feat(agent/sync): sync non-OpenRouter models from upstream endpoint

When upstream /v1/models returns models not yet in OpenRouter's catalog
(e.g. brand-new models like DeepSeek-V4), also upsert them through the
same pipeline (provider → model → version → pricing → capabilities →
parameter_profile) with inferred defaults, instead of silently dropping
them. Previously the direct-sync fallback only triggered when *zero*
OpenRouter matches existed.

libs/service/agent/sync.rs

@@ -5,7 +5,10 @@
 //!   2. Fetch full metadata (pricing, context_length, capabilities) for those
 //!      model IDs from OpenRouter's public `/api/v1/models` endpoint (no auth).
 //!   3. Upsert provider / model / version / pricing / capability / profile
-//!      records only for models the client can actually call.
+//!      records for models the client can actually call.
+//!   4. Models accessible from the user's endpoint but NOT in OpenRouter's
+//!      catalog ("stranger" models) are also upserted through the same
+//!      pipeline with inferred defaults (direct sync).
 //!
 //! Usage: call `start_sync_task()` to launch a background task that syncs
 //! immediately and then every 10 minutes. On app startup, run it once
@@ -704,7 +707,11 @@ impl AppService {
     /// Steps:
     ///   1. Call `client.models().list()` to get the set of accessible model IDs.
     ///   2. Fetch full model list from OpenRouter's public `/api/v1/models` endpoint.
-    ///   3. Keep only models whose ID appears in the accessible set, then upsert.
+    ///   3. Keep only models whose ID appears in the accessible set, then upsert
+    ///      with full OpenRouter metadata.
+    ///   4. Models NOT in OpenRouter's catalog are also upserted through the
+    ///      same pipeline (provider → model → version → pricing → capabilities
+    ///      → parameter_profile) with inferred defaults.
     pub async fn sync_upstream_models(
         &self,
         _ctx: &Session,
@@ -733,17 +740,21 @@ impl AppService {
             .filter(|m| m.id != "openrouter/auto")
             .collect();
 
-        let filtered_count = filtered.len();
+        // Identify "stranger" models: accessible from the user's endpoint but
+        // NOT present in OpenRouter's public catalog. These are also upserted
+        // through the same pipeline (provider → model → version → pricing →
+        // capabilities → parameter_profile) with inferred defaults.
+        let or_matched_ids: std::collections::HashSet<&str> = filtered
+            .iter()
+            .map(|m| m.id.as_str())
+            .collect();
+        let unknown_ids: Vec<&str> = available_ids
+            .iter()
+            .filter(|id| !or_matched_ids.contains(id.as_str()))
+            .map(|s| s.as_str())
+            .collect();
 
-        // Fallback: if no OpenRouter metadata matches, sync models directly from
-        // the user's endpoint (handles Bailian/MiniMax and other non-OpenRouter providers).
-        if filtered_count == 0 && !available_ids.is_empty() {
-            tracing::info!(
-                model_count = available_ids.len(),
-                "sync_upstream_models: no OpenRouter matches, falling back to direct sync"
-            );
-            return Ok(sync_models_direct(&self.db, &available_ids).await);
-        }
+        let filtered_count = filtered.len();
 
         let mut models_created = 0i64;
         let mut models_updated = 0i64;
@@ -752,6 +763,37 @@ impl AppService {
         let mut capabilities_created = 0i64;
         let mut profiles_created = 0i64;
 
+        // Sync stranger models (non-OpenRouter) through the direct pipeline.
+        if !unknown_ids.is_empty() {
+            tracing::info!(
+                unknown_count = unknown_ids.len(),
+                "sync_upstream_models: {} models not in OpenRouter catalog, syncing directly",
+                unknown_ids.len()
+            );
+            let unknown_set: std::collections::HashSet<String> =
+                unknown_ids.iter().map(|s| ToString::to_string(s)).collect();
+            let direct_result = sync_models_direct(&self.db, &unknown_set).await;
+            models_created += direct_result.models_created;
+            models_updated += direct_result.models_updated;
+            versions_created += direct_result.versions_created;
+            pricing_created += direct_result.pricing_created;
+            capabilities_created += direct_result.capabilities_created;
+            profiles_created += direct_result.profiles_created;
+        }
+
+        // If no OpenRouter metadata matched at all, the direct sync above
+        // already handled everything — return early.
+        if filtered_count == 0 {
+            return Ok(SyncModelsResponse {
+                models_created,
+                models_updated,
+                versions_created,
+                pricing_created,
+                capabilities_created,
+                profiles_created,
+            });
+        }
+
         for or_model in filtered {
             let provider_slug = extract_provider(&or_model.id);
             let provider = match upsert_provider(&self.db, provider_slug).await {
@@ -823,8 +865,10 @@ impl AppService {
             filtered_count,
             models_created,
             models_updated,
-            "sync_upstream_models: synced {} accessible models",
-            filtered_count
+            "sync_upstream_models: synced {} accessible models ({}) OpenRouter + ({}) direct",
+            filtered_count + unknown_ids.len(),
+            filtered_count,
+            unknown_ids.len()
         );
 
         Ok(SyncModelsResponse {
@@ -898,19 +942,21 @@ impl AppService {
             .filter(|m| m.id != "openrouter/auto")
             .collect();
 
+        // Identify "stranger" models: accessible from the user's endpoint but
+        // NOT present in OpenRouter's public catalog.
+        let or_matched_ids: std::collections::HashSet<&str> = filtered
+            .iter()
+            .map(|m| m.id.as_str())
+            .collect();
+        let unknown_ids: Vec<&str> = available_ids
+            .iter()
+            .filter(|id| !or_matched_ids.contains(id.as_str()))
+            .map(|s| s.as_str())
+            .collect();
+
         let filtered_count = filtered.len();
 
-        // Fallback: if no OpenRouter metadata matches, sync models directly from
-        // the user's endpoint (handles Bailian/MiniMax and other non-OpenRouter providers).
-        if filtered_count == 0 && !available_ids.is_empty() {
-            tracing::info!(
-                model_count = available_ids.len(),
-                "OpenRouter model sync: no matches, falling back to direct sync"
-            );
-            sync_models_direct(db, &available_ids).await;
-            return;
-        }
-
+        // Sync stranger models (non-OpenRouter) through the direct pipeline.
         let mut models_created = 0i64;
         let mut models_updated = 0i64;
         let mut versions_created = 0i64;
@@ -918,6 +964,39 @@ impl AppService {
         let mut capabilities_created = 0i64;
         let mut profiles_created = 0i64;
 
+        if !unknown_ids.is_empty() {
+            tracing::info!(
+                unknown_count = unknown_ids.len(),
+                "OpenRouter model sync: {} models not in OpenRouter catalog, syncing directly",
+                unknown_ids.len()
+            );
+            let unknown_set: std::collections::HashSet<String> =
+                unknown_ids.iter().map(|s| ToString::to_string(s)).collect();
+            let direct_result = sync_models_direct(db, &unknown_set).await;
+            models_created += direct_result.models_created;
+            models_updated += direct_result.models_updated;
+            versions_created += direct_result.versions_created;
+            pricing_created += direct_result.pricing_created;
+            capabilities_created += direct_result.capabilities_created;
+            profiles_created += direct_result.profiles_created;
+        }
+
+        // If no OpenRouter metadata matched at all, the direct sync above
+        // already handled everything — return early.
+        if filtered_count == 0 {
+            tracing::info!(
+                matched = filtered_count,
+                models_created,
+                models_updated,
+                versions_created,
+                pricing_created,
+                capabilities_created,
+                profiles_created,
+                "OpenRouter model sync complete (direct only)"
+            );
+            return;
+        }
+
         for or_model in filtered {
             let provider_slug = extract_provider(&or_model.id);
             let provider = match upsert_provider(db, provider_slug).await {
@@ -987,13 +1066,17 @@ impl AppService {
 
         tracing::info!(
             matched = filtered_count,
+            unknown = unknown_ids.len(),
             models_created,
             models_updated,
             versions_created,
             pricing_created,
             capabilities_created,
             profiles_created,
-            "OpenRouter model sync complete"
+            "OpenRouter model sync complete: {} total ({} OpenRouter + {} direct)",
+            filtered_count + unknown_ids.len(),
+            filtered_count,
+            unknown_ids.len()
         );
     }
 }