CLAUDE.md

@@ -533,6 +533,11 @@ POST /insights/chat                  (single-turn reply, non-streaming)
 POST /insights/chat/stream           (SSE: text / tool_call / tool_result / truncated / done)
 GET  /insights/chat/history?path=... (rendered transcript with tool invocations)
 POST /insights/chat/rewind           (truncate transcript at a rendered index)
+
+// User AI Prefs (owner defaults + passive mirror; src/prefs.rs)
+GET  /prefs                          (current row; auth)
+PUT  /prefs                          (partial update: omitted=unchanged, null=clear, value=set; auth)
+GET  /timezones                      (IANA zone list from chrono-tz; unauthenticated)
 ```
 
 **Request Types:**

CONTEXT.md

@@ -62,3 +62,25 @@ Insights and pre-computed artifacts grouped by calendar day (daily message summa
 reel pre-generation). Because it is keyed on the day, re-dating a photo across a day
 boundary invalidates the derived data for that day, which regenerates on its normal schedule.
 _Avoid_: cache, nightly output
+
+### Configuration
+
+**Owner default**:
+A per-request default the owner sets for the whole deployment — the model, backend,
+persona, reel span, or segment cap a client gets when it omits the field. Stored in the
+single `user_ai_prefs` row, hot-tunable without a redeploy, and global: every client that
+omits the field gets it, on every device. Never holds secrets.
+_Avoid_: user prefs, user settings, preferences (ambiguous with device-local client state)
+
+**Deployment config**:
+Values the server needs before it can start serving — bind address, database path, media
+roots, the URLs used to build clients — plus every secret. Lives in `.env` only; changing
+it is a redeploy, by design.
+_Avoid_: environment config, system config
+
+**Hard default**:
+The env/code fallback a request lands on when neither an explicit param nor a stored
+owner default supplies the field (e.g. backend `local`, persona `default`, reel span
+day). It is the floor under the owner-default layer: env changes keep working whenever a
+default column is NULL.
+_Avoid_: fallback value, env default

docs/adr/0002-owner-defaults-vs-env.md

@@ -0,0 +1,78 @@
+# ADR-0002: Owner defaults live in the DB row; .env keeps deployment config and secrets
+
+**Status**: accepted (2026-09-27)
+
+## Context
+
+All configuration lived in `.env`: bind address, media paths, Ollama/llama-swap/OpenRouter
+URLs, model names, `LLM_BACKEND`, the embedding contract, TTS voices. Two friction points:
+values the owner actually *tunes* (which model, which backend, which persona, how long a
+reel is) needed an edit + redeploy to change, and clients had no server-side fallback —
+"use the default" was expressed by omitting a field, and the default itself was invisible
+and untouchable at runtime. The `user_ai_prefs` single row (`id = 1`) already existed as
+a passive mirror of the last client reel request for the nightly pre-gen. #126/#127 gave
+it an explicit `PUT /prefs` surface and five owner-default columns (`default_model`,
+`default_backend`, `default_persona_id`, `reel_span`, `reel_max_segments`) — the first
+batch of non-private configuration moved out of `.env` and into the database.
+
+## Decision
+
+Configuration splits into two homes by one rule — *does the server need the value before
+it can start serving?*
+
+- **`.env` (deployment config)**: everything read at boot to bind, connect, or open —
+  `BIND_URL`, `DATABASE_URL`, media paths, `OLLAMA_*_URL` / `LLAMA_SWAP_URL`,
+  `LLM_BACKEND`, the embedding contract — and **all secrets** (`OPENROUTER_API_KEY`,
+  `SECRET_KEY`, `SMS_API_TOKEN`). Secrets never go in the database: the SQLite file gets
+  backed up, copied, and attached to bug reports.
+- **`user_ai_prefs` (owner defaults)**: per-request defaults resolved at request time —
+  the model/backend/persona a client gets when it omits the field, reel span and
+  segment cap, display zone, TTS voice. Hot-tunable via `PUT /prefs`, effective on the
+  next request, no redeploy. This row is the designated home for future non-private,
+  per-request configuration that outgrows `.env`. The table is a **single row per
+  deployment, not a per-user profile** — `user_ai_prefs` is a historical name; the
+  canonical term is *owner defaults* (single-user tool: owner = admin).
+
+Resolution order at every consumer: **explicit request param → stored owner default →
+hard default** (the env/code fallback). An explicit request always wins, so any client
+can override the owner's default per request. The env value stays the floor under the
+DB: a NULL `default_model` means "whatever the Ollama/llama-swap primary is", so env
+changes keep working.
+
+**Privacy posture is safe-by-default.** `default_backend` is the one owner default with
+network-egress consequences: `hybrid` routes insight generation and chat to OpenRouter.
+With no row or a NULL column, the hard default is `local` — data stays on the LAN.
+Leaving the network requires an explicit owner action (storing `hybrid`), and any client
+can still pin itself by sending `backend` explicitly. Owner defaults therefore split
+into two sensitivity classes: **harmless tuning** (span, segments, display zone, voice,
+model, persona) and **privacy/security** (`default_backend`), which any admin UI must
+surface with a warning, not a quiet dropdown.
+
+## Considered options
+
+- **Everything in `.env`** (status quo): safe, but tuning a model default means edit +
+  redeploy, and clients have no server-side default to fall back to — a picker's
+  "server default" is just the env value, invisible and unchangeable at runtime.
+  Rejected for per-request defaults.
+- **Everything in the DB**: boot-time values (bind address, DB path, the URLs used to
+  build the HTTP clients) become chicken-and-egg, and secrets sit next to tuning values
+  in a file that gets copied around. Rejected.
+- **A separate `system_config` table**: cleaner name, but a second single-row table
+  would be a synonym with extra migrations; the row and the `PUT /prefs` surface already
+  exist. Rejected — the row keeps its name, its *term* becomes owner defaults.
+
+## Non-obvious rules worth remembering
+
+- **Owner defaults are global to the deployment.** Every client that omits a field gets
+  the stored value, on every device. Storing `default_backend=hybrid` changes the
+  mobile app's behavior even for users who never touch the setting; a client that must
+  stay local should send `backend` explicitly (tracked for FileViewer-React #33).
+- **A stored value that is no longer consumable fails at write time, not request time** —
+  `PUT /prefs` validates `default_backend` against the canonical `BackendKind::parse`
+  and `reel_span` against the span enum before writing, so a bad default cannot 500
+  every request later.
+- **The passive reel mirror (`capture_prefs`) may only write the fields it owns** (voice,
+  library, display zone, effective span/segments); it carries the AI owner-defaults
+  through unchanged, because the upsert overwrites every column.
+- **NULL means "use the hard default", not "unset the env"** — the DB layer sits
+  *above* env, never instead of it.

migrations/2026-09-27-000000_add_owner_defaults/down.sql

@@ -0,0 +1,5 @@
+ALTER TABLE user_ai_prefs DROP COLUMN default_model;
+ALTER TABLE user_ai_prefs DROP COLUMN default_backend;
+ALTER TABLE user_ai_prefs DROP COLUMN default_persona_id;
+ALTER TABLE user_ai_prefs DROP COLUMN reel_span;
+ALTER TABLE user_ai_prefs DROP COLUMN reel_max_segments;

migrations/2026-09-27-000000_add_owner_defaults/up.sql

@@ -0,0 +1,8 @@
+-- Owner defaults: recurring per-user picks that were previously re-sent on
+-- every request. All nullable — NULL = keep the current per-request /
+-- hard-default behavior.
+ALTER TABLE user_ai_prefs ADD COLUMN default_model TEXT;
+ALTER TABLE user_ai_prefs ADD COLUMN default_backend TEXT;
+ALTER TABLE user_ai_prefs ADD COLUMN default_persona_id TEXT;
+ALTER TABLE user_ai_prefs ADD COLUMN reel_span TEXT;
+ALTER TABLE user_ai_prefs ADD COLUMN reel_max_segments INTEGER;

src/ai/daily_summary_job.rs

@@ -360,11 +360,17 @@ async fn generate_and_store_daily_summary(
     span.set_attribute(KeyValue::new("date", date.to_string()));
     span.set_attribute(KeyValue::new("contact", contact.to_string()));
     span.set_attribute(KeyValue::new("message_count", messages.len() as i64));
+    // Parent context for the child llm.request spans; the span itself stays
+    // borrowed here for the attribute updates below.
+    let llm_cx =
+        opentelemetry::Context::new().with_remote_span_context(span.span_context().clone());
 
     let (prompt, system_prompt) = build_daily_summary_prompt(contact, date, messages);
 
     // Generate summary with LLM
-    let summary = ollama.generate(&prompt, Some(system_prompt)).await?;
+    let summary = ollama
+        .generate(&llm_cx, &prompt, Some(system_prompt))
+        .await?;
 
     log::debug!(
         "Generated summary for {}: {}",
@@ -382,7 +388,9 @@ async fn generate_and_store_daily_summary(
     );
 
     // Embed the stripped summary (store original summary in DB)
-    let embedding = ollama.generate_embedding(&stripped_summary).await?;
+    let embedding = ollama
+        .generate_embedding(&llm_cx, &stripped_summary)
+        .await?;
 
     span.set_attribute(KeyValue::new(
         "embedding_dimensions",

src/ai/handlers.rs

@@ -384,6 +384,36 @@ pub struct ServerModels {
     pub default_model: String,
 }
 
+/// Apply the stored owner defaults to any request field the client omitted.
+/// Resolution order: request param → stored pref → hard default (applied
+/// downstream). The request always wins; the pref only fills gaps. One DAO
+/// read per request; a failed read silently falls back to hard defaults.
+fn apply_owner_defaults(
+    app_state: &AppState,
+    model: &mut Option<String>,
+    backend: &mut Option<String>,
+    persona_id: &mut Option<String>,
+) {
+    if model.is_some() && backend.is_some() && persona_id.is_some() {
+        return; // nothing to fill — skip the DB read
+    }
+    let Some(mut dao) = app_state.user_ai_prefs_dao.lock().ok() else {
+        return;
+    };
+    let Ok(Some(prefs)) = dao.get_prefs(&opentelemetry::Context::new()) else {
+        return;
+    };
+    if model.is_none() {
+        *model = prefs.default_model;
+    }
+    if backend.is_none() {
+        *backend = prefs.default_backend;
+    }
+    if persona_id.is_none() {
+        *persona_id = prefs.default_persona_id;
+    }
+}
+
 /// POST /insights/generate - Generate insight for a specific photo (async)
 #[post("/insights/generate")]
 pub async fn generate_insight_handler(
@@ -396,6 +426,16 @@ pub async fn generate_insight_handler(
     let tracer = global_tracer();
     let mut span = tracer.start_with_context("http.insights.generate", &parent_context);
 
+    let mut request = request.0;
+    // Stored owner defaults fill what the client omitted (request → pref →
+    // hard default; the hard default still applies downstream).
+    apply_owner_defaults(
+        &app_state,
+        &mut request.model,
+        &mut request.backend,
+        &mut request.persona_id,
+    );
+
     let normalized_path = normalize_path(&request.file_path);
     let library = app_state.primary_library();
     let gen_type = InsightGenerationType::Standard;
@@ -714,6 +754,16 @@ pub async fn generate_agentic_insight_handler(
     let tracer = global_tracer();
     let mut span = tracer.start_with_context("http.insights.generate_agentic", &parent_context);
 
+    let mut request = request.0;
+    // Stored owner defaults fill what the client omitted (request → pref →
+    // hard default; the hard default still applies downstream).
+    apply_owner_defaults(
+        &app_state,
+        &mut request.model,
+        &mut request.backend,
+        &mut request.persona_id,
+    );
+
     let normalized_path = normalize_path(&request.file_path);
     let library = app_state.primary_library();
     let gen_type = InsightGenerationType::Agentic;
@@ -1231,6 +1281,16 @@ pub async fn chat_turn_handler(
     let mut span = tracer.start_with_context("http.insights.chat", &parent_context);
     span.set_attribute(KeyValue::new("file_path", request.file_path.clone()));
 
+    let mut request = request.0;
+    // Stored owner defaults fill what the client omitted (request → pref →
+    // hard default; the hard default still applies downstream).
+    apply_owner_defaults(
+        &app_state,
+        &mut request.model,
+        &mut request.backend,
+        &mut request.persona_id,
+    );
+
     let library =
         match libraries::resolve_library_param_state(&app_state, request.library.as_deref()) {
             Ok(Some(lib)) => lib,
@@ -1607,10 +1667,22 @@ pub struct ChatSwitchBranchRequest {
 /// Returns `text/event-stream` with one event per chat stream event.
 #[post("/insights/chat/stream")]
 pub async fn chat_stream_handler(
+    http_request: HttpRequest,
     claims: Claims,
     request: web::Json<ChatTurnHttpRequest>,
     app_state: web::Data<AppState>,
 ) -> HttpResponse {
+    let parent_context = extract_context_from_request(&http_request);
+    let mut request = request.0;
+    // Stored owner defaults fill what the client omitted (request → pref →
+    // hard default; the hard default still applies downstream).
+    apply_owner_defaults(
+        &app_state,
+        &mut request.model,
+        &mut request.backend,
+        &mut request.persona_id,
+    );
+
     let library =
         match libraries::resolve_library_param_state(&app_state, request.library.as_deref()) {
             Ok(Some(lib)) => lib,
@@ -1646,7 +1718,7 @@ pub async fn chat_stream_handler(
     };
 
     let service = app_state.insight_chat.clone();
-    let events = service.chat_turn_stream(chat_req);
+    let events = service.chat_turn_stream(chat_req, parent_context);
 
     // Map ChatStreamEvent → SSE frame bytes.
     let sse_stream = futures::stream::StreamExt::map(events, |ev| {
@@ -1758,6 +1830,16 @@ pub async fn turn_async_handler(
     let mut span = tracer.start_with_context("http.insights.chat_turn_async", &parent_context);
     span.set_attribute(KeyValue::new("file_path", request.file_path.clone()));
 
+    let mut request = request.0;
+    // Stored owner defaults fill what the client omitted (request → pref →
+    // hard default; the hard default still applies downstream).
+    apply_owner_defaults(
+        &app_state,
+        &mut request.model,
+        &mut request.backend,
+        &mut request.persona_id,
+    );
+
     let library =
         match libraries::resolve_library_param_state(&app_state, request.library.as_deref()) {
             Ok(Some(lib)) => lib,
@@ -2277,3 +2359,95 @@ mod turn_replay_tests {
         assert_eq!(info.total_events_pushed, 2);
     }
 }
+
+#[cfg(test)]
+mod owner_defaults_tests {
+    use super::apply_owner_defaults;
+    use crate::database::models::UpsertUserAiPrefs;
+    use crate::database::{SqliteUserAiPrefsDao, UserAiPrefsDao};
+    use crate::state::AppState;
+    use std::sync::{Arc, Mutex};
+
+    fn prefs_state(prefs: Option<UpsertUserAiPrefs>) -> AppState {
+        let conn = Arc::new(Mutex::new(crate::database::test::in_memory_db_connection()));
+        let mut state = AppState::test_state();
+        state.user_ai_prefs_dao = Arc::new(Mutex::new(Box::new(
+            SqliteUserAiPrefsDao::from_connection(conn.clone()),
+        )));
+        if let Some(p) = prefs {
+            let mut dao = SqliteUserAiPrefsDao::from_connection(conn);
+            dao.upsert_prefs(&opentelemetry::Context::new(), &p)
+                .expect("seed prefs");
+        }
+        state
+    }
+
+    fn seeded_defaults() -> UpsertUserAiPrefs {
+        UpsertUserAiPrefs {
+            voice: None,
+            tz_offset_minutes: None,
+            library: None,
+            updated_at: 0,
+            display_tz: None,
+            default_model: Some("llama3.2:3b".into()),
+            default_backend: Some("hybrid".into()),
+            default_persona_id: Some("cabin".into()),
+            reel_span: None,
+            reel_max_segments: None,
+        }
+    }
+
+    #[actix_rt::test]
+    async fn request_wins_over_stored_pref() {
+        let state = prefs_state(Some(seeded_defaults()));
+        let mut model = Some("explicit".to_string());
+        let mut backend = Some("local".to_string());
+        let mut persona = Some("explicit-p".to_string());
+        apply_owner_defaults(&state, &mut model, &mut backend, &mut persona);
+        assert_eq!(model.as_deref(), Some("explicit"));
+        assert_eq!(backend.as_deref(), Some("local"));
+        assert_eq!(persona.as_deref(), Some("explicit-p"));
+    }
+
+    #[actix_rt::test]
+    async fn pref_fills_omitted_fields() {
+        let state = prefs_state(Some(seeded_defaults()));
+        let mut model = None;
+        let mut backend = None;
+        let mut persona = None;
+        apply_owner_defaults(&state, &mut model, &mut backend, &mut persona);
+        assert_eq!(model.as_deref(), Some("llama3.2:3b"));
+        assert_eq!(backend.as_deref(), Some("hybrid"));
+        assert_eq!(persona.as_deref(), Some("cabin"));
+    }
+
+    #[actix_rt::test]
+    async fn no_pref_row_leaves_fields_none() {
+        let state = prefs_state(None);
+        let mut model = None;
+        let mut backend = None;
+        let mut persona = None;
+        apply_owner_defaults(&state, &mut model, &mut backend, &mut persona);
+        // Hard defaults apply downstream; the pref layer fills nothing.
+        assert!(model.is_none());
+        assert!(backend.is_none());
+        assert!(persona.is_none());
+    }
+
+    #[actix_rt::test]
+    async fn pref_fills_only_the_fields_it_has() {
+        let state = prefs_state(Some(UpsertUserAiPrefs {
+            default_backend: None,
+            default_persona_id: Some("journal".into()),
+            ..seeded_defaults()
+        }));
+        let mut model = None;
+        let mut backend = Some("local".to_string());
+        let mut persona = None;
+        apply_owner_defaults(&state, &mut model, &mut backend, &mut persona);
+        assert_eq!(model.as_deref(), Some("llama3.2:3b"));
+        // Explicit request backend is untouched; pref had none anyway.
+        assert_eq!(backend.as_deref(), Some("local"));
+        assert_eq!(persona.as_deref(), Some("journal"));
+    }
+}

src/ai/insight_chat.rs

@@ -377,7 +377,7 @@ impl InsightChatService {
 
             let (response, prompt_tokens, eval_tokens) = backend
                 .chat()
-                .chat_with_tools(messages.clone(), tools.clone())
+                .chat_with_tools(&loop_cx, messages.clone(), tools.clone())
                 .await?;
             last_prompt_eval_count = prompt_tokens;
             last_eval_count = eval_tokens;
@@ -434,7 +434,7 @@ impl InsightChatService {
             let synthetic_idx = push_synthetic_final_prompt(&mut messages);
             let (final_response, prompt_tokens, eval_tokens) = backend
                 .chat()
-                .chat_with_tools(messages.clone(), vec![])
+                .chat_with_tools(&loop_cx, messages.clone(), vec![])
                 .await?;
             last_prompt_eval_count = prompt_tokens;
             last_eval_count = eval_tokens;
@@ -492,6 +492,7 @@ impl InsightChatService {
             let title_raw = backend
                 .chat()
                 .generate(
+                    &loop_cx,
                     &title_prompt,
                     Some(
                         "You are my long term memory assistant. Use only the information provided. Do not invent details.",
@@ -781,10 +782,11 @@ impl InsightChatService {
     pub fn chat_turn_stream(
         self: Arc<Self>,
         req: ChatTurnRequest,
+        cx: opentelemetry::Context,
     ) -> BoxStream<'static, ChatStreamEvent> {
         let svc = self;
         let s = async_stream::stream! {
-            match svc.chat_turn_stream_inner(req, Ok).await {
+            match svc.chat_turn_stream_inner(req, cx, Ok).await {
                 Ok(mut rx) => {
                     while let Some(ev) = rx.recv().await {
                         yield ev;
@@ -806,6 +808,7 @@ impl InsightChatService {
     async fn chat_turn_stream_inner<F>(
         self: Arc<Self>,
         req: ChatTurnRequest,
+        cx: opentelemetry::Context,
         _ev_mapper: F,
     ) -> Result<tokio::sync::mpsc::Receiver<ChatStreamEvent>>
     where
@@ -814,7 +817,7 @@ impl InsightChatService {
         let (tx, rx) = tokio::sync::mpsc::channel::<ChatStreamEvent>(64);
         let svc = self.clone();
         tokio::spawn(async move {
-            let result = svc.run_streaming_turn(req, tx.clone()).await;
+            let result = svc.run_streaming_turn(req, cx, tx.clone()).await;
             if let Err(e) = result {
                 let _ = tx.send(ChatStreamEvent::Error(format!("{}", e))).await;
             }
@@ -852,7 +855,7 @@ impl InsightChatService {
             span.set_attribute(KeyValue::new("library_id", library_id as i64));
 
             let result = svc
-                .run_streaming_turn_with_entry(req, entry_clone.clone())
+                .run_streaming_turn_with_entry(req, entry_clone.clone(), &mut span)
                 .await;
             if let Err(ref e) = result {
                 span.set_attribute(KeyValue::new("status", "error"));
@@ -882,6 +885,7 @@ impl InsightChatService {
         self: Arc<Self>,
         req: ChatTurnRequest,
         entry: Arc<TurnEntry>,
+        turn_span: &mut opentelemetry::global::BoxedSpan,
     ) -> Result<()> {
         if req.user_message.trim().is_empty() {
             bail!("user_message must not be empty");
@@ -911,11 +915,11 @@ impl InsightChatService {
 
         if req.regenerate || existing_insight.is_none() {
             return self
-                .run_bootstrap_streaming_with_entry(req, normalized, entry)
+                .run_bootstrap_streaming_with_entry(req, normalized, entry, turn_span)
                 .await;
         }
         let insight = existing_insight.expect("just checked Some above");
-        self.run_continuation_streaming_with_entry(req, normalized, insight, entry)
+        self.run_continuation_streaming_with_entry(req, normalized, insight, entry, turn_span)
             .await
     }
 
@@ -926,6 +930,7 @@ impl InsightChatService {
         normalized: String,
         insight: crate::database::models::PhotoInsight,
         entry: Arc<TurnEntry>,
+        turn_span: &mut opentelemetry::global::BoxedSpan,
     ) -> Result<()> {
         let active_persona = req
             .persona_id
@@ -981,6 +986,12 @@ impl InsightChatService {
         };
         let backend = self.generator.resolve_backend(kind, &overrides).await?;
         let model_used = backend.model().to_string();
+        // llm.* attributes on the turn-level span (issue #130): clients can't
+        // know local-vs-hybrid, but the turn knows its resolved backend.
+        turn_span.set_attribute(KeyValue::new("llm.backend", kind.as_str()));
+        turn_span.set_attribute(KeyValue::new("llm.model", model_used.clone()));
+        let turn_cx = opentelemetry::Context::new()
+            .with_remote_span_context(turn_span.span_context().clone());
 
         let local_first_user_has_image = messages
             .iter()
@@ -1030,6 +1041,7 @@ impl InsightChatService {
                 &active_persona,
                 max_iterations,
                 &entry,
+                &turn_cx,
             )
             .await?;
         let AgenticLoopOutcome {
@@ -1146,6 +1158,7 @@ impl InsightChatService {
         req: ChatTurnRequest,
         normalized: String,
         entry: Arc<TurnEntry>,
+        turn_span: &mut opentelemetry::global::BoxedSpan,
     ) -> Result<()> {
         let active_persona = req
             .persona_id
@@ -1171,6 +1184,11 @@ impl InsightChatService {
         };
         let backend = self.generator.resolve_backend(kind, &overrides).await?;
         let model_used = backend.model().to_string();
+        // llm.* attributes on the turn-level span (issue #130).
+        turn_span.set_attribute(KeyValue::new("llm.backend", kind.as_str()));
+        turn_span.set_attribute(KeyValue::new("llm.model", model_used.clone()));
+        let turn_cx = opentelemetry::Context::new()
+            .with_remote_span_context(turn_span.span_context().clone());
 
         let image_base64: Option<String> = self.generator.load_image_as_base64(&normalized).ok();
 
@@ -1190,7 +1208,7 @@ impl InsightChatService {
 
         let visual_block = if !backend.images_inline {
             match image_base64.as_deref() {
-                Some(b64) => match backend.local().describe_image(b64).await {
+                Some(b64) => match backend.local().describe_image(&turn_cx, b64).await {
                     Ok(desc) => {
                         format!("Visual description (from local vision model):\n{}\n", desc)
                     }
@@ -1245,6 +1263,7 @@ impl InsightChatService {
                 &active_persona,
                 max_iterations,
                 &entry,
+                &turn_cx,
             )
             .await?;
         let AgenticLoopOutcome {
@@ -1333,6 +1352,7 @@ impl InsightChatService {
         active_persona: &str,
         max_iterations: usize,
         entry: &Arc<TurnEntry>,
+        cx: &opentelemetry::Context,
     ) -> Result<AgenticLoopOutcome> {
         crate::ai::insight_chat::run_streaming_agentic_loop_with_entry(
             &self.generator,
@@ -1345,6 +1365,7 @@ impl InsightChatService {
             active_persona,
             max_iterations,
             entry,
+            cx,
         )
         .await
     }
@@ -1352,6 +1373,7 @@ impl InsightChatService {
     async fn run_streaming_turn(
         self: Arc<Self>,
         req: ChatTurnRequest,
+        cx: opentelemetry::Context,
         tx: tokio::sync::mpsc::Sender<ChatStreamEvent>,
     ) -> Result<()> {
         if req.user_message.trim().is_empty() {
@@ -1381,17 +1403,17 @@ impl InsightChatService {
         // library bootstraps cleanly, even when another library has
         // an insight for the same rel_path.
         let existing_insight = {
-            let cx = opentelemetry::Context::new();
+            let db_cx = opentelemetry::Context::new();
             let mut dao = self.insight_dao.lock().expect("Unable to lock InsightDao");
-            dao.get_current_insight_for_library(&cx, req.library_id, &normalized)
+            dao.get_current_insight_for_library(&db_cx, req.library_id, &normalized)
                 .map_err(|e| anyhow!("failed to load insight: {:?}", e))?
         };
 
         if req.regenerate || existing_insight.is_none() {
-            return self.run_bootstrap_streaming(req, normalized, tx).await;
+            return self.run_bootstrap_streaming(req, normalized, cx, tx).await;
         }
         let insight = existing_insight.expect("just checked Some above");
-        self.run_continuation_streaming(req, normalized, insight, tx)
+        self.run_continuation_streaming(req, normalized, cx, insight, tx)
             .await
     }
 
@@ -1403,6 +1425,7 @@ impl InsightChatService {
         &self,
         req: ChatTurnRequest,
         normalized: String,
+        cx: opentelemetry::Context,
         insight: crate::database::models::PhotoInsight,
         tx: tokio::sync::mpsc::Sender<ChatStreamEvent>,
     ) -> Result<()> {
@@ -1500,6 +1523,7 @@ impl InsightChatService {
                 &active_persona,
                 max_iterations,
                 &tx,
+                &cx,
             )
             .await?;
         let AgenticLoopOutcome {
@@ -1611,6 +1635,7 @@ impl InsightChatService {
         &self,
         req: ChatTurnRequest,
         normalized: String,
+        cx: opentelemetry::Context,
         tx: tokio::sync::mpsc::Sender<ChatStreamEvent>,
     ) -> Result<()> {
         let active_persona = req
@@ -1667,7 +1692,7 @@ impl InsightChatService {
         // images_inline backends send images directly to the chat model.
         let visual_block = if !backend.images_inline {
             match image_base64.as_deref() {
-                Some(b64) => match backend.local().describe_image(b64).await {
+                Some(b64) => match backend.local().describe_image(&cx, b64).await {
                     Ok(desc) => {
                         format!("Visual description (from local vision model):\n{}\n", desc)
                     }
@@ -1734,6 +1759,7 @@ impl InsightChatService {
                 &active_persona,
                 max_iterations,
                 &tx,
+                &cx,
             )
             .await?;
         let AgenticLoopOutcome {
@@ -1816,6 +1842,7 @@ impl InsightChatService {
         active_persona: &str,
         max_iterations: usize,
         tx: &tokio::sync::mpsc::Sender<ChatStreamEvent>,
+        cx: &opentelemetry::Context,
     ) -> Result<AgenticLoopOutcome> {
         let mut tool_calls_made = 0usize;
         let mut iterations_used = 0usize;
@@ -1834,7 +1861,7 @@ impl InsightChatService {
 
             let mut stream = backend
                 .chat()
-                .chat_with_tools_stream(messages.clone(), tools.clone())
+                .chat_with_tools_stream(cx, messages.clone(), tools.clone())
                 .await?;
 
             let mut final_message: Option<ChatMessage> = None;
@@ -1928,7 +1955,7 @@ impl InsightChatService {
             let synthetic_idx = push_synthetic_final_prompt(messages);
             let mut stream = backend
                 .chat()
-                .chat_with_tools_stream(messages.clone(), vec![])
+                .chat_with_tools_stream(cx, messages.clone(), vec![])
                 .await?;
             let mut final_message: Option<ChatMessage> = None;
             while let Some(ev) = stream.next().await {
@@ -2308,6 +2335,7 @@ pub async fn run_streaming_agentic_loop_with_entry(
     active_persona: &str,
     max_iterations: usize,
     entry: &Arc<TurnEntry>,
+    cx: &opentelemetry::Context,
 ) -> Result<AgenticLoopOutcome> {
     let mut tool_calls_made = 0usize;
     let mut iterations_used = 0usize;
@@ -2337,7 +2365,7 @@ pub async fn run_streaming_agentic_loop_with_entry(
 
         let mut stream = backend
             .chat()
-            .chat_with_tools_stream(messages.clone(), tools.clone())
+            .chat_with_tools_stream(cx, messages.clone(), tools.clone())
             .await?;
 
         let mut final_message: Option<ChatMessage> = None;
@@ -2420,7 +2448,7 @@ pub async fn run_streaming_agentic_loop_with_entry(
         let synthetic_idx = push_synthetic_final_prompt(messages);
         let mut stream = backend
             .chat()
-            .chat_with_tools_stream(messages.clone(), vec![])
+            .chat_with_tools_stream(cx, messages.clone(), vec![])
             .await?;
         let mut final_message: Option<ChatMessage> = None;
         while let Some(ev) = stream.next().await {

src/ai/insight_generator.rs

@@ -646,7 +646,8 @@ impl InsightGenerator {
         // (`LLM_BACKEND` switch). Must match the backend that populated the
         // daily-summary embeddings or similarity search will be garbage.
         let query_embedding =
-            crate::ai::embed_query(&self.ollama, self.llamacpp.as_deref(), &query).await?;
+            crate::ai::embed_query(&search_cx, &self.ollama, self.llamacpp.as_deref(), &query)
+                .await?;
 
         // Search for similar daily summaries with time-based weighting
         // This prioritizes summaries temporally close to the query date
@@ -712,7 +713,8 @@ impl InsightGenerator {
         // Must use the same backend that populated the daily-summary
         // embeddings or similarity search is garbage (see embed_one docs).
         let query_embedding =
-            crate::ai::embed_query(&self.ollama, self.llamacpp.as_deref(), query).await?;
+            crate::ai::embed_query(&search_cx, &self.ollama, self.llamacpp.as_deref(), query)
+                .await?;
 
         let mut summary_dao = self
             .daily_summary_dao
@@ -798,7 +800,9 @@ impl InsightGenerator {
         let calendar_cx = parent_cx.with_span(span);
 
         let query_embedding = if let Some(loc) = location {
-            match crate::ai::embed_query(&self.ollama, self.llamacpp.as_deref(), loc).await {
+            match crate::ai::embed_query(&calendar_cx, &self.ollama, self.llamacpp.as_deref(), loc)
+                .await
+            {
                 Ok(emb) => Some(emb),
                 Err(e) => {
                     log::warn!("Failed to generate embedding for location '{}': {}", loc, e);
@@ -969,18 +973,23 @@ impl InsightGenerator {
             )
         };
 
-        let query_embedding =
+        let query_embedding = match crate::ai::embed_query(
-            match crate::ai::embed_query(&self.ollama, self.llamacpp.as_deref(), &query_text).await
+            &search_cx,
-            {
+            &self.ollama,
-                Ok(emb) => emb,
+            self.llamacpp.as_deref(),
-                Err(e) => {
+            &query_text,
-                    log::warn!("Failed to generate search embedding: {}", e);
+        )
-                    search_cx.span().set_status(Status::Error {
+        .await
-                        description: e.to_string().into(),
+        {
-                    });
+            Ok(emb) => emb,
-                    return Ok(None);
+            Err(e) => {
-                }
+                log::warn!("Failed to generate search embedding: {}", e);
-            };
+                search_cx.span().set_status(Status::Error {
+                    description: e.to_string().into(),
+                });
+                return Ok(None);
+            }
+        };
 
         let searches = {
             let mut dao = self
@@ -1254,7 +1263,10 @@ impl InsightGenerator {
 
         // Generate brief photo description for RAG enrichment (vision models only)
         let photo_description: Option<String> = if let Some(ref img_b64) = image_base64 {
-            match ollama_client.generate_photo_description(img_b64).await {
+            match ollama_client
+                .generate_photo_description(&insight_cx, img_b64)
+                .await
+            {
                 Ok(desc) => {
                     log::info!("Photo description for RAG enrichment: {}", desc);
                     Some(desc)
@@ -1323,7 +1335,7 @@ impl InsightGenerator {
             if !immediate_messages.is_empty() {
                 // Step 2: Extract topics from immediate messages to enrich RAG query
                 let topics = self
-                    .extract_topics_from_messages(&immediate_messages, &ollama_client)
+                    .extract_topics_from_messages(&immediate_messages, &ollama_client, &insight_cx)
                     .await;
 
                 log::info!("Extracted topics for query enrichment: {:?}", topics);
@@ -1379,6 +1391,7 @@ impl InsightGenerator {
                                 &immediate_messages,
                                 &ollama_client,
                                 custom_system_prompt.as_deref(),
+                                &insight_cx,
                             )
                             .await
                             .unwrap_or_else(|| String::from("No immediate context"));
@@ -1388,6 +1401,7 @@ impl InsightGenerator {
                                 &historical_messages,
                                 &ollama_client,
                                 custom_system_prompt.as_deref(),
+                                &insight_cx,
                             )
                             .await
                             .unwrap_or_else(|| String::from("No historical context"));
@@ -1406,6 +1420,7 @@ impl InsightGenerator {
                                 &immediate_messages,
                                 &ollama_client,
                                 custom_system_prompt.as_deref(),
+                                &insight_cx,
                             )
                             .await;
                     }
@@ -1416,6 +1431,7 @@ impl InsightGenerator {
                                 &immediate_messages,
                                 &ollama_client,
                                 custom_system_prompt.as_deref(),
+                                &insight_cx,
                             )
                             .await;
                     }
@@ -1441,6 +1457,7 @@ impl InsightGenerator {
                                 &rag_messages,
                                 &ollama_client,
                                 custom_system_prompt.as_deref(),
+                                &insight_cx,
                             )
                             .await;
                     }
@@ -1473,7 +1490,7 @@ impl InsightGenerator {
             if !sms_messages.is_empty() {
                 match self
                     .sms_client
-                    .summarize_context(&sms_messages, &ollama_client)
+                    .summarize_context(&sms_messages, &ollama_client, &insight_cx)
                     .await
                 {
                     Ok(summary) => {
@@ -1557,6 +1574,7 @@ impl InsightGenerator {
         // 10. Generate summary first, then derive title from the summary
         let summary = ollama_client
             .generate_photo_summary(
+                &insight_cx,
                 date_taken,
                 location.as_deref(),
                 contact.as_deref(),
@@ -1567,7 +1585,7 @@ impl InsightGenerator {
             .await?;
 
         let title = ollama_client
-            .generate_photo_title(&summary, custom_system_prompt.as_deref())
+            .generate_photo_title(&insight_cx, &summary, custom_system_prompt.as_deref())
             .await?;
 
         log::info!("Generated title: {}", title);
@@ -1634,6 +1652,7 @@ impl InsightGenerator {
         &self,
         messages: &[crate::ai::SmsMessage],
         ollama: &OllamaClient,
+        cx: &opentelemetry::Context,
     ) -> Vec<String> {
         if messages.is_empty() {
             return Vec::new();
@@ -1669,7 +1688,7 @@ Return ONLY the comma-separated list, nothing else."#,
         );
 
         match ollama
-            .generate(&prompt, Some("You are an entity extraction assistant. Extract proper nouns, people, places, and domain-specific terms that provide context."))
+            .generate(cx, &prompt, Some("You are an entity extraction assistant. Extract proper nouns, people, places, and domain-specific terms that provide context."))
             .await
         {
             Ok(response) => {
@@ -1704,6 +1723,7 @@ Return ONLY the comma-separated list, nothing else."#,
         messages: &[String],
         ollama: &OllamaClient,
         custom_system: Option<&str>,
+        cx: &opentelemetry::Context,
     ) -> Option<String> {
         if messages.is_empty() {
             return None;
@@ -1724,7 +1744,7 @@ Return ONLY the summary, nothing else."#,
         let system = custom_system
             .unwrap_or("You are a context summarization assistant. Be concise and factual.");
 
-        match ollama.generate(&prompt, Some(system)).await {
+        match ollama.generate(cx, &prompt, Some(system)).await {
             Ok(summary) => Some(summary),
             Err(e) => {
                 log::warn!("Failed to summarize messages: {}", e);
@@ -1740,6 +1760,7 @@ Return ONLY the summary, nothing else."#,
         messages: &[crate::ai::SmsMessage],
         ollama: &OllamaClient,
         custom_system: Option<&str>,
+        cx: &opentelemetry::Context,
     ) -> Option<String> {
         if messages.is_empty() {
             return None;
@@ -1785,7 +1806,7 @@ Return ONLY the summary, nothing else."#,
             "You are a context summarization assistant. Be detailed and factual, preserving important context.",
         );
 
-        match ollama.generate(&prompt, Some(system)).await {
+        match ollama.generate(cx, &prompt, Some(system)).await {
             Ok(summary) => Some(summary),
             Err(e) => {
                 log::warn!("Failed to summarize immediate context: {}", e);
@@ -1829,7 +1850,7 @@ Return ONLY the summary, nothing else."#,
             "get_file_tags" => self.tool_get_file_tags(arguments, cx).await,
             "get_faces_in_photo" => self.tool_get_faces_in_photo(arguments, cx).await,
             "describe_photo" => {
-                self.tool_describe_photo(backend.local(), image_base64)
+                self.tool_describe_photo(backend.local(), image_base64, cx)
                     .await
             }
             "reverse_geocode" => self.tool_reverse_geocode(arguments).await,
@@ -1880,7 +1901,7 @@ Return ONLY the summary, nothing else."#,
         &self,
         args: &serde_json::Value,
         local: &dyn LlmClient,
-        _cx: &opentelemetry::Context,
+        cx: &opentelemetry::Context,
     ) -> String {
         let query = match args.get("query").and_then(|v| v.as_str()) {
             Some(q) => q.to_string(),
@@ -1944,7 +1965,10 @@ Return ONLY the summary, nothing else."#,
         };
 
         let final_results = if rerank_enabled && results.len() > limit {
-            match self.rerank_with_llm(&query, &results, limit, local).await {
+            match self
+                .rerank_with_llm(cx, &query, &results, limit, local)
+                .await
+            {
                 Ok(reordered) => reordered,
                 Err(e) => {
                     log::warn!("rerank failed, using vector order: {}", e);
@@ -1967,6 +1991,7 @@ Return ONLY the summary, nothing else."#,
     /// tool_result. On parse failure we fall back to the input order.
     async fn rerank_with_llm(
         &self,
+        cx: &opentelemetry::Context,
         query: &str,
         candidates: &[String],
         limit: usize,
@@ -2010,7 +2035,7 @@ Return ONLY the summary, nothing else."#,
         let started = std::time::Instant::now();
         let system =
             Some("You are a terse relevance ranker. You output only numbers separated by commas.");
-        let response = local.generate(&prompt, system, None).await?;
+        let response = local.generate(cx, &prompt, system, None).await?;
         log::info!(
             "rerank: finished in {} ms (prompt={} chars)",
             started.elapsed().as_millis(),
@@ -2620,10 +2645,11 @@ Return ONLY the summary, nothing else."#,
         &self,
         local: &dyn LlmClient,
         image_base64: &Option<String>,
+        cx: &opentelemetry::Context,
     ) -> String {
         log::info!("tool_describe_photo: generating visual description");
         match image_base64 {
-            Some(img) => match local.describe_image(img).await {
+            Some(img) => match local.describe_image(cx, img).await {
                 Ok(desc) => desc,
                 Err(e) => format!("Error describing photo: {}", e),
             },
@@ -3031,6 +3057,7 @@ Return ONLY the summary, nothing else."#,
         // configured local backend.
         let embed_text = format!("{} {}", name, description);
         let embedding: Option<Vec<u8>> = match crate::ai::embed_document(
+            cx,
             &self.ollama,
             self.llamacpp.as_deref(),
             &embed_text,
@@ -4277,7 +4304,7 @@ Return ONLY the summary, nothing else."#,
         // the local backend so non-text work stays off OpenRouter.
         let inlined_visual_description: Option<String> = if !backend.images_inline {
             match image_base64.as_deref() {
-                Some(b64) => match backend.local().describe_image(b64).await {
+                Some(b64) => match backend.local().describe_image(&insight_cx, b64).await {
                     Ok(desc) => {
                         log::info!("{}: vision describe succeeded ({} chars)", kind, desc.len());
                         Some(desc)
@@ -4397,7 +4424,7 @@ Return ONLY the summary, nothing else."#,
 
             let (response, prompt_tokens, eval_tokens) = backend
                 .chat()
-                .chat_with_tools(messages.clone(), tools.clone())
+                .chat_with_tools(&loop_cx, messages.clone(), tools.clone())
                 .await?;
 
             last_prompt_eval_count = prompt_tokens;
@@ -4466,7 +4493,7 @@ Return ONLY the summary, nothing else."#,
             )));
             let (final_response, prompt_tokens, eval_tokens) = backend
                 .chat()
-                .chat_with_tools(messages.clone(), tools.clone())
+                .chat_with_tools(&loop_cx, messages.clone(), tools.clone())
                 .await?;
             last_prompt_eval_count = prompt_tokens;
             last_eval_count = eval_tokens;
@@ -4499,7 +4526,7 @@ Return ONLY the summary, nothing else."#,
                 }
                 let (resp2, pt2, et2) = backend
                     .chat()
-                    .chat_with_tools(messages.clone(), vec![])
+                    .chat_with_tools(&loop_cx, messages.clone(), vec![])
                     .await?;
                 last_prompt_eval_count = pt2;
                 last_eval_count = et2;
@@ -4637,6 +4664,7 @@ Return ONLY the summary, nothing else."#,
         tools: Vec<Tool>,
         max_iter: usize,
     ) -> Result<String> {
+        let loop_cx = opentelemetry::Context::current();
         let mut final_content = String::new();
 
         for iteration in 0..max_iter {
@@ -4644,7 +4672,7 @@ Return ONLY the summary, nothing else."#,
 
             let (response, _prompt_tokens, _eval_tokens) = backend
                 .chat()
-                .chat_with_tools(messages.clone(), tools.clone())
+                .chat_with_tools(&loop_cx, messages.clone(), tools.clone())
                 .await?;
 
             // Sanitize tool call arguments before pushing back into history.
@@ -4686,7 +4714,7 @@ Return ONLY the summary, nothing else."#,
                             "",
                             0,
                             "",
-                            &opentelemetry::Context::new(),
+                            &loop_cx,
                         )
                         .await;
                     messages.push(ChatMessage::tool_result(result));
@@ -4710,7 +4738,7 @@ Return ONLY the summary, nothing else."#,
             ));
             let (final_response, _, _) = backend
                 .chat()
-                .chat_with_tools(messages.clone(), tools.clone())
+                .chat_with_tools(&loop_cx, messages.clone(), tools.clone())
                 .await?;
             // Same bounded tool round as the insight loop: execute any last
             // tool call, then force a tool-less turn for the actual answer.
@@ -4728,14 +4756,14 @@ Return ONLY the summary, nothing else."#,
                             "",
                             0,
                             "",
-                            &opentelemetry::Context::new(),
+                            &loop_cx,
                         )
                         .await;
                     messages.push(ChatMessage::tool_result(result));
                 }
                 let (resp2, _, _) = backend
                     .chat()
-                    .chat_with_tools(messages.clone(), vec![])
+                    .chat_with_tools(&loop_cx, messages.clone(), vec![])
                     .await?;
                 final_content = resp2.content;
             } else {

src/ai/llamacpp.rs

@@ -20,8 +20,11 @@
 // insight_generator / insight_chat.
 #![allow(dead_code)]
 
-use anyhow::{Context, Result, anyhow, bail};
+use anyhow::Context as _;
+use anyhow::{Result, anyhow, bail};
 use async_trait::async_trait;
+use opentelemetry::trace::{Span, Status};
+use opentelemetry::{Context, KeyValue};
 use reqwest::Client;
 use serde::Deserialize;
 use serde_json::{Value, json};
@@ -486,6 +489,7 @@ impl LlamaCppClient {
     async fn chat_completion_with_model(
         &self,
         model: &str,
+        mut span: Option<&mut opentelemetry::global::BoxedSpan>,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<(ChatMessage, Option<i32>, Option<i32>)> {
@@ -517,19 +521,31 @@ impl LlamaCppClient {
         // Wait for any TTS synthesis to release the GPU before the request
         // timeout starts (see ai::gpu).
         let _gpu = crate::ai::gpu::llm_lease().await;
-        let resp = self
+        let mut request = self.client.post(&url).json(&Value::Object(body));
-            .client
+        if let Some(s) = span.as_mut() {
-            .post(&url)
+            request = crate::otel::with_traceparent(request, s.span_context());
-            .json(&Value::Object(body))
+        }
-            .send()
+        let resp = match request.send().await {
-            .await
+            Ok(r) => r,
-            .with_context(|| format!("POST {} failed", url))?;
+            Err(e) => {
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("llama-swap request failed"));
+                }
+                return Err(e).with_context(|| format!("POST {} failed", url));
+            }
+        };
 
         if !resp.status().is_success() {
             let status = resp.status();
+            crate::otel::record_http_failure(
+                span.as_deref_mut(),
+                status.as_u16(),
+                &format!("llama-swap chat request failed: {status}"),
+            );
             let body = resp.text().await.unwrap_or_default();
             bail!("llama-swap chat request failed: {} — {}", status, body);
         }
+        crate::otel::set_http_status(span.as_deref_mut(), resp.status().as_u16());
 
         let parsed: Value = resp.json().await.context("parsing chat response")?;
         let choice = parsed
@@ -562,14 +578,35 @@ impl LlamaCppClient {
 
         log_timings(&parsed, prompt_tokens, completion_tokens);
 
+        if let Some(s) = span.as_mut() {
+            if let Some(pt) = prompt_tokens {
+                s.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = completion_tokens {
+                s.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            s.set_status(Status::Ok);
+        }
+
         Ok((chat_msg, prompt_tokens, completion_tokens))
     }
+
+    /// Start the `llm.request` span for an outgoing llama-swap call.
+    fn start_request_span(
+        &self,
+        cx: &Context,
+        operation: &str,
+        model: &str,
+    ) -> opentelemetry::global::BoxedSpan {
+        crate::otel::start_llm_request_span(cx, "llamacpp", operation, model)
+    }
 }
 
 #[async_trait]
 impl LlmClient for LlamaCppClient {
     async fn generate(
         &self,
+        cx: &Context,
         prompt: &str,
         system: Option<&str>,
         images: Option<Vec<String>>,
@@ -582,12 +619,15 @@ impl LlmClient for LlamaCppClient {
         user.images = images;
         messages.push(user);
 
-        let (reply, _, _) = self.chat_with_tools(messages, Vec::new()).await?;
+        // Delegates to chat_with_tools, which starts the llm.request span
+        // (operation=chat) — one span per HTTP request.
+        let (reply, _, _) = self.chat_with_tools(cx, messages, Vec::new()).await?;
         Ok(reply.content)
     }
 
     async fn chat_with_tools(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<(ChatMessage, Option<i32>, Option<i32>)> {
@@ -597,15 +637,25 @@ impl LlmClient for LlamaCppClient {
             messages.len(),
             tools.len()
         );
-        self.chat_completion_with_model(&self.primary_model.clone(), messages, tools)
+        let mut span = self.start_request_span(cx, "chat", &self.primary_model);
-            .await
+        self.chat_completion_with_model(
+            &self.primary_model.clone(),
+            Some(&mut span),
+            messages,
+            tools,
+        )
+        .await
     }
 
     async fn chat_with_tools_stream(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<BoxStream<'static, Result<LlmStreamEvent>>> {
+        // The span outlives this method — it moves into the stream
+        // generator and is finalized when the response drains.
+        let mut span = self.start_request_span(cx, "chat_stream", &self.primary_model);
         let url = format!("{}/chat/completions", self.base_url);
         let roles: Vec<&str> = messages.iter().map(|m| m.role.as_str()).collect();
         log::debug!(
@@ -639,19 +689,24 @@ impl LlmClient for LlamaCppClient {
         // timeout starts (see ai::gpu). The guard is moved into the stream
         // below so the lease spans the whole generation, not just the send.
         let gpu = crate::ai::gpu::llm_lease().await;
-        let resp = self
+        let mut request = self.client.post(&url).json(&Value::Object(body));
-            .client
+        request = crate::otel::with_traceparent(request, span.span_context());
-            .post(&url)
+        let resp = request
-            .json(&Value::Object(body))
             .send()
             .await
             .with_context(|| format!("POST {} failed", url))?;
 
         if !resp.status().is_success() {
             let status = resp.status();
+            crate::otel::record_http_failure(
+                Some(&mut span),
+                status.as_u16(),
+                &format!("llama-swap stream request failed: {status}"),
+            );
             let body = resp.text().await.unwrap_or_default();
             bail!("llama-swap stream request failed: {} — {}", status, body);
         }
+        crate::otel::set_http_status(Some(&mut span), resp.status().as_u16());
 
         let byte_stream = resp.bytes_stream();
         let stream = async_stream::stream! {
@@ -760,6 +815,13 @@ impl LlmClient for LlamaCppClient {
                 tool_calls,
                 images: None,
             };
+            if let Some(pt) = prompt_tokens {
+                span.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = completion_tokens {
+                span.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            span.set_status(Status::Ok);
             yield Ok(LlmStreamEvent::Done {
                 message,
                 prompt_eval_count: prompt_tokens,
@@ -770,7 +832,8 @@ impl LlmClient for LlamaCppClient {
         Ok(Box::pin(stream))
     }
 
-    async fn generate_embeddings(&self, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
+    async fn generate_embeddings(&self, cx: &Context, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
+        let mut span = self.start_request_span(cx, "embeddings", &self.embedding_model);
         // Deliberately NO GPU lease: the embed slot sits in llama-swap's
         // always-resident group and never participates in a model swap, so
         // leasing here would only stall searches behind a queued synthesis.
@@ -780,19 +843,27 @@ impl LlmClient for LlamaCppClient {
             "input": texts,
         });
 
-        let resp = self
+        let mut request = self.client.post(&url).json(&body);
-            .client
+        request = crate::otel::with_traceparent(request, span.span_context());
-            .post(&url)
+        let resp = match request.send().await {
-            .json(&body)
+            Ok(r) => r,
-            .send()
+            Err(e) => {
-            .await
+                span.set_status(Status::error("llama-swap embedding request failed"));
-            .with_context(|| format!("POST {} failed", url))?;
+                return Err(e).with_context(|| format!("POST {} failed", url));
+            }
+        };
 
         if !resp.status().is_success() {
             let status = resp.status();
+            crate::otel::record_http_failure(
+                Some(&mut span),
+                status.as_u16(),
+                &format!("llama-swap embedding request failed: {status}"),
+            );
             let body = resp.text().await.unwrap_or_default();
             bail!("llama-swap embedding request failed: {} — {}", status, body);
         }
+        crate::otel::set_http_status(Some(&mut span), resp.status().as_u16());
 
         #[derive(Deserialize)]
         struct EmbedResponse {
@@ -803,11 +874,20 @@ impl LlmClient for LlamaCppClient {
             embedding: Vec<f32>,
         }
 
-        let parsed: EmbedResponse = resp.json().await.context("parsing embed response")?;
+        let parsed: EmbedResponse = match resp.json().await {
+            Ok(p) => p,
+            Err(e) => {
+                span.set_status(Status::error("parsing embed response failed"));
+                return Err(e).context("parsing embed response");
+            }
+        };
+        span.set_status(Status::Ok);
         Ok(parsed.data.into_iter().map(|i| i.embedding).collect())
     }
 
-    async fn describe_image(&self, image_base64: &str) -> Result<String> {
+    async fn describe_image(&self, cx: &Context, image_base64: &str) -> Result<String> {
+        // Own span with operation=describe_image — routes through the vision slot.
+        let mut span = self.start_request_span(cx, "describe_image", &self.vision_model);
         let prompt = "Briefly describe what you see in this image in 1-2 sentences. \
                       Focus on the people, location, and activity.";
         let system = "You are a scene description assistant. Be concise and factual.";
@@ -823,7 +903,12 @@ impl LlmClient for LlamaCppClient {
         ];
 
         let (reply, _, _) = self
-            .chat_completion_with_model(&self.vision_model.clone(), messages, Vec::new())
+            .chat_completion_with_model(
+                &self.vision_model.clone(),
+                Some(&mut span),
+                messages,
+                Vec::new(),
+            )
             .await?;
         Ok(reply.content)
     }

src/ai/llm_client.rs

@@ -1,6 +1,7 @@
 use anyhow::Result;
 use async_trait::async_trait;
 use futures::stream::BoxStream;
+use opentelemetry::Context;
 use serde::{Deserialize, Serialize};
 
 /// Provider-agnostic surface for LLM backends (Ollama, OpenRouter, …).
@@ -10,6 +11,12 @@ use serde::{Deserialize, Serialize};
 /// when a provider requires it (OpenAI-compatible APIs do), and `images`
 /// stays as base64 strings here and is rewritten into content-parts where
 /// needed.
+///
+/// Every method that performs an HTTP call takes the caller's tracing
+/// `Context` as its first parameter so the impl can start an `llm.request`
+/// span parented under the caller's feature span — the codebase threads
+/// contexts explicitly rather than relying on ambient `Context::current()`
+/// at deep call sites.
 // First consumer lands in a later PR (OpenRouter impl + hybrid mode routing).
 #[allow(dead_code)]
 #[async_trait]
@@ -18,6 +25,7 @@ pub trait LlmClient: Send + Sync {
     /// base64 images (ignored by providers without vision support).
     async fn generate(
         &self,
+        cx: &Context,
         prompt: &str,
         system: Option<&str>,
         images: Option<Vec<String>>,
@@ -27,6 +35,7 @@ pub trait LlmClient: Send + Sync {
     /// (which may contain tool_calls) plus optional prompt/eval token counts.
     async fn chat_with_tools(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<(ChatMessage, Option<i32>, Option<i32>)>;
@@ -39,16 +48,17 @@ pub trait LlmClient: Send + Sync {
     /// as one `Done` event.
     async fn chat_with_tools_stream(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<BoxStream<'static, Result<LlmStreamEvent>>>;
 
     /// Batch embedding generation. Dimensionality is provider/model specific.
-    async fn generate_embeddings(&self, texts: &[&str]) -> Result<Vec<Vec<f32>>>;
+    async fn generate_embeddings(&self, cx: &Context, texts: &[&str]) -> Result<Vec<Vec<f32>>>;
 
     /// One-shot vision description of an image. Used to convert images into
     /// plain text for the hybrid-mode conversation flow.
-    async fn describe_image(&self, image_base64: &str) -> Result<String>;
+    async fn describe_image(&self, cx: &Context, image_base64: &str) -> Result<String>;
 
     /// Enumerate available models with their capabilities.
     async fn list_models(&self) -> Result<Vec<ModelCapabilities>>;

src/ai/local_llm.rs

@@ -47,12 +47,24 @@ impl LocalLlm {
     /// pick query vs document — retrieval models treat the two sides
     /// differently and an unmarked embed invites prefix-mismatch bugs.
     pub async fn embed_query(&self, text: &str) -> Result<Vec<f32>> {
-        super::embed_query(&self.ollama, self.llamacpp.as_deref(), text).await
+        super::embed_query(
+            &opentelemetry::Context::new(),
+            &self.ollama,
+            self.llamacpp.as_deref(),
+            text,
+        )
+        .await
     }
 
     /// Embed corpus text (applies `EMBED_DOCUMENT_PREFIX`).
     pub async fn embed_document(&self, text: &str) -> Result<Vec<f32>> {
-        super::embed_document(&self.ollama, self.llamacpp.as_deref(), text).await
+        super::embed_document(
+            &opentelemetry::Context::new(),
+            &self.ollama,
+            self.llamacpp.as_deref(),
+            text,
+        )
+        .await
     }
 
     /// Single-shot local text generation via the `LLM_BACKEND`-selected
@@ -60,14 +72,23 @@ impl LocalLlm {
     pub async fn generate(&self, prompt: &str, system: Option<&str>) -> Result<String> {
         if super::local_backend_is_llamacpp() {
             if let Some(lc) = self.llamacpp.as_deref() {
-                return <LlamaCppClient as LlmClient>::generate(lc, prompt, system, None).await;
+                return <LlamaCppClient as LlmClient>::generate(
+                    lc,
+                    &opentelemetry::Context::new(),
+                    prompt,
+                    system,
+                    None,
+                )
+                .await;
             }
             anyhow::bail!(
                 "LLM_BACKEND=llamacpp but LlamaCppClient is unconfigured — \
                  set LLAMA_SWAP_URL or switch to LLM_BACKEND=ollama"
             );
         }
-        self.ollama.generate(prompt, system).await
+        self.ollama
+            .generate(&opentelemetry::Context::new(), prompt, system)
+            .await
     }
 
     /// Label identifying which backend + model produces embeddings right

src/ai/mod.rs

@@ -115,24 +115,26 @@ fn embed_prefix(key: &str) -> String {
 /// `Instruct: <task>\nQuery: `. Must pair with the document prefix the
 /// corpus was embedded with or similarity degrades.
 pub async fn embed_query(
+    cx: &opentelemetry::Context,
     ollama: &OllamaClient,
     llamacpp: Option<&LlamaCppClient>,
     text: &str,
 ) -> anyhow::Result<Vec<f32>> {
     let prefixed = format!("{}{}", embed_prefix("EMBED_QUERY_PREFIX"), text);
-    embed_one(ollama, llamacpp, &prefixed).await
+    embed_one(cx, ollama, llamacpp, &prefixed).await
 }
 
 /// Embed corpus text (the stored side of retrieval). Applies
 /// `EMBED_DOCUMENT_PREFIX` (default empty; nomic v1.5 wants
 /// `search_document: `, Qwen3-Embedding wants none).
 pub async fn embed_document(
+    cx: &opentelemetry::Context,
     ollama: &OllamaClient,
     llamacpp: Option<&LlamaCppClient>,
     text: &str,
 ) -> anyhow::Result<Vec<f32>> {
     let prefixed = format!("{}{}", embed_prefix("EMBED_DOCUMENT_PREFIX"), text);
-    embed_one(ollama, llamacpp, &prefixed).await
+    embed_one(cx, ollama, llamacpp, &prefixed).await
 }
 
 /// Embed a batch of strings via the configured local backend. Routes
@@ -140,30 +142,32 @@ pub async fn embed_document(
 /// configured), else Ollama. See [`local_backend_is_llamacpp`] for the
 /// rationale on consistency.
 pub async fn embed_many(
+    cx: &opentelemetry::Context,
     ollama: &OllamaClient,
     llamacpp: Option<&LlamaCppClient>,
     texts: &[&str],
 ) -> anyhow::Result<Vec<Vec<f32>>> {
     if local_backend_is_llamacpp() {
         if let Some(lc) = llamacpp {
-            return <LlamaCppClient as LlmClient>::generate_embeddings(lc, texts).await;
+            return <LlamaCppClient as LlmClient>::generate_embeddings(lc, cx, texts).await;
         }
         anyhow::bail!(
             "LLM_BACKEND=llamacpp but LlamaCppClient is unconfigured — \
              set LLAMA_SWAP_URL or switch to LLM_BACKEND=ollama"
         );
     }
-    ollama.generate_embeddings(texts).await
+    ollama.generate_embeddings(cx, texts).await
 }
 
 /// Embed one string via the configured local backend. Single-text
 /// convenience over [`embed_many`].
 pub async fn embed_one(
+    cx: &opentelemetry::Context,
     ollama: &OllamaClient,
     llamacpp: Option<&LlamaCppClient>,
     text: &str,
 ) -> anyhow::Result<Vec<f32>> {
-    let mut vecs = embed_many(ollama, llamacpp, &[text]).await?;
+    let mut vecs = embed_many(cx, ollama, llamacpp, &[text]).await?;
     vecs.pop()
         .ok_or_else(|| anyhow::anyhow!("embedding backend returned no embeddings"))
 }

src/ai/nl_query.rs

@@ -259,10 +259,13 @@ pub async fn translate_nl_query(
     nl: &str,
     tag_vocab: &[(i32, String)],
     today: chrono::NaiveDate,
+    cx: &opentelemetry::Context,
 ) -> Result<StructuredQuery> {
     let system = build_system_prompt(tag_vocab, today);
     let messages = vec![ChatMessage::system(system), ChatMessage::user(nl)];
-    let (msg, _, _) = client.chat_with_tools(messages, Vec::<Tool>::new()).await?;
+    let (msg, _, _) = client
+        .chat_with_tools(cx, messages, Vec::<Tool>::new())
+        .await?;
     parse_response(&msg.content, tag_vocab)
 }
 

src/ai/ollama.rs

@@ -1,6 +1,9 @@
-use anyhow::{Context, Result};
+use anyhow::Context as _;
+use anyhow::Result;
 use async_trait::async_trait;
 use chrono::NaiveDate;
+use opentelemetry::trace::{Span, Status};
+use opentelemetry::{Context, KeyValue};
 use reqwest::Client;
 use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
@@ -365,6 +368,7 @@ impl OllamaClient {
 
     async fn try_generate(
         &self,
+        mut span: Option<&mut opentelemetry::global::BoxedSpan>,
         url: &str,
         model: &str,
         prompt: &str,
@@ -382,15 +386,30 @@ impl OllamaClient {
             think,
         };
 
-        let response = self
+        let mut request_builder = self
             .client
             .post(format!("{}/api/generate", url))
-            .json(&request)
+            .json(&request);
-            .send()
+        if let Some(s) = span.as_mut() {
-            .await?;
+            request_builder = crate::otel::with_traceparent(request_builder, s.span_context());
+        }
+        let response = match request_builder.send().await {
+            Ok(r) => r,
+            Err(e) => {
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("Ollama request failed"));
+                }
+                return Err(e.into());
+            }
+        };
 
         if !response.status().is_success() {
             let status = response.status();
+            crate::otel::record_http_failure(
+                span.as_deref_mut(),
+                status.as_u16(),
+                &format!("Ollama request failed: {status}"),
+            );
             let error_body = response.text().await.unwrap_or_default();
             return Err(anyhow::anyhow!(
                 "Ollama request failed: {} - {}",
@@ -398,8 +417,26 @@ impl OllamaClient {
                 error_body
             ));
         }
+        crate::otel::set_http_status(span.as_deref_mut(), response.status().as_u16());
 
-        let result: OllamaResponse = response.json().await?;
+        let result: OllamaResponse = match response.json().await {
+            Ok(r) => r,
+            Err(e) => {
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("parsing Ollama response failed"));
+                }
+                return Err(e.into());
+            }
+        };
+        if let Some(s) = span.as_mut() {
+            if let Some(pt) = result.prompt_eval_count {
+                s.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = result.eval_count {
+                s.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            s.set_status(Status::Ok);
+        }
         log_chat_metrics(
             result.prompt_eval_count,
             result.prompt_eval_duration,
@@ -409,32 +446,55 @@ impl OllamaClient {
         Ok(result.response)
     }
 
-    pub async fn generate(&self, prompt: &str, system: Option<&str>) -> Result<String> {
+    /// Start the `llm.request` span for an outgoing Ollama call.
-        self.generate_with_images(prompt, system, None).await
+    fn start_request_span(
+        &self,
+        cx: &Context,
+        operation: &str,
+        model: &str,
+    ) -> opentelemetry::global::BoxedSpan {
+        crate::otel::start_llm_request_span(cx, "ollama", operation, model)
+    }
+
+    pub async fn generate(
+        &self,
+        cx: &Context,
+        prompt: &str,
+        system: Option<&str>,
+    ) -> Result<String> {
+        self.generate_with_images(cx, prompt, system, None).await
     }
 
     #[allow(dead_code)]
-    pub async fn generate_no_think(&self, prompt: &str, system: Option<&str>) -> Result<String> {
+    pub async fn generate_no_think(
-        self.generate_with_options(prompt, system, None, Some(false))
+        &self,
+        cx: &Context,
+        prompt: &str,
+        system: Option<&str>,
+    ) -> Result<String> {
+        self.generate_with_options(cx, prompt, system, None, Some(false), "generate")
             .await
     }
 
     pub async fn generate_with_images(
         &self,
+        cx: &Context,
         prompt: &str,
         system: Option<&str>,
         images: Option<Vec<String>>,
     ) -> Result<String> {
-        self.generate_with_options(prompt, system, images, None)
+        self.generate_with_options(cx, prompt, system, images, None, "generate")
             .await
     }
 
     async fn generate_with_options(
         &self,
+        cx: &Context,
         prompt: &str,
         system: Option<&str>,
         images: Option<Vec<String>>,
         think: Option<bool>,
+        operation: &str,
     ) -> Result<String> {
         log::debug!("=== Ollama Request ===");
         log::debug!("Primary model: {}", self.primary_model);
@@ -447,6 +507,10 @@ impl OllamaClient {
         }
         log::debug!("=====================");
 
+        // Span for the whole generate attempt; the model attribute
+        // reflects whichever server actually answered.
+        let mut span = self.start_request_span(cx, operation, &self.primary_model);
+
         // Try primary server first with primary model
         log::info!(
             "Attempting to generate with primary server: {} (model: {})",
@@ -455,6 +519,7 @@ impl OllamaClient {
         );
         let primary_result = self
             .try_generate(
+                Some(&mut span),
                 &self.primary_url,
                 &self.primary_model,
                 prompt,
@@ -483,8 +548,10 @@ impl OllamaClient {
                         fallback_url,
                         fallback_model
                     );
+                    span.set_attribute(KeyValue::new("llm.model", fallback_model.to_string()));
                     match self
                         .try_generate(
+                            Some(&mut span),
                             fallback_url,
                             fallback_model,
                             prompt,
@@ -531,6 +598,7 @@ impl OllamaClient {
     /// Generate a title for a single photo based on its generated summary
     pub async fn generate_photo_title(
         &self,
+        cx: &Context,
         summary: &str,
         custom_system: Option<&str>,
     ) -> Result<String> {
@@ -546,7 +614,7 @@ Capture the key moment or theme. Return ONLY the title, nothing else."#,
         let system = custom_system.unwrap_or("You are my long term memory assistant. Use only the information provided. Do not invent details.");
 
         let title = self
-            .generate_with_images(&prompt, Some(system), None)
+            .generate_with_images(cx, &prompt, Some(system), None)
             .await?;
         // Models decorate despite "Return ONLY the title": quotes, bold
         // markers, sometimes a "Title:" label.
@@ -563,6 +631,7 @@ Capture the key moment or theme. Return ONLY the title, nothing else."#,
     /// Generate a summary for a single photo based on its context
     pub async fn generate_photo_summary(
         &self,
+        cx: &Context,
         date: NaiveDate,
         location: Option<&str>,
         contact: Option<&str>,
@@ -645,21 +714,32 @@ Analyze the image and use specific details from both the visual content and the
         let system = custom_system.unwrap_or("You are a memory refreshing assistant who is able to provide insights through analyzing past conversations. Use only the information provided. Do not invent details.");
 
         let images = image_base64.map(|img| vec![img]);
-        self.generate_with_images(&prompt, Some(system), images)
+        self.generate_with_images(cx, &prompt, Some(system), images)
             .await
     }
 
     /// Generate a brief visual description of a photo for use in RAG query enrichment.
     /// Returns 1-2 sentences describing people, location, and activity visible in the image.
     /// Only called when the model has vision capabilities.
-    pub async fn generate_photo_description(&self, image_base64: &str) -> Result<String> {
+    pub async fn generate_photo_description(
+        &self,
+        cx: &Context,
+        image_base64: &str,
+    ) -> Result<String> {
         let prompt = "Briefly describe what you see in this image in 1-2 sentences. \
                       Focus on the people, location, and activity.";
         let system = "You are a scene description assistant. Be concise and factual.";
         let images = vec![image_base64.to_string()];
 
         let description = self
-            .generate_with_images(prompt, Some(system), Some(images))
+            .generate_with_options(
+                cx,
+                prompt,
+                Some(system),
+                Some(images),
+                None,
+                "describe_image",
+            )
             .await?;
 
         Ok(description.trim().to_string())
@@ -672,9 +752,11 @@ Analyze the image and use specific details from both the visual content and the
     /// iteration of the agent loop.
     pub async fn chat_with_tools(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<(ChatMessage, Option<i32>, Option<i32>)> {
+        let mut span = self.start_request_span(cx, "chat", &self.primary_model);
         let order = self.attempt_order();
         let mut errors: Vec<String> = Vec::new();
 
@@ -685,8 +767,11 @@ Analyze the image and use specific details from both the visual content and the
                 url,
                 model
             );
+            if *label == "fallback" {
+                span.set_attribute(KeyValue::new("llm.model", model.to_string()));
+            }
             match self
-                .try_chat_with_tools(url, messages.clone(), tools.clone())
+                .try_chat_with_tools(Some(&mut span), url, messages.clone(), tools.clone())
                 .await
             {
                 Ok(result) => {
@@ -728,18 +813,22 @@ Analyze the image and use specific details from both the visual content and the
     /// any, live on the final message).
     pub async fn chat_with_tools_stream(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<BoxStream<'static, Result<LlmStreamEvent>>> {
         // Same preference logic as `chat_with_tools`. Only the initial
         // connection is retried across servers — once the stream begins,
-        // mid-stream errors propagate to the caller.
+        // mid-stream errors propagate to the caller. Each attempt gets its
+        // own `llm.request` span, created inside
+        // `try_chat_with_tools_stream` so the winning attempt's span moves
+        // into the stream and is finalized when the response drains.
         let order = self.attempt_order();
         let mut last_err: Option<anyhow::Error> = None;
 
         for (label, url, _model) in &order {
             match self
-                .try_chat_with_tools_stream(url, messages.clone(), tools.clone())
+                .try_chat_with_tools_stream(cx, url, messages.clone(), tools.clone())
                 .await
             {
                 Ok(s) => {
@@ -759,6 +848,7 @@ Analyze the image and use specific details from both the visual content and the
 
     async fn try_chat_with_tools_stream(
         &self,
+        cx: &Context,
         base_url: &str,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
@@ -773,6 +863,11 @@ Analyze the image and use specific details from both the visual content and the
         };
         let options = self.build_options();
 
+        // The span outlives this method — the stream yields until the
+        // response fully drains, so the owned span moves into the
+        // generator and is finalized there.
+        let mut span = self.start_request_span(cx, "chat_stream", model);
+
         let request_body = OllamaChatRequest {
             model,
             messages: &messages,
@@ -781,16 +876,26 @@ Analyze the image and use specific details from both the visual content and the
             options,
         };
 
-        let response = self
+        let response = {
-            .client
+            let mut rb = self.client.post(&url).json(&request_body);
-            .post(&url)
+            rb = crate::otel::with_traceparent(rb, span.span_context());
-            .json(&request_body)
+            match rb.send().await {
-            .send()
+                Ok(r) => r,
-            .await
+                Err(e) => {
-            .with_context(|| format!("Failed to connect to Ollama at {}", url))?;
+                    span.set_status(Status::error("Ollama stream connect failed"));
+                    return Err(e)
+                        .with_context(|| format!("Failed to connect to Ollama at {}", url));
+                }
+            }
+        };
 
         if !response.status().is_success() {
             let status = response.status();
+            crate::otel::record_http_failure(
+                Some(&mut span),
+                status.as_u16(),
+                &format!("Ollama stream request failed: {status}"),
+            );
             let body = response.text().await.unwrap_or_default();
             anyhow::bail!(
                 "Ollama stream request failed with status {}: {}",
@@ -798,6 +903,7 @@ Analyze the image and use specific details from both the visual content and the
                 body
             );
         }
+        crate::otel::set_http_status(Some(&mut span), response.status().as_u16());
 
         // Ollama streams NDJSON: each line is a full `OllamaStreamChunk`.
         // We buffer partial lines across chunks from the byte stream.
@@ -879,6 +985,13 @@ Analyze the image and use specific details from both the visual content and the
                 eval_count,
                 eval_duration,
             );
+            if let Some(pt) = prompt_eval_count {
+                span.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = eval_count {
+                span.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            span.set_status(Status::Ok);
             let message = ChatMessage {
                 role,
                 content: accumulated,
@@ -897,6 +1010,7 @@ Analyze the image and use specific details from both the visual content and the
 
     async fn try_chat_with_tools(
         &self,
+        mut span: Option<&mut opentelemetry::global::BoxedSpan>,
         base_url: &str,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
@@ -924,16 +1038,30 @@ Analyze the image and use specific details from both the visual content and the
             .unwrap_or_else(|e| format!("<serialization error: {}>", e));
         log::debug!("chat_with_tools request body: {}", request_json);
 
-        let response = self
+        let response = {
-            .client
+            let mut rb = self.client.post(&url).json(&request_body);
-            .post(&url)
+            if let Some(s) = span.as_mut() {
-            .json(&request_body)
+                rb = crate::otel::with_traceparent(rb, s.span_context());
-            .send()
+            }
-            .await
+            match rb.send().await {
-            .with_context(|| format!("Failed to connect to Ollama at {}", url))?;
+                Ok(r) => r,
+                Err(e) => {
+                    if let Some(s) = span.as_mut() {
+                        s.set_status(Status::error("Ollama chat connect failed"));
+                    }
+                    return Err(e)
+                        .with_context(|| format!("Failed to connect to Ollama at {}", url));
+                }
+            }
+        };
 
         if !response.status().is_success() {
             let status = response.status();
+            crate::otel::record_http_failure(
+                span.as_deref_mut(),
+                status.as_u16(),
+                &format!("Ollama chat request failed: {status}"),
+            );
             let body = response.text().await.unwrap_or_default();
             // warn, not error — the outer `chat_with_tools` may recover via
             // the fallback server. When both fail, the outer layer emits the
@@ -950,11 +1078,26 @@ Analyze the image and use specific details from both the visual content and the
                 body
             );
         }
+        crate::otel::set_http_status(span.as_deref_mut(), response.status().as_u16());
 
-        let chat_response: OllamaChatResponse = response
+        let chat_response: OllamaChatResponse = match response.json().await {
-            .json()
+            Ok(r) => r,
-            .await
+            Err(e) => {
-            .with_context(|| "Failed to parse Ollama chat response")?;
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("parsing Ollama chat response failed"));
+                }
+                return Err(e).with_context(|| "Failed to parse Ollama chat response");
+            }
+        };
+        if let Some(s) = span.as_mut() {
+            if let Some(pt) = chat_response.prompt_eval_count {
+                s.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = chat_response.eval_count {
+                s.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            s.set_status(Status::Ok);
+        }
 
         // Log performance counters returned by Ollama. Durations are
         // reported in nanoseconds; we render ms + tokens/sec for skim-ability
@@ -976,8 +1119,8 @@ Analyze the image and use specific details from both the visual content and the
 
     /// Generate an embedding vector for text using nomic-embed-text:v1.5
     /// Returns a 768-dimensional vector as Vec<f32>
-    pub async fn generate_embedding(&self, text: &str) -> Result<Vec<f32>> {
+    pub async fn generate_embedding(&self, cx: &Context, text: &str) -> Result<Vec<f32>> {
-        let embeddings = self.generate_embeddings(&[text]).await?;
+        let embeddings = self.generate_embeddings(cx, &[text]).await?;
         embeddings
             .into_iter()
             .next()
@@ -987,9 +1130,11 @@ Analyze the image and use specific details from both the visual content and the
     /// Generate embeddings for multiple texts in a single API call (batch mode)
     /// Returns a vector of 768-dimensional vectors
     /// This is much more efficient than calling generate_embedding multiple times
-    pub async fn generate_embeddings(&self, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
+    pub async fn generate_embeddings(&self, cx: &Context, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
         let embedding_model = EMBEDDING_MODEL;
 
+        let mut span = self.start_request_span(cx, "embeddings", embedding_model);
+
         log::debug!("=== Ollama Batch Embedding Request ===");
         log::debug!("Model: {}", embedding_model);
         log::debug!("Batch size: {} texts", texts.len());
@@ -1003,7 +1148,7 @@ Analyze the image and use specific details from both the visual content and the
             embedding_model
         );
         let primary_result = self
-            .try_generate_embeddings(&self.primary_url, embedding_model, texts)
+            .try_generate_embeddings(Some(&mut span), &self.primary_url, embedding_model, texts)
             .await;
 
         let embeddings = match primary_result {
@@ -1026,7 +1171,12 @@ Analyze the image and use specific details from both the visual content and the
                         embedding_model
                     );
                     match self
-                        .try_generate_embeddings(fallback_url, embedding_model, texts)
+                        .try_generate_embeddings(
+                            Some(&mut span),
+                            fallback_url,
+                            embedding_model,
+                            texts,
+                        )
                         .await
                     {
                         Ok(embeddings) => {
@@ -1073,6 +1223,7 @@ Analyze the image and use specific details from both the visual content and the
     /// Internal helper to try generating embeddings for multiple texts from a specific server
     async fn try_generate_embeddings(
         &self,
+        mut span: Option<&mut opentelemetry::global::BoxedSpan>,
         url: &str,
         model: &str,
         texts: &[&str],
@@ -1082,15 +1233,32 @@ Analyze the image and use specific details from both the visual content and the
             input: texts.iter().map(|s| s.to_string()).collect(),
         };
 
-        let response = self
+        let response = {
-            .client
+            let mut rb = self
-            .post(format!("{}/api/embed", url))
+                .client
-            .json(&request)
+                .post(format!("{}/api/embed", url))
-            .send()
+                .json(&request);
-            .await?;
+            if let Some(s) = span.as_mut() {
+                rb = crate::otel::with_traceparent(rb, s.span_context());
+            }
+            match rb.send().await {
+                Ok(r) => r,
+                Err(e) => {
+                    if let Some(s) = span.as_mut() {
+                        s.set_status(Status::error("Ollama embedding request failed"));
+                    }
+                    return Err(e.into());
+                }
+            }
+        };
 
         if !response.status().is_success() {
             let status = response.status();
+            crate::otel::record_http_failure(
+                span.as_deref_mut(),
+                status.as_u16(),
+                &format!("Ollama batch embedding request failed: {status}"),
+            );
             let error_body = response.text().await.unwrap_or_default();
             return Err(anyhow::anyhow!(
                 "Ollama batch embedding request failed: {} - {}",
@@ -1098,8 +1266,20 @@ Analyze the image and use specific details from both the visual content and the
                 error_body
             ));
         }
+        crate::otel::set_http_status(span.as_deref_mut(), response.status().as_u16());
 
-        let result: OllamaEmbedResponse = response.json().await?;
+        let result: OllamaEmbedResponse = match response.json().await {
+            Ok(r) => r,
+            Err(e) => {
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("parsing Ollama embed response failed"));
+                }
+                return Err(e.into());
+            }
+        };
+        if let Some(s) = span.as_mut() {
+            s.set_status(Status::Ok);
+        }
         Ok(result.embeddings)
     }
 }
@@ -1108,35 +1288,38 @@ Analyze the image and use specific details from both the visual content and the
 impl LlmClient for OllamaClient {
     async fn generate(
         &self,
+        cx: &Context,
         prompt: &str,
         system: Option<&str>,
         images: Option<Vec<String>>,
     ) -> Result<String> {
-        self.generate_with_images(prompt, system, images).await
+        self.generate_with_images(cx, prompt, system, images).await
     }
 
     async fn chat_with_tools(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<(ChatMessage, Option<i32>, Option<i32>)> {
-        OllamaClient::chat_with_tools(self, messages, tools).await
+        OllamaClient::chat_with_tools(self, cx, messages, tools).await
     }
 
     async fn chat_with_tools_stream(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<BoxStream<'static, Result<LlmStreamEvent>>> {
-        OllamaClient::chat_with_tools_stream(self, messages, tools).await
+        OllamaClient::chat_with_tools_stream(self, cx, messages, tools).await
     }
 
-    async fn generate_embeddings(&self, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
+    async fn generate_embeddings(&self, cx: &Context, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
-        OllamaClient::generate_embeddings(self, texts).await
+        OllamaClient::generate_embeddings(self, cx, texts).await
     }
 
-    async fn describe_image(&self, image_base64: &str) -> Result<String> {
+    async fn describe_image(&self, cx: &Context, image_base64: &str) -> Result<String> {
-        self.generate_photo_description(image_base64).await
+        self.generate_photo_description(cx, image_base64).await
     }
 
     async fn list_models(&self) -> Result<Vec<ModelCapabilities>> {

src/ai/openrouter.rs

@@ -2,8 +2,11 @@
 // the translation helpers directly.
 #![allow(dead_code)]
 
-use anyhow::{Context, Result, anyhow, bail};
+use anyhow::Context as _;
+use anyhow::{Result, anyhow, bail};
 use async_trait::async_trait;
+use opentelemetry::trace::{Span, Status};
+use opentelemetry::{Context, KeyValue};
 use reqwest::Client;
 use serde::Deserialize;
 use serde_json::{Value, json};
@@ -14,6 +17,7 @@ use std::time::{Duration, Instant};
 use crate::ai::llm_client::{
     ChatMessage, LlmClient, LlmStreamEvent, ModelCapabilities, Tool, ToolCall, ToolCallFunction,
 };
+
 use futures::stream::{BoxStream, StreamExt};
 
 const DEFAULT_BASE_URL: &str = "https://openrouter.ai/api/v1";
@@ -144,6 +148,19 @@ impl OpenRouterClient {
         b
     }
 
+    /// Start the `llm.request` span for an outgoing call. `operation` is
+    /// `chat`, `embeddings`, or `describe` — the call shape, not the entry
+    /// point (a `describe_image` funneled through `chat_with_tools` is
+    /// still a `describe`).
+    fn start_request_span(
+        &self,
+        cx: &Context,
+        operation: &str,
+        model: &str,
+    ) -> opentelemetry::global::BoxedSpan {
+        crate::otel::start_llm_request_span(cx, "openrouter", operation, model)
+    }
+
     /// Translate canonical messages to the OpenAI-compatible wire shape.
     ///
     /// Walks in order so it can attach `tool_call_id` to `role=tool` messages
@@ -312,6 +329,7 @@ impl OpenRouterClient {
 impl LlmClient for OpenRouterClient {
     async fn generate(
         &self,
+        cx: &Context,
         prompt: &str,
         system: Option<&str>,
         images: Option<Vec<String>>,
@@ -324,90 +342,34 @@ impl LlmClient for OpenRouterClient {
         user.images = images;
         messages.push(user);
 
-        let (reply, _, _) = self.chat_with_tools(messages, Vec::new()).await?;
+        // Delegates to chat_with_tools, which starts the llm.request span
+        // (operation=chat) — one span per HTTP request.
+        let (reply, _, _) = self.chat_with_tools(cx, messages, Vec::new()).await?;
         Ok(reply.content)
     }
 
     async fn chat_with_tools(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<(ChatMessage, Option<i32>, Option<i32>)> {
-        let url = format!("{}/chat/completions", self.base_url);
+        let mut span = self.start_request_span(cx, "chat", &self.primary_model);
-        let mut body = serde_json::Map::new();
+        self.chat_with_tools_impl(Some(&mut span), messages, tools)
-        body.insert("model".into(), Value::String(self.primary_model.clone()));
-        body.insert(
-            "messages".into(),
-            Value::Array(Self::messages_to_openai(&messages)),
-        );
-        body.insert("stream".into(), Value::Bool(false));
-        if !tools.is_empty() {
-            body.insert(
-                "tools".into(),
-                serde_json::to_value(&tools).context("serializing tools")?,
-            );
-        }
-        for (k, v) in self.build_options() {
-            body.insert(k.into(), v);
-        }
-
-        log::info!(
-            "OpenRouter chat_with_tools: model={} messages={} tools={}",
-            self.primary_model,
-            messages.len(),
-            tools.len()
-        );
-
-        let resp = self
-            .authed(self.client.post(&url))
-            .json(&Value::Object(body))
-            .send()
             .await
-            .with_context(|| format!("POST {} failed", url))?;
-
-        if !resp.status().is_success() {
-            let status = resp.status();
-            let body = resp.text().await.unwrap_or_default();
-            bail!("OpenRouter chat request failed: {} — {}", status, body);
-        }
-
-        let parsed: Value = resp.json().await.context("parsing chat response")?;
-        let choice = parsed
-            .get("choices")
-            .and_then(|v| v.as_array())
-            .and_then(|a| a.first())
-            .ok_or_else(|| {
-                anyhow!(
-                    "response missing choices[0]: {}",
-                    extract_openrouter_error_detail(&parsed)
-                )
-            })?;
-        let msg = choice.get("message").ok_or_else(|| {
-            anyhow!(
-                "choices[0] missing message: {}",
-                extract_openrouter_error_detail(&parsed)
-            )
-        })?;
-        let chat_msg = Self::openai_message_to_chat(msg)?;
-
-        let usage = parsed.get("usage");
-        let prompt_tokens = usage
-            .and_then(|u| u.get("prompt_tokens"))
-            .and_then(|v| v.as_i64())
-            .map(|n| n as i32);
-        let completion_tokens = usage
-            .and_then(|u| u.get("completion_tokens"))
-            .and_then(|v| v.as_i64())
-            .map(|n| n as i32);
-
-        Ok((chat_msg, prompt_tokens, completion_tokens))
     }
 
     async fn chat_with_tools_stream(
         &self,
+        cx: &Context,
         messages: Vec<ChatMessage>,
         tools: Vec<Tool>,
     ) -> Result<BoxStream<'static, Result<LlmStreamEvent>>> {
+        // The span outlives this method — the stream yields until the
+        // response fully drains, so the owned span moves into the
+        // generator and is finalized there.
+        let mut span = self.start_request_span(cx, "chat_stream", &self.primary_model);
+
         let url = format!("{}/chat/completions", self.base_url);
         let mut body = serde_json::Map::new();
         body.insert("model".into(), Value::String(self.primary_model.clone()));
@@ -432,18 +394,26 @@ impl LlmClient for OpenRouterClient {
             body.insert(k.into(), v);
         }
 
-        let resp = self
+        let mut request = self
             .authed(self.client.post(&url))
-            .json(&Value::Object(body))
+            .json(&Value::Object(body));
+        request = crate::otel::with_traceparent(request, span.span_context());
+        let resp = request
             .send()
             .await
             .with_context(|| format!("POST {} failed", url))?;
 
         if !resp.status().is_success() {
             let status = resp.status();
+            crate::otel::record_http_failure(
+                Some(&mut span),
+                status.as_u16(),
+                &format!("OpenRouter stream request failed: {status}"),
+            );
             let body = resp.text().await.unwrap_or_default();
             bail!("OpenRouter stream request failed: {} — {}", status, body);
         }
+        crate::otel::set_http_status(Some(&mut span), resp.status().as_u16());
 
         // OpenAI-compat SSE stream. Each event is `data: <json>\n\n`, with
         // `data: [DONE]` signalling completion. Tool calls arrive as
@@ -467,6 +437,7 @@ impl LlmClient for OpenRouterClient {
                 let chunk = match chunk {
                     Ok(b) => b,
                     Err(e) => {
+                        span.set_status(Status::error(format!("stream read failed: {e}")));
                         yield Err(anyhow!("stream read failed: {}", e));
                         return;
                     }
@@ -517,6 +488,9 @@ impl LlmClient for OpenRouterClient {
                                 .get("completion_tokens")
                                 .and_then(|n| n.as_i64())
                                 .map(|n| n as i32);
+                            if let Some(cost) = usage.get("cost").and_then(|c| c.as_f64()) {
+                                span.set_attribute(KeyValue::new("llm.cost_usd", cost));
+                            }
                         }
 
                         let Some(choices) = v.get("choices").and_then(|c| c.as_array())
@@ -606,6 +580,13 @@ impl LlmClient for OpenRouterClient {
                 tool_calls,
                 images: None,
             };
+            if let Some(pt) = prompt_tokens {
+                span.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = completion_tokens {
+                span.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            span.set_status(Status::Ok);
             yield Ok(LlmStreamEvent::Done {
                 message,
                 prompt_eval_count: prompt_tokens,
@@ -616,25 +597,35 @@ impl LlmClient for OpenRouterClient {
         Ok(Box::pin(stream))
     }
 
-    async fn generate_embeddings(&self, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
+    async fn generate_embeddings(&self, cx: &Context, texts: &[&str]) -> Result<Vec<Vec<f32>>> {
+        let mut span = self.start_request_span(cx, "embeddings", &self.embedding_model);
         let url = format!("{}/embeddings", self.base_url);
         let body = json!({
             "model": self.embedding_model,
             "input": texts,
         });
 
-        let resp = self
+        let mut request = self.authed(self.client.post(&url)).json(&body);
-            .authed(self.client.post(&url))
+        request = crate::otel::with_traceparent(request, span.span_context());
-            .json(&body)
+        let resp = request
             .send()
             .await
-            .with_context(|| format!("POST {} failed", url))?;
+            .with_context(|| format!("POST {} failed", url))
+            .inspect_err(|_| {
+                span.set_status(Status::error("OpenRouter embedding request failed"));
+            })?;
 
         if !resp.status().is_success() {
             let status = resp.status();
+            crate::otel::record_http_failure(
+                Some(&mut span),
+                status.as_u16(),
+                &format!("OpenRouter embedding request failed: {status}"),
+            );
             let body = resp.text().await.unwrap_or_default();
             bail!("OpenRouter embedding request failed: {} — {}", status, body);
         }
+        crate::otel::set_http_status(Some(&mut span), resp.status().as_u16());
 
         #[derive(Deserialize)]
         struct EmbedResponse {
@@ -646,18 +637,28 @@ impl LlmClient for OpenRouterClient {
         }
 
         let parsed: EmbedResponse = resp.json().await.context("parsing embed response")?;
+        span.set_status(Status::Ok);
         Ok(parsed.data.into_iter().map(|i| i.embedding).collect())
     }
 
-    async fn describe_image(&self, image_base64: &str) -> Result<String> {
+    async fn describe_image(&self, cx: &Context, image_base64: &str) -> Result<String> {
+        // Own span with operation=describe_image — the funnel into
+        // chat_with_tools_impl keeps the request on a single span.
+        let mut span = self.start_request_span(cx, "describe_image", &self.primary_model);
         let prompt = "Briefly describe what you see in this image in 1-2 sentences. \
                       Focus on the people, location, and activity.";
-        self.generate(
+        let mut messages: Vec<ChatMessage> = Vec::new();
-            prompt,
+        messages.push(ChatMessage::system(
-            Some("You are a scene description assistant. Be concise and factual."),
+            "You are a scene description assistant. Be concise and factual.",
-            Some(vec![image_base64.to_string()]),
+        ));
-        )
+        let mut user = ChatMessage::user(prompt);
-        .await
+        user.images = Some(vec![image_base64.to_string()]);
+        messages.push(user);
+
+        let (reply, _, _) = self
+            .chat_with_tools_impl(Some(&mut span), messages, Vec::new())
+            .await?;
+        Ok(reply.content)
     }
 
     async fn list_models(&self) -> Result<Vec<ModelCapabilities>> {
@@ -711,6 +712,124 @@ impl LlmClient for OpenRouterClient {
     }
 }
 
+impl OpenRouterClient {
+    /// HTTP core of `chat_with_tools`. `span` is the `llm.request` span
+    /// started by the caller (or a funnelling entry point like
+    /// `describe_image`) so attributes from the response land on the right
+    /// operation label.
+    async fn chat_with_tools_impl(
+        &self,
+        mut span: Option<&mut opentelemetry::global::BoxedSpan>,
+        messages: Vec<ChatMessage>,
+        tools: Vec<Tool>,
+    ) -> Result<(ChatMessage, Option<i32>, Option<i32>)> {
+        let url = format!("{}/chat/completions", self.base_url);
+        let mut body = serde_json::Map::new();
+        body.insert("model".into(), Value::String(self.primary_model.clone()));
+        body.insert(
+            "messages".into(),
+            Value::Array(Self::messages_to_openai(&messages)),
+        );
+        body.insert("stream".into(), Value::Bool(false));
+        if !tools.is_empty() {
+            body.insert(
+                "tools".into(),
+                serde_json::to_value(&tools).context("serializing tools")?,
+            );
+        }
+        for (k, v) in self.build_options() {
+            body.insert(k.into(), v);
+        }
+
+        log::info!(
+            "OpenRouter chat_with_tools: model={} messages={} tools={}",
+            self.primary_model,
+            messages.len(),
+            tools.len()
+        );
+
+        let mut request = self
+            .authed(self.client.post(&url))
+            .json(&Value::Object(body));
+        if let Some(s) = span.as_mut() {
+            request = crate::otel::with_traceparent(request, s.span_context());
+        }
+        let resp = match request.send().await {
+            Ok(r) => r,
+            Err(e) => {
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("OpenRouter request failed"));
+                }
+                return Err(e).with_context(|| format!("POST {} failed", url));
+            }
+        };
+
+        if !resp.status().is_success() {
+            let status = resp.status();
+            crate::otel::record_http_failure(
+                span.as_deref_mut(),
+                status.as_u16(),
+                &format!("OpenRouter chat request failed: {status}"),
+            );
+            let body = resp.text().await.unwrap_or_default();
+            bail!("OpenRouter chat request failed: {} — {}", status, body);
+        }
+        crate::otel::set_http_status(span.as_deref_mut(), resp.status().as_u16());
+
+        let parsed: Value = match resp.json().await {
+            Ok(v) => v,
+            Err(e) => {
+                if let Some(s) = span.as_mut() {
+                    s.set_status(Status::error("parsing chat response failed"));
+                }
+                return Err(e).context("parsing chat response");
+            }
+        };
+        let choice = parsed
+            .get("choices")
+            .and_then(|v| v.as_array())
+            .and_then(|a| a.first())
+            .ok_or_else(|| {
+                anyhow!(
+                    "response missing choices[0]: {}",
+                    extract_openrouter_error_detail(&parsed)
+                )
+            })?;
+        let msg = choice.get("message").ok_or_else(|| {
+            anyhow!(
+                "choices[0] missing message: {}",
+                extract_openrouter_error_detail(&parsed)
+            )
+        })?;
+        let chat_msg = Self::openai_message_to_chat(msg)?;
+
+        let usage = parsed.get("usage");
+        let prompt_tokens = usage
+            .and_then(|u| u.get("prompt_tokens"))
+            .and_then(|v| v.as_i64())
+            .map(|n| n as i32);
+        let completion_tokens = usage
+            .and_then(|u| u.get("completion_tokens"))
+            .and_then(|v| v.as_i64())
+            .map(|n| n as i32);
+        // OpenRouter reports per-request USD cost in `usage.cost`.
+        if let Some(s) = span.as_mut() {
+            if let Some(pt) = prompt_tokens {
+                s.set_attribute(KeyValue::new("llm.tokens.input", pt as i64));
+            }
+            if let Some(ct) = completion_tokens {
+                s.set_attribute(KeyValue::new("llm.tokens.output", ct as i64));
+            }
+            if let Some(cost) = usage.and_then(|u| u.get("cost")).and_then(|c| c.as_f64()) {
+                s.set_attribute(KeyValue::new("llm.cost_usd", cost));
+            }
+            s.set_status(Status::Ok);
+        }
+
+        Ok((chat_msg, prompt_tokens, completion_tokens))
+    }
+}
+
 /// Extract a diagnostic fragment from an OpenRouter response body that
 /// doesn't match the expected `{choices: [...]}` shape. OpenRouter will
 /// sometimes return 200 OK with `{"error": {"message": "...", "code": ...}}`

src/ai/persona_chat.rs

@@ -26,7 +26,7 @@ use actix_web::{HttpRequest, HttpResponse, Responder, delete, get, post, web};
 use anyhow::{Context, Result, anyhow, bail};
 use chrono::Utc;
 use opentelemetry::KeyValue;
-use opentelemetry::trace::{Span, Status, Tracer};
+use opentelemetry::trace::{Span, Status, TraceContextExt, Tracer};
 use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 use std::sync::{Arc, Mutex as StdMutex};
@@ -685,6 +685,7 @@ impl PersonaChatSession {
                     entry_clone.clone(),
                     user_id,
                     library_id,
+                    &mut span,
                 )
                 .await;
             match result {
@@ -721,6 +722,7 @@ impl PersonaChatSession {
         entry: Arc<TurnEntry>,
         user_id: i32,
         _library_id: i32,
+        turn_span: &mut opentelemetry::global::BoxedSpan,
     ) -> Result<()> {
         // Per-conversation mutex — serializes this turn against a
         // concurrent rewind or branch switch on the same transcript.
@@ -808,6 +810,12 @@ impl PersonaChatSession {
             .resolve_backend(backend_kind, &overrides)
             .await?;
         let model_used = backend.model().to_string();
+        // llm.* attributes on the turn-level span (issue #130), plus the
+        // parent context for the child llm.request spans.
+        turn_span.set_attribute(KeyValue::new("llm.backend", backend.kind.as_str()));
+        turn_span.set_attribute(KeyValue::new("llm.model", model_used.clone()));
+        let turn_cx = opentelemetry::Context::new()
+            .with_remote_span_context(turn_span.span_context().clone());
 
         // Build a no-photo tool catalog. `current_gate_opts_for_persona`
         // already probes each tool's backing table for presence and
@@ -874,6 +882,7 @@ impl PersonaChatSession {
             &persona.persona_id,
             max_iterations,
             &entry,
+            &turn_cx,
         )
         .await
         .map_err(|e| anyhow!("persona chat loop failed: {e:?}"))?;
@@ -942,7 +951,7 @@ impl PersonaChatSession {
         // list falls back to the persona name while the title is empty.
         if needs_title {
             let title = self
-                .generate_title(&backend, &messages)
+                .generate_title(&backend, &messages, &turn_cx)
                 .await
                 .unwrap_or_else(|| fallback_title(&messages));
             if !title.is_empty() {
@@ -1047,6 +1056,7 @@ impl PersonaChatSession {
         &self,
         backend: &crate::ai::backend::ResolvedBackend,
         messages: &[ChatMessage],
+        cx: &opentelemetry::Context,
     ) -> Option<String> {
         let user_message = messages.iter().find(|m| m.role == "user")?;
         // The last assistant message with actual prose — earlier ones may be
@@ -1057,7 +1067,7 @@ impl PersonaChatSession {
             .find(|m| m.role == "assistant" && !m.content.trim().is_empty())?;
 
         let prompt = title_prompt(&user_message.content, &reply.content);
-        match backend.chat().generate(&prompt, None, None).await {
+        match backend.chat().generate(cx, &prompt, None, None).await {
             Ok(raw) => {
                 let title = sanitize_title(&crate::ai::llm_client::strip_think_blocks(&raw));
                 if title.is_empty() { None } else { Some(title) }

src/ai/sms_client.rs

@@ -324,6 +324,7 @@ impl SmsApiClient {
         &self,
         messages: &[SmsMessage],
         ollama: &OllamaClient,
+        cx: &opentelemetry::Context,
     ) -> Result<String> {
         if messages.is_empty() {
             return Ok(String::from("No messages on this day"));
@@ -356,6 +357,7 @@ Summary:"#,
 
         ollama
             .generate(
+                cx,
                 &prompt,
                 // Some("You are a summarizer for the purposes of jogging my memory and highlighting events and situations."),
                 Some("You are the keeper of memories, ingest the context and give me a casual summary of the moment."),

src/bin/test_daily_summary.rs

@@ -220,7 +220,9 @@ async fn main() -> Result<()> {
 
         println!("Generating summary...");
 
-        let summary = ollama.generate(&prompt, Some(system_prompt)).await?;
+        let summary = ollama
+            .generate(&opentelemetry::Context::new(), &prompt, Some(system_prompt))
+            .await?;
 
         println!("\n📝 GENERATED SUMMARY:");
         println!("─────────────────────────────────────────");
@@ -236,7 +238,9 @@ async fn main() -> Result<()> {
             );
 
             println!("\nGenerating embedding...");
-            let embedding = ollama.generate_embedding(&stripped).await?;
+            let embedding = ollama
+                .generate_embedding(&opentelemetry::Context::new(), &stripped)
+                .await?;
             println!("✓ Embedding generated ({} dimensions)", embedding.len());
 
             println!("Saving to database...");

src/database/models.rs

@@ -579,6 +579,19 @@ pub struct UserAiPrefs {
     /// default (`camera_tz::DEFAULT_DISPLAY_TZ`). The legacy
     /// `tz_offset_minutes` column stays frozen for rollback.
     pub display_tz: Option<String>,
+    /// Last/explicit insight + chat model choice (local id or OpenRouter id).
+    /// NULL = current per-request / env default behavior.
+    pub default_model: Option<String>,
+    /// `"local"` | `"hybrid"` — insight/chat backend. NULL = per-request /
+    /// hard default ("local").
+    pub default_backend: Option<String>,
+    /// Active persona for persona chat. NULL = hard default ("default").
+    pub default_persona_id: Option<String>,
+    /// Reel span: `"day"` | `"week"` | `"month"`. NULL = per-request /
+    /// hard default (day).
+    pub reel_span: Option<String>,
+    /// Reel segment cap. NULL = per-request / hard default.
+    pub reel_max_segments: Option<i32>,
 }
 
 #[derive(Insertable, Debug, Clone, serde::Deserialize, serde::Serialize)]
@@ -589,4 +602,9 @@ pub struct UpsertUserAiPrefs {
     pub library: Option<String>,
     pub updated_at: i64,
     pub display_tz: Option<String>,
+    pub default_model: Option<String>,
+    pub default_backend: Option<String>,
+    pub default_persona_id: Option<String>,
+    pub reel_span: Option<String>,
+    pub reel_max_segments: Option<i32>,
 }

src/database/schema.rs

@@ -293,6 +293,11 @@ diesel::table! {
         library -> Nullable<Text>,
         updated_at -> BigInt,
         display_tz -> Nullable<Text>,
+        default_model -> Nullable<Text>,
+        default_backend -> Nullable<Text>,
+        default_persona_id -> Nullable<Text>,
+        reel_span -> Nullable<Text>,
+        reel_max_segments -> Nullable<Integer>,
     }
 }
 

src/database/user_ai_prefs_dao.rs

@@ -100,6 +100,11 @@ impl UserAiPrefsDao for SqliteUserAiPrefsDao {
                     dsl::tz_offset_minutes.eq(&prefs.tz_offset_minutes),
                     dsl::library.eq(&prefs.library),
                     dsl::display_tz.eq(&prefs.display_tz),
+                    dsl::default_model.eq(&prefs.default_model),
+                    dsl::default_backend.eq(&prefs.default_backend),
+                    dsl::default_persona_id.eq(&prefs.default_persona_id),
+                    dsl::reel_span.eq(&prefs.reel_span),
+                    dsl::reel_max_segments.eq(&prefs.reel_max_segments),
                     dsl::updated_at.eq(&prefs.updated_at),
                 ))
                 .execute(connection.deref_mut())
@@ -147,6 +152,11 @@ mod tests {
             library: Some("1".to_string()),
             updated_at: now,
             display_tz: Some("America/New_York".to_string()),
+            default_model: Some("llama3.2:3b".to_string()),
+            default_backend: Some("hybrid".to_string()),
+            default_persona_id: Some("cabin".to_string()),
+            reel_span: Some("week".to_string()),
+            reel_max_segments: Some(24),
         };
         dao.upsert_prefs(&ctx(), &prefs).unwrap();
 
@@ -156,6 +166,11 @@ mod tests {
         assert_eq!(row.tz_offset_minutes, Some(-480));
         assert_eq!(row.library, Some("1".to_string()));
         assert_eq!(row.display_tz, Some("America/New_York".to_string()));
+        assert_eq!(row.default_model, Some("llama3.2:3b".to_string()));
+        assert_eq!(row.default_backend, Some("hybrid".to_string()));
+        assert_eq!(row.default_persona_id, Some("cabin".to_string()));
+        assert_eq!(row.reel_span, Some("week".to_string()));
+        assert_eq!(row.reel_max_segments, Some(24));
         assert_eq!(row.updated_at, now);
     }
 
@@ -171,6 +186,11 @@ mod tests {
             library: Some("1".to_string()),
             updated_at: now1,
             display_tz: Some("America/New_York".to_string()),
+            default_model: Some("llama3.2:3b".to_string()),
+            default_backend: Some("local".to_string()),
+            default_persona_id: None,
+            reel_span: Some("week".to_string()),
+            reel_max_segments: Some(24),
         };
         dao.upsert_prefs(&ctx(), &prefs1).unwrap();
 
@@ -180,6 +200,11 @@ mod tests {
             library: None,
             updated_at: now2,
             display_tz: None,
+            default_model: None,
+            default_backend: None,
+            default_persona_id: None,
+            reel_span: None,
+            reel_max_segments: None,
         };
         dao.upsert_prefs(&ctx(), &prefs2).unwrap();
 
@@ -188,6 +213,11 @@ mod tests {
         assert_eq!(row.tz_offset_minutes, Some(-300));
         assert!(row.library.is_none());
         assert!(row.display_tz.is_none());
+        // Owner-default columns are overwritten to NULL like the rest.
+        assert!(row.default_model.is_none());
+        assert!(row.default_backend.is_none());
+        assert!(row.reel_span.is_none());
+        assert!(row.reel_max_segments.is_none());
         assert_eq!(row.updated_at, now2);
     }
 
@@ -202,6 +232,11 @@ mod tests {
             library: None,
             updated_at: now,
             display_tz: Some("Europe/Berlin".to_string()),
+            default_model: None,
+            default_backend: None,
+            default_persona_id: Some("default".to_string()),
+            reel_span: Some("month".to_string()),
+            reel_max_segments: None,
         };
         dao.upsert_prefs(&ctx(), &prefs).unwrap();
 
@@ -210,5 +245,9 @@ mod tests {
         assert!(row.voice.is_none());
         assert!(row.library.is_none());
         assert_eq!(row.display_tz, Some("Europe/Berlin".to_string()));
+        assert!(row.default_model.is_none());
+        assert_eq!(row.default_persona_id, Some("default".to_string()));
+        assert_eq!(row.reel_span, Some("month".to_string()));
+        assert!(row.reel_max_segments.is_none());
     }
 }

src/knowledge.rs

@@ -859,7 +859,8 @@ async fn synthesize_merge<D: KnowledgeDao + 'static>(
 
     let ollama = app_state.ollama.clone();
     let model_used = ollama.primary_model.clone();
-    let proposed = match ollama.generate(&prompt, Some(system)).await {
+    let cx = opentelemetry::Context::current();
+    let proposed = match ollama.generate(&cx, &prompt, Some(system)).await {
         Ok(out) => {
             // Strip the framing models reach for even with explicit
             // "no preamble" guidance: leading "Merged description:"

src/memories.rs

@@ -108,6 +108,31 @@ pub enum MemoriesSpan {
     Month,
 }
 
+/// Wire/storage string for the span (`"day"` | `"week"` | `"month"`) — the
+/// same lowercase names as the serde rename, so a stored pref round-trips
+/// through `MemoriesRequest.span` unchanged.
+impl std::fmt::Display for MemoriesSpan {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.write_str(match self {
+            MemoriesSpan::Day => "day",
+            MemoriesSpan::Week => "week",
+            MemoriesSpan::Month => "month",
+        })
+    }
+}
+
+impl std::str::FromStr for MemoriesSpan {
+    type Err = String;
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        match s.trim().to_ascii_lowercase().as_str() {
+            "day" => Ok(MemoriesSpan::Day),
+            "week" => Ok(MemoriesSpan::Week),
+            "month" => Ok(MemoriesSpan::Month),
+            other => Err(format!("invalid span {other:?} (expected day|week|month)")),
+        }
+    }
+}
+
 #[derive(Deserialize)]
 pub struct MemoriesRequest {
     pub span: Option<MemoriesSpan>,

src/otel.rs

@@ -2,7 +2,9 @@ use actix_web::HttpRequest;
 use actix_web::http::header::HeaderMap;
 use opentelemetry::global::{BoxedSpan, BoxedTracer};
 use opentelemetry::propagation::TextMapPropagator;
-use opentelemetry::trace::{Span, Status, Tracer};
+use opentelemetry::trace::{
+    Span, SpanBuilder, SpanContext, SpanKind, Status, TraceContextExt, Tracer,
+};
 use opentelemetry::{Context, KeyValue, global};
 use opentelemetry_appender_log::OpenTelemetryLogBridge;
 use opentelemetry_otlp::WithExportConfig;
@@ -77,11 +79,78 @@ impl<'a> opentelemetry::propagation::Extractor for HeaderExtractor<'a> {
         self.0.keys().map(|k| k.as_str()).collect()
     }
 }
+/// Start a client span for a single LLM call. Every provider funnels its
+/// LlmClient methods through this so the span shape (name, `SpanKind::Client`,
+/// provider / operation / model attributes) is identical across
+/// openrouter / llamacpp / ollama. `operation` is one of `chat`,
+/// `chat_stream`, `embeddings`, `describe_image`, `generate`.
+pub fn start_llm_request_span(
+    cx: &Context,
+    provider: &str,
+    operation: &str,
+    model: &str,
+) -> BoxedSpan {
+    let builder = SpanBuilder::from_name("llm.request").with_kind(SpanKind::Client);
+    let mut span = global_tracer().build_with_context(builder, cx);
+    span.set_attribute(KeyValue::new("llm.provider", provider.to_string()));
+    span.set_attribute(KeyValue::new("llm.operation", operation.to_string()));
+    span.set_attribute(KeyValue::new("llm.model", model.to_string()));
+    span
+}
+
 pub fn extract_context_from_request(req: &HttpRequest) -> Context {
     let propagator = TraceContextPropagator::new();
     propagator.extract(&HeaderExtractor(req.headers()))
 }
 
+/// Build the W3C trace-context headers to attach to an outgoing HTTP
+/// request so a downstream service (llama-swap, Ollama) can parent its own
+/// spans onto this trace. Returns an empty vec when the context carries no
+/// valid span — e.g. debug builds, where the global tracer is a no-op and
+/// injecting an all-zero traceparent would produce garbage parent spans.
+pub fn traceparent_headers(context: &Context) -> Vec<(String, String)> {
+    if !context.span().span_context().is_valid() {
+        return Vec::new();
+    }
+    let mut headers: std::collections::HashMap<String, String> = std::collections::HashMap::new();
+    TraceContextPropagator::new().inject_context(context, &mut headers);
+    // The propagator writes an empty tracestate header when there is no
+    // state — drop empty-valued headers rather than sending noise.
+    headers.into_iter().filter(|(_, v)| !v.is_empty()).collect()
+}
+
+/// Attach the W3C traceparent header to an outgoing request so a downstream
+/// service parents its own spans onto this trace. No-op when the span context
+/// is not valid — e.g. debug builds, where the global tracer is a no-op and an
+/// all-zero traceparent would produce garbage parent spans.
+pub fn with_traceparent(
+    builder: reqwest::RequestBuilder,
+    span_context: &SpanContext,
+) -> reqwest::RequestBuilder {
+    let mut b = builder;
+    for (k, v) in
+        traceparent_headers(&Context::new().with_remote_span_context(span_context.clone()))
+    {
+        b = b.header(k, v);
+    }
+    b
+}
+
+/// Record the HTTP status code on the span (success or failure path).
+pub fn set_http_status(span: Option<&mut BoxedSpan>, status: u16) {
+    if let Some(s) = span {
+        s.set_attribute(KeyValue::new("http.status_code", status as i64));
+    }
+}
+
+/// Record an HTTP failure on the span: status code plus an error status.
+pub fn record_http_failure(span: Option<&mut BoxedSpan>, status: u16, message: &str) {
+    if let Some(s) = span {
+        s.set_attribute(KeyValue::new("http.status_code", status as i64));
+        s.set_status(Status::error(message.to_string()));
+    }
+}
+
 pub fn trace_db_call<F, O>(
     context: &Context,
     query_type: &str,
@@ -110,3 +179,59 @@ where
 
     result
 }
+
+#[cfg(test)]
+mod traceparent_tests {
+    use super::*;
+    use opentelemetry::trace::{
+        SpanContext, SpanId, TraceContextExt, TraceFlags, TraceId, TraceState,
+    };
+
+    /// A context carrying a span context with known ids — no global
+    /// provider needed, so tests can't pollute process state.
+    fn context_with_flags(flags: TraceFlags) -> Context {
+        let trace_id = TraceId::from_bytes([
+            0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e,
+            0x0f, 0x10,
+        ]);
+        let span_id = SpanId::from_bytes([0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08]);
+        let span_context = SpanContext::new(trace_id, span_id, flags, false, TraceState::default());
+        Context::new().with_remote_span_context(span_context)
+    }
+
+    #[test]
+    fn traceparent_headers_sampled_context() {
+        let cx = context_with_flags(TraceFlags::SAMPLED);
+        let headers = traceparent_headers(&cx);
+        assert_eq!(
+            headers,
+            vec![(
+                "traceparent".to_string(),
+                "00-0102030405060708090a0b0c0d0e0f10-0102030405060708-01".to_string(),
+            )]
+        );
+    }
+
+    #[test]
+    fn traceparent_headers_not_sampled_context() {
+        // A valid-but-unsampled context still propagates; the flags byte
+        // carries the state.
+        let cx = context_with_flags(TraceFlags::NOT_SAMPLED);
+        let headers = traceparent_headers(&cx);
+        assert_eq!(
+            headers,
+            vec![(
+                "traceparent".to_string(),
+                "00-0102030405060708090a0b0c0d0e0f10-0102030405060708-00".to_string(),
+            )]
+        );
+    }
+
+    #[test]
+    fn traceparent_headers_invalid_context_is_empty() {
+        // No span (debug-build no-op tracer) → no header, never an
+        // all-zero traceparent.
+        let cx = Context::new();
+        assert!(traceparent_headers(&cx).is_empty());
+    }
+}

src/prefs.rs

@@ -8,15 +8,19 @@
 //! `GET /timezones` — the IANA zone list from chrono-tz, for picker UIs
 //! (unauthenticated; static data).
 //!
-//! Scope: existing columns only (voice, library, display_tz) — no migration.
+//! Surface: all persisted prefs — the passive mirror (voice, library,
-//! The legacy `tz_offset_minutes` column is preserved verbatim on every
+//! display_tz) and the owner defaults (default_model, default_backend,
-//! write. `POST /reels`'s `capture_prefs` stays as the incidental writer;
+//! default_persona_id, reel_span, reel_max_segments). Consumers resolve
-//! both writers hit the same single row, so they're idempotent. Owner
+//! `request param → stored pref → hard default`; this endpoint is the only
-//! defaults (model/backend/persona/reels) are a separate follow-on ticket
+//! place the owner-default fields are written. `POST /reels`'s
-//! that adds the columns.
+//! `capture_prefs` stays as the incidental writer for the mirror fields
+//! (plus the effective reel span/segments); both writers hit the same
+//! single row, so they're idempotent. The legacy `tz_offset_minutes`
+//! column is preserved verbatim on every write.
 
 use actix_web::{HttpResponse, web};
 use chrono_tz::Tz;
+use serde::de::IntoDeserializer;
 use serde::{Deserialize, Serialize};
 use serde_json::json;
 use std::str::FromStr;
@@ -24,6 +28,7 @@ use std::time::{SystemTime, UNIX_EPOCH};
 
 use crate::data::Claims;
 use crate::database::models::{UpsertUserAiPrefs, UserAiPrefs};
+use crate::memories::MemoriesSpan;
 use crate::state::AppState;
 
 // ---------------------------------------------------------------------------
@@ -35,6 +40,18 @@ pub struct PrefsResponse {
     pub voice: Option<String>,
     pub library: Option<String>,
     pub display_tz: Option<String>,
+    /// Last/explicit insight + chat model choice. NULL = per-request /
+    /// env default behavior.
+    pub default_model: Option<String>,
+    /// `"local"` | `"hybrid"`. NULL = per-request / hard default.
+    pub default_backend: Option<String>,
+    /// Active persona for persona chat. NULL = hard default ("default").
+    pub default_persona_id: Option<String>,
+    /// Reel span: `"day"` | `"week"` | `"month"`. NULL = per-request /
+    /// hard default (day).
+    pub reel_span: Option<String>,
+    /// Reel segment cap. NULL = per-request / hard default.
+    pub reel_max_segments: Option<i32>,
     pub updated_at: Option<i64>,
 }
 
@@ -43,41 +60,59 @@ pub struct PrefsResponse {
 /// explicit null and a missing key apart on `Option<Option<T>>` (both come
 /// out as outer `None`), so this carries a hand-rolled visitor.
 #[derive(Clone, Debug, Default)]
-pub enum TriState {
+pub enum TriState<T> {
     #[default]
     Absent,
     Null,
-    Val(String),
+    Val(T),
 }
 
-impl<'de> Deserialize<'de> for TriState {
+impl<'de, T: Deserialize<'de>> Deserialize<'de> for TriState<T> {
-    fn deserialize<D>(deserializer: D) -> Result<TriState, D::Error>
+    fn deserialize<D>(deserializer: D) -> Result<TriState<T>, D::Error>
     where
         D: serde::Deserializer<'de>,
     {
-        struct Visitor;
+        struct Visitor<T>(std::marker::PhantomData<T>);
-        impl<'de> serde::de::Visitor<'de> for Visitor {
+        impl<'de, T: Deserialize<'de>> serde::de::Visitor<'de> for Visitor<T> {
-            type Value = TriState;
+            type Value = TriState<T>;
             fn expecting(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result {
-                f.write_str("a string or null")
+                f.write_str("a value or null")
             }
             // serde_json routes an explicit JSON null to visit_unit/visit_none;
             // a missing key never calls deserialize — `#[serde(default)]`
-            // supplies Absent instead.
+            // supplies Absent instead. Concrete values are re-deserialized into
-            fn visit_unit<E: serde::de::Error>(self) -> Result<TriState, E> {
+            // T through single-value deserializers (serde_json's
+            // deserialize_any dispatches directly to visit_str/visit_i64/…
+            // — visit_some only fires on the deserialize_option path).
+            fn visit_unit<E: serde::de::Error>(self) -> Result<TriState<T>, E> {
                 Ok(TriState::Null)
             }
-            fn visit_none<E: serde::de::Error>(self) -> Result<TriState, E> {
+            fn visit_none<E: serde::de::Error>(self) -> Result<TriState<T>, E> {
                 Ok(TriState::Null)
             }
-            fn visit_str<E: serde::de::Error>(self, v: &str) -> Result<TriState, E> {
+            fn visit_some<D2>(self, d: D2) -> Result<TriState<T>, D2::Error>
-                Ok(TriState::Val(v.to_string()))
+            where
+                D2: serde::Deserializer<'de>,
+            {
+                Ok(TriState::Val(T::deserialize(d)?))
             }
-            fn visit_string<E: serde::de::Error>(self, v: String) -> Result<TriState, E> {
+            // `into_deserializer()` gives a single-value deserializer whose
-                Ok(TriState::Val(v))
+            // error type is the visitor's E, so T::deserialize lands on
+            // Result<T, E> with no conversion.
+            fn visit_str<E: serde::de::Error>(self, v: &str) -> Result<TriState<T>, E> {
+                T::deserialize(v.into_deserializer()).map(TriState::Val)
+            }
+            fn visit_string<E: serde::de::Error>(self, v: String) -> Result<TriState<T>, E> {
+                T::deserialize(v.as_str().into_deserializer()).map(TriState::Val)
+            }
+            fn visit_i64<E: serde::de::Error>(self, v: i64) -> Result<TriState<T>, E> {
+                T::deserialize(v.into_deserializer()).map(TriState::Val)
+            }
+            fn visit_u64<E: serde::de::Error>(self, v: u64) -> Result<TriState<T>, E> {
+                T::deserialize(v.into_deserializer()).map(TriState::Val)
             }
         }
-        deserializer.deserialize_any(Visitor)
+        deserializer.deserialize_any(Visitor(std::marker::PhantomData))
     }
 }
 
@@ -86,11 +121,21 @@ impl<'de> Deserialize<'de> for TriState {
 #[derive(Deserialize, Debug, Default)]
 pub struct UpdatePrefsRequest {
     #[serde(default)]
-    pub voice: TriState,
+    pub voice: TriState<String>,
     #[serde(default)]
-    pub library: TriState,
+    pub library: TriState<String>,
     #[serde(default)]
-    pub display_tz: TriState,
+    pub display_tz: TriState<String>,
+    #[serde(default)]
+    pub default_model: TriState<String>,
+    #[serde(default)]
+    pub default_backend: TriState<String>,
+    #[serde(default)]
+    pub default_persona_id: TriState<String>,
+    #[serde(default)]
+    pub reel_span: TriState<String>,
+    #[serde(default)]
+    pub reel_max_segments: TriState<i32>,
 }
 
 #[derive(Serialize, Debug)]
@@ -109,6 +154,18 @@ fn is_valid_display_tz(name: &str) -> bool {
     Tz::from_str(name.trim()).is_ok()
 }
 
+/// A stored `default_backend` must be consumable downstream, i.e. accepted
+/// by the canonical `BackendKind::parse` ("local"/"hybrid", case-insensitive).
+/// Delegating to that parser keeps the write-time validator and the read-time
+/// consumer from drifting apart; anything else stored here would 500 on every
+/// insight/chat request instead of failing once at write time. The blank
+/// string the parser tolerates (empty request field = app default) is
+/// rejected here — a stored default is an explicit choice, and NULL already
+/// means "no default".
+fn is_valid_backend(name: &str) -> bool {
+    !name.trim().is_empty() && crate::ai::backend::BackendKind::parse(name).is_ok()
+}
+
 /// The full set of zone names `PUT /prefs` will accept — `chrono_tz::TZ_VARIANTS`
 /// is exactly the domain of `Tz::from_str`, so the picker list and the
 /// validation check can never drift apart.
@@ -119,14 +176,13 @@ pub fn all_timezones() -> Vec<String> {
         .collect()
 }
 
-/// Merge a partial update onto the existing row. Returns
+/// Merge a partial update onto the existing row. An omitted field keeps
-/// `(voice, library, display_tz, changed)`. An omitted field keeps the
+/// the stored value (or `None` when there is no row); a present field wins.
-/// stored value (or `None` when there is no row); a present field wins.
+fn tri_value<T: Clone>(
-fn tri_value(
+    field: &TriState<T>,
-    field: &TriState,
     existing: Option<&UserAiPrefs>,
-    keep: impl FnOnce(&UserAiPrefs) -> Option<String>,
+    keep: impl FnOnce(&UserAiPrefs) -> Option<T>,
-) -> Option<String> {
+) -> Option<T> {
     match field {
         TriState::Absent => existing.and_then(keep),
         TriState::Null => None,
@@ -134,17 +190,43 @@ fn tri_value(
     }
 }
 
-fn merge_patch(
+/// The eight partial-update fields after a merge (everything except the
-    existing: Option<&UserAiPrefs>,
+/// frozen `tz_offset_minutes`, which the handler copies through verbatim).
-    patch: &UpdatePrefsRequest,
+struct MergedPrefs {
-) -> (Option<String>, Option<String>, Option<String>, bool) {
+    voice: Option<String>,
-    let voice = tri_value(&patch.voice, existing, |e| e.voice.clone());
+    library: Option<String>,
-    let library = tri_value(&patch.library, existing, |e| e.library.clone());
+    display_tz: Option<String>,
-    let display_tz = tri_value(&patch.display_tz, existing, |e| e.display_tz.clone());
+    default_model: Option<String>,
+    default_backend: Option<String>,
+    default_persona_id: Option<String>,
+    reel_span: Option<String>,
+    reel_max_segments: Option<i32>,
+}
+
+fn merge_patch(existing: Option<&UserAiPrefs>, patch: &UpdatePrefsRequest) -> (MergedPrefs, bool) {
+    let merged = MergedPrefs {
+        voice: tri_value(&patch.voice, existing, |e| e.voice.clone()),
+        library: tri_value(&patch.library, existing, |e| e.library.clone()),
+        display_tz: tri_value(&patch.display_tz, existing, |e| e.display_tz.clone()),
+        default_model: tri_value(&patch.default_model, existing, |e| e.default_model.clone()),
+        default_backend: tri_value(&patch.default_backend, existing, |e| {
+            e.default_backend.clone()
+        }),
+        default_persona_id: tri_value(&patch.default_persona_id, existing, |e| {
+            e.default_persona_id.clone()
+        }),
+        reel_span: tri_value(&patch.reel_span, existing, |e| e.reel_span.clone()),
+        reel_max_segments: tri_value(&patch.reel_max_segments, existing, |e| e.reel_max_segments),
+    };
     let changed = !matches!(patch.voice, TriState::Absent)
         || !matches!(patch.library, TriState::Absent)
-        || !matches!(patch.display_tz, TriState::Absent);
+        || !matches!(patch.display_tz, TriState::Absent)
-    (voice, library, display_tz, changed)
+        || !matches!(patch.default_model, TriState::Absent)
+        || !matches!(patch.default_backend, TriState::Absent)
+        || !matches!(patch.default_persona_id, TriState::Absent)
+        || !matches!(patch.reel_span, TriState::Absent)
+        || !matches!(patch.reel_max_segments, TriState::Absent);
+    (merged, changed)
 }
 
 fn row_to_response(row: Option<&UserAiPrefs>) -> PrefsResponse {
@@ -153,12 +235,22 @@ fn row_to_response(row: Option<&UserAiPrefs>) -> PrefsResponse {
             voice: row.voice.clone(),
             library: row.library.clone(),
             display_tz: row.display_tz.clone(),
+            default_model: row.default_model.clone(),
+            default_backend: row.default_backend.clone(),
+            default_persona_id: row.default_persona_id.clone(),
+            reel_span: row.reel_span.clone(),
+            reel_max_segments: row.reel_max_segments,
             updated_at: Some(row.updated_at),
         },
         None => PrefsResponse {
             voice: None,
             library: None,
             display_tz: None,
+            default_model: None,
+            default_backend: None,
+            default_persona_id: None,
+            reel_span: None,
+            reel_max_segments: None,
             updated_at: None,
         },
     }
@@ -192,7 +284,8 @@ pub async fn put_prefs(
 ) -> HttpResponse {
     let patch = body.0;
 
-    // Validate before any write: an invalid zone name → 400, nothing written.
+    // Validate before any write: an invalid value → 400, nothing written
+    // (not even the other fields in the same body).
     if let TriState::Val(name) = &patch.display_tz
         && !is_valid_display_tz(name)
     {
@@ -200,6 +293,20 @@ pub async fn put_prefs(
             "error": format!("invalid display_tz: {name}")
         }));
     }
+    if let TriState::Val(name) = &patch.default_backend
+        && !is_valid_backend(name)
+    {
+        return HttpResponse::BadRequest().json(json!({
+            "error": format!("invalid default_backend: {name} (expected local|hybrid)")
+        }));
+    }
+    if let TriState::Val(name) = &patch.reel_span
+        && MemoriesSpan::from_str(name).is_err()
+    {
+        return HttpResponse::BadRequest().json(json!({
+            "error": format!("invalid reel_span: {name} (expected day|week|month)")
+        }));
+    }
 
     let mut dao = match app_state.user_ai_prefs_dao.lock() {
         Ok(dao) => dao,
@@ -219,7 +326,7 @@ pub async fn put_prefs(
         }
     };
 
-    let (voice, library, display_tz, changed) = merge_patch(existing.as_ref(), &patch);
+    let (merged, changed) = merge_patch(existing.as_ref(), &patch);
     if !changed {
         // Empty body: no-op, return the current row without bumping updated_at.
         return HttpResponse::Ok().json(row_to_response(existing.as_ref()));
@@ -231,18 +338,28 @@ pub async fn put_prefs(
         .as_secs() as i64;
 
     let resp = PrefsResponse {
-        voice: voice.clone(),
+        voice: merged.voice.clone(),
-        library: library.clone(),
+        library: merged.library.clone(),
-        display_tz: display_tz.clone(),
+        display_tz: merged.display_tz.clone(),
+        default_model: merged.default_model.clone(),
+        default_backend: merged.default_backend.clone(),
+        default_persona_id: merged.default_persona_id.clone(),
+        reel_span: merged.reel_span.clone(),
+        reel_max_segments: merged.reel_max_segments,
         updated_at: Some(now),
     };
     let upsert = UpsertUserAiPrefs {
-        voice,
+        voice: merged.voice,
         // Frozen legacy column: preserved verbatim, never written by this
         // surface.
         tz_offset_minutes: existing.as_ref().and_then(|e| e.tz_offset_minutes),
-        library,
+        library: merged.library,
-        display_tz,
+        display_tz: merged.display_tz,
+        default_model: merged.default_model,
+        default_backend: merged.default_backend,
+        default_persona_id: merged.default_persona_id,
+        reel_span: merged.reel_span,
+        reel_max_segments: merged.reel_max_segments,
         updated_at: now,
     };
     match dao.upsert_prefs(&ctx, &upsert) {
@@ -311,6 +428,27 @@ mod tests {
             library: library.map(str::to_string),
             updated_at: 1_700_000_000,
             display_tz: display_tz.map(str::to_string),
+            default_model: None,
+            default_backend: None,
+            default_persona_id: None,
+            reel_span: None,
+            reel_max_segments: None,
+        }
+    }
+
+    /// A row with owner defaults populated, for the owner-default tests.
+    fn seeded_row_with_defaults() -> UpsertUserAiPrefs {
+        UpsertUserAiPrefs {
+            voice: Some("grandma".to_string()),
+            tz_offset_minutes: Some(-240),
+            library: Some("1".to_string()),
+            updated_at: 1_700_000_000,
+            display_tz: Some("America/New_York".to_string()),
+            default_model: Some("llama3.2:3b".to_string()),
+            default_backend: Some("hybrid".to_string()),
+            default_persona_id: Some("cabin".to_string()),
+            reel_span: Some("week".to_string()),
+            reel_max_segments: Some(24),
         }
     }
 
@@ -363,6 +501,11 @@ mod tests {
                 "voice": null,
                 "library": null,
                 "display_tz": null,
+                "default_model": null,
+                "default_backend": null,
+                "default_persona_id": null,
+                "reel_span": null,
+                "reel_max_segments": null,
                 "updated_at": null
             })
         );
@@ -682,43 +825,242 @@ mod tests {
 
     // --- merge_patch (pure) --------------------------------------------------
 
-    #[test]
+    fn row_with_defaults() -> UserAiPrefs {
-    fn merge_patch_omitted_keeps_existing() {
+        UserAiPrefs {
-        let row = UserAiPrefs {
             id: 1,
             voice: Some("a".into()),
             tz_offset_minutes: None,
             library: Some("b".into()),
             updated_at: 0,
             display_tz: Some("c".into()),
-        };
+            default_model: Some("m1".into()),
-        let (voice, library, display_tz, changed) =
+            default_backend: Some("local".into()),
-            merge_patch(Some(&row), &UpdatePrefsRequest::default());
+            default_persona_id: Some("p1".into()),
-        assert_eq!(voice.as_deref(), Some("a"));
+            reel_span: Some("week".into()),
-        assert_eq!(library.as_deref(), Some("b"));
+            reel_max_segments: Some(24),
-        assert_eq!(display_tz.as_deref(), Some("c"));
+        }
+    }
+
+    #[test]
+    fn merge_patch_omitted_keeps_existing() {
+        let row = row_with_defaults();
+        let (merged, changed) = merge_patch(Some(&row), &UpdatePrefsRequest::default());
+        assert_eq!(merged.voice.as_deref(), Some("a"));
+        assert_eq!(merged.library.as_deref(), Some("b"));
+        assert_eq!(merged.display_tz.as_deref(), Some("c"));
+        assert_eq!(merged.default_model.as_deref(), Some("m1"));
+        assert_eq!(merged.default_backend.as_deref(), Some("local"));
+        assert_eq!(merged.default_persona_id.as_deref(), Some("p1"));
+        assert_eq!(merged.reel_span.as_deref(), Some("week"));
+        assert_eq!(merged.reel_max_segments, Some(24));
         assert!(!changed);
     }
 
     #[test]
     fn merge_patch_present_field_wins_including_null() {
-        let row = UserAiPrefs {
+        let row = row_with_defaults();
-            id: 1,
-            voice: Some("a".into()),
-            tz_offset_minutes: None,
-            library: None,
-            updated_at: 0,
-            display_tz: None,
-        };
         let patch = UpdatePrefsRequest {
             voice: TriState::Null,                     // clear
             library: TriState::Val(String::from("x")), // set
             display_tz: TriState::Absent,
+            default_backend: TriState::Null, // clear an owner default
+            reel_span: TriState::Val(String::from("month")),
+            reel_max_segments: TriState::Val(12),
+            ..UpdatePrefsRequest::default()
         };
-        let (voice, library, display_tz, changed) = merge_patch(Some(&row), &patch);
+        let (merged, changed) = merge_patch(Some(&row), &patch);
-        assert!(voice.is_none());
+        assert!(merged.voice.is_none());
-        assert_eq!(library.as_deref(), Some("x"));
+        assert_eq!(merged.library.as_deref(), Some("x"));
-        assert!(display_tz.is_none());
+        // Absent field keeps the stored value.
+        assert_eq!(merged.display_tz.as_deref(), Some("c"));
+        assert!(merged.default_backend.is_none());
+        assert_eq!(merged.reel_span.as_deref(), Some("month"));
+        assert_eq!(merged.reel_max_segments, Some(12));
+        // Untouched owner defaults survive.
+        assert_eq!(merged.default_model.as_deref(), Some("m1"));
+        assert_eq!(merged.default_persona_id.as_deref(), Some("p1"));
         assert!(changed);
     }
+
+    // --- PUT /prefs: owner defaults ------------------------------------------
+
+    #[actix_rt::test]
+    async fn put_prefs_owner_defaults_set_and_persist() {
+        let (state, mut seed_dao) = prefs_state();
+        seed(
+            &mut seed_dao,
+            seeded_row(Some("grandma"), Some("1"), Some("America/New_York")),
+        );
+        let app = init_service(
+            App::new()
+                .service(
+                    web::resource("/prefs")
+                        .route(web::get().to(get_prefs))
+                        .route(web::put().to(put_prefs)),
+                )
+                .service(web::resource("/timezones").route(web::get().to(list_timezones)))
+                .app_data(state.clone()),
+        )
+        .await;
+        let (header, value) = auth();
+        let req = TestRequest::put()
+            .uri("/prefs")
+            .insert_header((header, value))
+            .set_json(serde_json::json!({
+                "default_model": "llama3.2:3b",
+                "default_backend": "hybrid",
+                "default_persona_id": "cabin",
+                "reel_span": "week",
+                "reel_max_segments": 24
+            }))
+            .to_request();
+        let resp = call_service(&app, req).await;
+        assert_eq!(resp.status(), 200);
+
+        let mut dao = state.user_ai_prefs_dao.lock().unwrap();
+        let row = dao
+            .get_prefs(&opentelemetry::Context::new())
+            .unwrap()
+            .unwrap();
+        assert_eq!(row.default_model.as_deref(), Some("llama3.2:3b"));
+        assert_eq!(row.default_backend.as_deref(), Some("hybrid"));
+        assert_eq!(row.default_persona_id.as_deref(), Some("cabin"));
+        assert_eq!(row.reel_span.as_deref(), Some("week"));
+        assert_eq!(row.reel_max_segments, Some(24));
+        // Mirror fields untouched by the owner-default patch.
+        assert_eq!(row.voice.as_deref(), Some("grandma"));
+        assert_eq!(row.library.as_deref(), Some("1"));
+    }
+
+    #[actix_rt::test]
+    async fn put_prefs_owner_defaults_null_clears() {
+        let (state, mut seed_dao) = prefs_state();
+        seed(&mut seed_dao, seeded_row_with_defaults());
+        let app = init_service(
+            App::new()
+                .service(
+                    web::resource("/prefs")
+                        .route(web::get().to(get_prefs))
+                        .route(web::put().to(put_prefs)),
+                )
+                .service(web::resource("/timezones").route(web::get().to(list_timezones)))
+                .app_data(state.clone()),
+        )
+        .await;
+        let (header, value) = auth();
+        let req = TestRequest::put()
+            .uri("/prefs")
+            .insert_header((header, value))
+            .set_json(serde_json::json!({
+                "default_model": null,
+                "default_backend": null,
+                "default_persona_id": null,
+                "reel_span": null,
+                "reel_max_segments": null
+            }))
+            .to_request();
+        let resp = call_service(&app, req).await;
+        assert_eq!(resp.status(), 200);
+
+        let mut dao = state.user_ai_prefs_dao.lock().unwrap();
+        let row = dao
+            .get_prefs(&opentelemetry::Context::new())
+            .unwrap()
+            .unwrap();
+        assert!(row.default_model.is_none());
+        assert!(row.default_backend.is_none());
+        assert!(row.default_persona_id.is_none());
+        assert!(row.reel_span.is_none());
+        assert!(row.reel_max_segments.is_none());
+        // Sibling fields survive a clear patch.
+        assert_eq!(row.voice.as_deref(), Some("grandma"));
+    }
+
+    #[actix_rt::test]
+    async fn put_prefs_invalid_backend_returns_400_and_writes_nothing() {
+        let (state, mut seed_dao) = prefs_state();
+        seed(
+            &mut seed_dao,
+            seeded_row(Some("grandma"), Some("1"), Some("America/New_York")),
+        );
+        let app = init_service(
+            App::new()
+                .service(
+                    web::resource("/prefs")
+                        .route(web::get().to(get_prefs))
+                        .route(web::put().to(put_prefs)),
+                )
+                .service(web::resource("/timezones").route(web::get().to(list_timezones)))
+                .app_data(state.clone()),
+        )
+        .await;
+        let (header, value) = auth();
+        let req = TestRequest::put()
+            .uri("/prefs")
+            .insert_header((header, value))
+            .set_json(serde_json::json!({"default_backend": "quantum", "voice": "dad"}))
+            .to_request();
+        let resp = call_service(&app, req).await;
+        assert_eq!(resp.status(), 400);
+
+        let mut dao = state.user_ai_prefs_dao.lock().unwrap();
+        let row = dao
+            .get_prefs(&opentelemetry::Context::new())
+            .unwrap()
+            .unwrap();
+        // Nothing written — not even the valid field in the same body.
+        assert_eq!(row.voice.as_deref(), Some("grandma"));
+        assert!(row.default_backend.is_none());
+        assert_eq!(row.updated_at, 1_700_000_000);
+    }
+
+    #[test]
+    fn backend_validator_delegates_to_canonical_parser() {
+        // Case-insensitive like BackendKind::parse (a stored pref must always
+        // be consumable downstream), but the blank string the parser tolerates
+        // for request fields is rejected — NULL already means "no default".
+        assert!(is_valid_backend("local"));
+        assert!(is_valid_backend("hybrid"));
+        assert!(is_valid_backend("Hybrid"));
+        assert!(is_valid_backend("  LOCAL "));
+        assert!(!is_valid_backend("quantum"));
+        assert!(!is_valid_backend(""));
+        assert!(!is_valid_backend("   "));
+    }
+
+    #[actix_rt::test]
+    async fn put_prefs_invalid_reel_span_returns_400_and_writes_nothing() {
+        let (state, mut seed_dao) = prefs_state();
+        seed(
+            &mut seed_dao,
+            seeded_row(Some("grandma"), Some("1"), Some("America/New_York")),
+        );
+        let app = init_service(
+            App::new()
+                .service(
+                    web::resource("/prefs")
+                        .route(web::get().to(get_prefs))
+                        .route(web::put().to(put_prefs)),
+                )
+                .service(web::resource("/timezones").route(web::get().to(list_timezones)))
+                .app_data(state.clone()),
+        )
+        .await;
+        let (header, value) = auth();
+        let req = TestRequest::put()
+            .uri("/prefs")
+            .insert_header((header, value))
+            .set_json(serde_json::json!({"reel_span": "decade"}))
+            .to_request();
+        let resp = call_service(&app, req).await;
+        assert_eq!(resp.status(), 400);
+
+        let mut dao = state.user_ai_prefs_dao.lock().unwrap();
+        let row = dao
+            .get_prefs(&opentelemetry::Context::new())
+            .unwrap()
+            .unwrap();
+        assert!(row.reel_span.is_none());
+        assert_eq!(row.updated_at, 1_700_000_000);
+    }
 }

src/reels/mod.rs

@@ -18,6 +18,7 @@ pub mod selector;
 
 use std::collections::HashMap;
 use std::path::{Path, PathBuf};
+use std::str::FromStr;
 use std::sync::{LazyLock, Mutex, Mutex as StdMutex};
 use std::time::{Duration, Instant};
 
@@ -71,13 +72,38 @@ pub fn normalize_library_key(libs: &[Library], param: Option<&str>) -> String {
     }
 }
 
+/// Reel span / segment-cap resolution: explicit request param → stored
+/// owner pref → hard default. A stored span that no longer parses (e.g.
+/// hand-edited) is skipped rather than erroring — the hard default still
+/// applies. Non-positive stored segment caps are ignored the same way.
+fn resolve_reel_params(
+    req_span: Option<MemoriesSpan>,
+    req_max_segments: Option<usize>,
+    prefs: Option<&crate::database::models::UserAiPrefs>,
+) -> (MemoriesSpan, usize) {
+    let pref_span = prefs
+        .and_then(|p| p.reel_span.as_deref())
+        .and_then(|s| MemoriesSpan::from_str(s).ok());
+    let pref_max_segments = prefs
+        .and_then(|p| p.reel_max_segments)
+        .filter(|v| *v > 0)
+        .map(|v| v as usize);
+    let span = req_span.or(pref_span).unwrap_or(MemoriesSpan::Day);
+    let max_segments = req_max_segments
+        .or(pref_max_segments)
+        .unwrap_or(selector::DEFAULT_MAX_SEGMENTS);
+    (span, max_segments)
+}
+
 /// Best-effort: mirror the latest client reel params into `user_ai_prefs`
 /// so the nightly pre-gen scheduler can pick them up. Never fails the
 /// caller regardless of DB errors.
 fn capture_prefs(
     app_state: &AppState,
-    req: &web::Json<CreateReelRequest>,
+    req: &CreateReelRequest,
     library_param: Option<&str>,
+    effective_span: MemoriesSpan,
+    effective_max_segments: usize,
 ) -> Result<(), anyhow::Error> {
     use crate::database::models::UpsertUserAiPrefs;
     let now = std::time::SystemTime::now()
@@ -93,6 +119,21 @@ fn capture_prefs(
     };
     let mut dao = app_state.user_ai_prefs_dao.lock().expect("lock");
     let ctx = opentelemetry::Context::new();
+    // The row passively mirrors the latest request's EFFECTIVE reel params
+    // (after the request → pref → default resolution done by the caller), so
+    // the stored reel_span / reel_max_segments track what was last used —
+    // idempotent when the pref itself supplied the effective value.
+    //
+    // The owner-default fields (model/backend/persona) are owned by PUT /prefs,
+    // not by reel requests. The upsert overwrites every column, so we must
+    // carry the existing values through unchanged — writing None here would
+    // silently wipe the owner's stored defaults on every reel creation. If the
+    // read fails we bail without writing: this is best-effort, and clobbering
+    // owner defaults on a transient error is worse than skipping a mirror.
+    let existing = match dao.get_prefs(&ctx) {
+        Ok(row) => row,
+        Err(e) => return Err(anyhow::anyhow!("failed to read prefs before mirror: {e}")),
+    };
     dao.upsert_prefs(
         &ctx,
         &UpsertUserAiPrefs {
@@ -104,6 +145,11 @@ fn capture_prefs(
             library: Some(library),
             updated_at: now as i64,
             display_tz: req.display_tz.clone().filter(|s| !s.is_empty()),
+            default_model: existing.as_ref().and_then(|e| e.default_model.clone()),
+            default_backend: existing.as_ref().and_then(|e| e.default_backend.clone()),
+            default_persona_id: existing.as_ref().and_then(|e| e.default_persona_id.clone()),
+            reel_span: Some(effective_span.to_string()),
+            reel_max_segments: Some(effective_max_segments as i32),
         },
     )
     .map_err(|e| anyhow::anyhow!("failed to upsert user_ai_prefs: {e}"))
@@ -415,20 +461,21 @@ pub async fn create_reel_handler(
         }));
     }
 
-    let span = req.span.unwrap_or(MemoriesSpan::Day);
+    // Owner defaults: explicit request param → stored pref → hard default.
-    let max_segments = req.max_segments.unwrap_or(selector::DEFAULT_MAX_SEGMENTS);
+    // The stored row is read once and reused for the span, the segment cap,
-    // Grouping zone: explicit request param → stored pref → app default
+    // and the display zone. The legacy device-offset param is still mirrored
-    // (ADR-0001). The legacy device-offset param is still mirrored into the
+    // into the frozen prefs column by `capture_prefs` for rollback but no
-    // frozen prefs column by `capture_prefs` for rollback but no longer
+    // longer drives selection.
-    // drives selection.
+    let stored_prefs = app_state
+        .user_ai_prefs_dao
+        .lock()
+        .ok()
+        .and_then(|mut dao| dao.get_prefs(&opentelemetry::Context::new()).ok().flatten());
+    let (span, max_segments) =
+        resolve_reel_params(req.span, req.max_segments, stored_prefs.as_ref());
     let display_zone_name = {
         let req_tz = req.display_tz.clone().filter(|s| !s.trim().is_empty());
-        let stored = app_state
+        let stored = stored_prefs.as_ref().and_then(|p| p.display_tz.clone());
-            .user_ai_prefs_dao
-            .lock()
-            .ok()
-            .and_then(|mut dao| dao.get_prefs(&opentelemetry::Context::new()).ok().flatten())
-            .and_then(|p| p.display_tz);
         crate::camera_tz::display_zone_for(req_tz.as_deref().or(stored.as_deref()))
     };
     let selector = ReelSelector::Memories {
@@ -489,7 +536,7 @@ pub async fn create_reel_handler(
             },
         );
         // Capture params for passive prefs mirror (best-effort, never fails).
-        let _ = capture_prefs(&app_state, &req, req.library.as_deref());
+        let _ = capture_prefs(&app_state, &req, req.library.as_deref(), span, max_segments);
         return HttpResponse::Accepted().json(ReelJobCreatedResponse {
             job_id: job_id.to_string(),
             status: ReelJobStatus::Done,
@@ -549,7 +596,7 @@ pub async fn create_reel_handler(
     with_job(job_id, |job| job.abort = Some(handle.abort_handle()));
 
     // Capture params for passive prefs mirror (best-effort, never fails).
-    let _ = capture_prefs(&app_state, &req, req.library.as_deref());
+    let _ = capture_prefs(&app_state, &req, req.library.as_deref(), span, max_segments);
 
     HttpResponse::Accepted().json(ReelJobCreatedResponse {
         job_id: job_id.to_string(),
@@ -1333,9 +1380,121 @@ fn sweep_reel_cache(app_state: &AppState) -> usize {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::database::models::UserAiPrefs;
     use crate::libraries::Library;
     use chrono::TimeZone;
 
+    fn prefs_row(span: Option<&str>, max_segments: Option<i32>) -> UserAiPrefs {
+        UserAiPrefs {
+            id: 1,
+            voice: None,
+            tz_offset_minutes: None,
+            library: None,
+            updated_at: 0,
+            display_tz: None,
+            default_model: None,
+            default_backend: None,
+            default_persona_id: None,
+            reel_span: span.map(str::to_string),
+            reel_max_segments: max_segments,
+        }
+    }
+
+    #[test]
+    fn resolve_reel_params_request_wins_over_pref() {
+        let prefs = prefs_row(Some("month"), Some(10));
+        let (span, max) = resolve_reel_params(Some(MemoriesSpan::Day), Some(5), Some(&prefs));
+        assert_eq!(span, MemoriesSpan::Day);
+        assert_eq!(max, 5);
+    }
+
+    #[test]
+    fn resolve_reel_params_pref_fills_omitted_request() {
+        let prefs = prefs_row(Some("week"), Some(24));
+        let (span, max) = resolve_reel_params(None, None, Some(&prefs));
+        assert_eq!(span, MemoriesSpan::Week);
+        assert_eq!(max, 24);
+    }
+
+    #[test]
+    fn resolve_reel_params_no_pref_falls_to_hard_defaults() {
+        let (span, max) = resolve_reel_params(None, None, None);
+        assert_eq!(span, MemoriesSpan::Day);
+        assert_eq!(max, selector::DEFAULT_MAX_SEGMENTS);
+    }
+
+    #[test]
+    fn resolve_reel_params_invalid_or_nonpositive_pref_skipped() {
+        // Hand-edited / corrupt stored values fall through to hard defaults
+        // instead of erroring the request.
+        let bad_span = prefs_row(Some("decade"), Some(0));
+        let (span, max) = resolve_reel_params(None, None, Some(&bad_span));
+        assert_eq!(span, MemoriesSpan::Day);
+        assert_eq!(max, selector::DEFAULT_MAX_SEGMENTS);
+    }
+
+    /// Regression: reel creation is a PASSIVE mirror. The upsert overwrites
+    /// every column, so `capture_prefs` must carry the PUT /prefs-owned
+    /// default_model / default_backend / default_persona_id through unchanged
+    /// — a reel request has no business wiping the owner's AI defaults.
+    #[actix_rt::test]
+    async fn reel_capture_preserves_owner_defaults() {
+        use crate::database::models::UpsertUserAiPrefs;
+        use crate::database::{SqliteUserAiPrefsDao, UserAiPrefsDao};
+        use crate::state::AppState;
+        use std::sync::{Arc, Mutex};
+
+        let conn = Arc::new(Mutex::new(crate::database::test::in_memory_db_connection()));
+        let mut state = AppState::test_state();
+        state.user_ai_prefs_dao = Arc::new(Mutex::new(Box::new(
+            SqliteUserAiPrefsDao::from_connection(conn.clone()),
+        )));
+        {
+            let mut dao = SqliteUserAiPrefsDao::from_connection(conn.clone());
+            dao.upsert_prefs(
+                &opentelemetry::Context::new(),
+                &UpsertUserAiPrefs {
+                    voice: None,
+                    tz_offset_minutes: None,
+                    library: None,
+                    updated_at: 0,
+                    display_tz: None,
+                    default_model: Some("llama3.2:3b".into()),
+                    default_backend: Some("hybrid".into()),
+                    default_persona_id: Some("cabin".into()),
+                    reel_span: None,
+                    reel_max_segments: None,
+                },
+            )
+            .expect("seed owner defaults");
+        }
+
+        let req = CreateReelRequest {
+            span: Some(MemoriesSpan::Week),
+            timezone_offset_minutes: Some(-240),
+            display_tz: None,
+            library: None,
+            voice: Some("grandma".into()),
+            max_segments: Some(12),
+        };
+        capture_prefs(&state, &req, None, MemoriesSpan::Week, 12).expect("capture");
+
+        let row = state
+            .user_ai_prefs_dao
+            .lock()
+            .expect("lock")
+            .get_prefs(&opentelemetry::Context::new())
+            .expect("read")
+            .expect("row");
+        // The mirrored reel params landed...
+        assert_eq!(row.reel_span.as_deref(), Some("week"));
+        assert_eq!(row.reel_max_segments, Some(12));
+        // ...and the owner defaults survived untouched.
+        assert_eq!(row.default_model.as_deref(), Some("llama3.2:3b"));
+        assert_eq!(row.default_backend.as_deref(), Some("hybrid"));
+        assert_eq!(row.default_persona_id.as_deref(), Some("cabin"));
+    }
+
     fn photo(p: &str, lib: i32) -> SegmentMedia {
         SegmentMedia::Photo {
             rel_path: p.to_string(),

src/reels/script.rs

@@ -281,8 +281,9 @@ pub async fn generate_script(
     beats: &[PlannedBeat],
 ) -> Result<ReelScript> {
     let (system, user) = build_script_messages(meta, beats);
+    let cx = opentelemetry::Context::current();
     let raw = client
-        .generate(&user, Some(&system), None)
+        .generate(&cx, &user, Some(&system), None)
         .await
         .context("LLM script generation failed")?;
     Ok(parse_script_response(&raw, beats.len()))

src/unified_search.rs

@@ -211,7 +211,7 @@ pub async fn unified_search<TagD: TagDao>(
     log::info!("unified_search: translating with model={}", backend.model());
 
     let today = chrono::Utc::now().date_naive();
-    let sq = match translate_nl_query(backend.chat(), &nl, &tag_vocab, today).await {
+    let sq = match translate_nl_query(backend.chat(), &nl, &tag_vocab, today, &ctx).await {
         Ok(sq) => sq,
         Err(e) => {
             log::warn!("unified_search: translate_nl_query failed: {e:?}");