now = chrono::Utc::now().timestamp();
-            diesel::insert_into(clip_failures)
-                .values((
-                    library_id.eq(library_id_val),
-                    rel_path.eq(rel_path_val),
-                    content_hash.eq(content_hash_val),
-                    error.eq(error_val),
-                    failed_at.eq(now),
-                ))
-                .on_conflict((library_id, rel_path))
-                .do_update()
-                .set((
-                    content_hash.eq(content_hash_val),
-                    error.eq(error_val),
-                    failed_at.eq(now),
-                ))
-                .execute(connection.deref_mut())
-                .map(|_| ())
-                .map_err(|e| anyhow::anyhow!("Insert error: {}", e))
-        })
-        .map_err(|e| DbError::log(DbErrorKind::InsertError, e))
-    }
-
     fn list_clip_index(
         &mut self,
         library_ids_val: &[i32],
@@ -2616,37 +2559,6 @@ mod exif_dao_tests {
         SqliteExifDao::from_connection(conn)
     }
 
-    #[test]
-    fn clip_failure_marker_skips_row_until_its_bytes_change() {
-        let mut dao = setup_two_libraries();
-        insert_row(&mut dao, 1, "bad.jpg", None);
-        insert_row(&mut dao, 1, "good.jpg", None);
-        insert_row(&mut dao, 2, "bad.jpg", None);
-        for (lib, rel) in [(1, "bad.jpg"), (1, "good.jpg"), (2, "bad.jpg")] {
-            dao.backfill_content_hash(lib, rel, "h1", 1).unwrap();
-        }
-        let rels = |dao: &mut SqliteExifDao, lib| -> Vec<String> {
-            dao.list_clip_unencoded_candidates(lib, 10)
-                .unwrap()
-                .into_iter()
-                .map(|(rel, _)| rel)
-                .collect()
-        };
-
-        dao.mark_clip_failed(1, "bad.jpg", "h1", "decode_failed")
-            .unwrap();
-        assert_eq!(rels(&mut dao, 1), vec!["good.jpg"]);
-        // Same path in another library is a different file.
-        assert_eq!(rels(&mut dao, 2), vec!["bad.jpg"]);
-        // Re-marking upserts rather than erroring on the primary key.
-        dao.mark_clip_failed(1, "bad.jpg", "h1", "decode_failed")
-            .unwrap();
-
-        // The file was replaced: new bytes get another try.
-        dao.backfill_content_hash(1, "bad.jpg", "h2", 1).unwrap();
-        assert_eq!(rels(&mut dao, 1), vec!["bad.jpg", "good.jpg"]);
-    }
-
     #[test]
     fn get_all_with_date_taken_union_returns_all_libraries() {
         let mut dao = setup_two_libraries();

src/database/schema.rs

@@ -18,16 +18,6 @@ diesel::table! {
     }
 }
 
-diesel::table! {
-    clip_failures (library_id, rel_path) {
-        library_id -> Integer,
-        rel_path -> Text,
-        content_hash -> Text,
-        error -> Text,
-        failed_at -> BigInt,
-    }
-}
-
 diesel::table! {
     daily_conversation_summaries (id) {
         id -> Integer,
@@ -358,7 +348,6 @@ diesel::table! {
 diesel::joinable!(entity_facts -> photo_insights (source_insight_id));
 diesel::joinable!(entity_photo_links -> entities (entity_id));
 diesel::joinable!(entity_photo_links -> libraries (library_id));
-diesel::joinable!(clip_failures -> libraries (library_id));
 diesel::joinable!(face_detections -> libraries (library_id));
 diesel::joinable!(face_detections -> persons (person_id));
 diesel::joinable!(image_exif -> libraries (library_id));
@@ -371,7 +360,6 @@ diesel::joinable!(video_preview_clips -> libraries (library_id));
 
 diesel::allow_tables_to_appear_in_same_query!(
     calendar_events,
-    clip_failures,
     daily_conversation_summaries,
     entities,
     entity_facts,

src/files.rs

@@ -1910,16 +1910,6 @@ mod tests {
             Ok(())
         }
 
-        fn mark_clip_failed(
-            &mut self,
-            _library_id: i32,
-            _rel_path: &str,
-            _content_hash: &str,
-            _error: &str,
-        ) -> Result<(), DbError> {
-            Ok(())
-        }
-
         fn list_clip_index(
             &mut self,
             _library_ids: &[i32],