feat(auth): replace cookie/api-key auth with JWT Bearer tokens, separate UI from API

- Add JWT Bearer token validation to Rust API via OIDC provider JWKS
  with automatic key rotation and 1-hour cache
- Remove x-api-key auth support and built-in web UI from furumi-web-player,
  leaving it as a pure API server
- Add /auth/token endpoint to Node player server to expose OIDC access
  tokens to the frontend
- Move Node player auth endpoints from /api/* to /auth/* to avoid
  path conflicts with Rust API
- Add static file serving to Node Express server for production
  single-container deployment
- Fix SameSite=Strict cookie issue breaking OIDC redirect flow (use Lax)
- Add Dockerfile.node-player with multi-stage Node.js build
- Add CI workflows for node-player Docker image (dev + release)
- Optimize Rust Dockerfiles with dependency caching layer
- Update docker-compose with OIDC env vars and OLLAMA_MODEL support
- Cherry-pick agent LLM client fixes from DEV branch

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

furumi-agent/src/ingest/normalize.rs

@@ -25,16 +25,37 @@ pub async fn normalize(
 ) -> anyhow::Result<NormalizedFields> {
     let user_message = build_user_message(raw, hints, similar_artists, similar_albums, folder_ctx);
 
+    let schema = normalize_schema();
     let response = call_ollama(
         &state.config.ollama_url,
         &state.config.ollama_model,
         &state.system_prompt,
         &user_message,
         state.config.ollama_auth.as_deref(),
+        0.5,
+        512,
+        Some(("normalized_metadata", schema.clone())),
     )
     .await?;
 
-    parse_response(&response)
+    match parse_response(&response) {
+        Ok(fields) => Ok(fields),
+        Err(e) => {
+            tracing::warn!(error = %e, "LLM parse failed, retrying with higher frequency_penalty");
+            let response2 = call_ollama(
+                &state.config.ollama_url,
+                &state.config.ollama_model,
+                &state.system_prompt,
+                &user_message,
+                state.config.ollama_auth.as_deref(),
+                1.5,
+                512,
+                Some(("normalized_metadata", schema)),
+            )
+            .await?;
+            parse_response(&response2)
+        }
+    }
 }
 
 fn build_user_message(
@@ -113,32 +134,49 @@ fn build_user_message(
 }
 
 #[derive(Serialize)]
-struct OllamaRequest {
+struct ChatRequest {
     model: String,
-    messages: Vec<OllamaMessage>,
-    format: String,
+    messages: Vec<ChatMessage>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    response_format: Option<ChatResponseFormat>,
     stream: bool,
-    options: OllamaOptions,
+    temperature: f64,
+    max_tokens: u32,
+    frequency_penalty: f64,
 }
 
 #[derive(Serialize)]
-struct OllamaMessage {
+struct ChatMessage {
     role: String,
     content: String,
 }
 
 #[derive(Serialize)]
-struct OllamaOptions {
-    temperature: f64,
+struct ChatResponseFormat {
+    #[serde(rename = "type")]
+    kind: String,
+    json_schema: JsonSchemaWrapper,
+}
+
+#[derive(Serialize)]
+struct JsonSchemaWrapper {
+    name: String,
+    strict: bool,
+    schema: serde_json::Value,
 }
 
 #[derive(Deserialize)]
-struct OllamaResponse {
-    message: OllamaResponseMessage,
+struct ChatResponse {
+    choices: Vec<ChatChoice>,
 }
 
 #[derive(Deserialize)]
-struct OllamaResponseMessage {
+struct ChatChoice {
+    message: ChatResponseMessage,
+}
+
+#[derive(Deserialize)]
+struct ChatResponseMessage {
     content: String,
 }
 
@@ -148,30 +186,40 @@ pub async fn call_ollama(
     system_prompt: &str,
     user_message: &str,
     auth: Option<&str>,
+    frequency_penalty: f64,
+    max_tokens: u32,
+    schema: Option<(&str, serde_json::Value)>,
 ) -> anyhow::Result<String> {
     let client = reqwest::Client::builder()
         .timeout(std::time::Duration::from_secs(120))
         .build()?;
 
-    let request = OllamaRequest {
+    let response_format = schema.map(|(name, schema)| ChatResponseFormat {
+        kind: "json_schema".to_owned(),
+        json_schema: JsonSchemaWrapper { name: name.to_owned(), strict: true, schema },
+    });
+
+    let request = ChatRequest {
         model: model.to_owned(),
         messages: vec![
-            OllamaMessage {
+            ChatMessage {
                 role: "system".to_owned(),
                 content: system_prompt.to_owned(),
             },
-            OllamaMessage {
+            ChatMessage {
                 role: "user".to_owned(),
                 content: user_message.to_owned(),
             },
         ],
-        format: "json".to_owned(),
+        response_format,
         stream: false,
-        options: OllamaOptions { temperature: 0.1 },
+        temperature: 0.1,
+        max_tokens,
+        frequency_penalty,
     };
 
-    let url = format!("{}/api/chat", base_url.trim_end_matches('/'));
-    tracing::info!(%url, model, prompt_len = user_message.len(), "Calling Ollama API...");
+    let url = format!("{}/v1/chat/completions", base_url.trim_end_matches('/'));
+    tracing::info!(%url, model, prompt_len = user_message.len(), "Calling LLM API...");
 
     let start = std::time::Instant::now();
     let mut req = client.post(&url).json(&request);
@@ -184,18 +232,45 @@ pub async fn call_ollama(
     if !resp.status().is_success() {
         let status = resp.status();
         let body = resp.text().await.unwrap_or_default();
-        tracing::error!(%status, body = &body[..body.len().min(500)], "Ollama API error");
-        anyhow::bail!("Ollama returned {}: {}", status, body);
+        tracing::error!(%status, body = &body[..body.len().min(500)], "LLM API error");
+        anyhow::bail!("LLM returned {}: {}", status, body);
     }
 
-    let ollama_resp: OllamaResponse = resp.json().await?;
+    let chat_resp: ChatResponse = resp.json().await?;
+    let content = chat_resp
+        .choices
+        .into_iter()
+        .next()
+        .ok_or_else(|| anyhow::anyhow!("LLM returned empty choices"))?
+        .message
+        .content;
     tracing::info!(
         elapsed_ms = elapsed.as_millis() as u64,
-        response_len = ollama_resp.message.content.len(),
-        "Ollama response received"
+        response_len = content.len(),
+        "LLM response received"
     );
-    tracing::debug!(raw_response = %ollama_resp.message.content, "LLM raw output");
-    Ok(ollama_resp.message.content)
+    tracing::debug!(raw_response = %content, "LLM raw output");
+    Ok(content)
+}
+
+fn normalize_schema() -> serde_json::Value {
+    serde_json::json!({
+        "type": "object",
+        "properties": {
+            "artist":           { "type": ["string", "null"] },
+            "album":            { "type": ["string", "null"] },
+            "title":            { "type": ["string", "null"] },
+            "year":             { "type": ["integer", "null"] },
+            "track_number":     { "type": ["integer", "null"] },
+            "genre":            { "type": ["string", "null"] },
+            "featured_artists": { "type": "array", "items": { "type": "string" } },
+            "release_kind":     { "type": ["string", "null"] },
+            "confidence":       { "type": ["number", "null"] },
+            "notes":            { "type": ["string", "null"] }
+        },
+        "required": ["artist", "album", "title", "year", "track_number", "genre", "featured_artists", "release_kind", "confidence", "notes"],
+        "additionalProperties": false
+    })
 }
 
 /// Parse the LLM JSON response into NormalizedFields.
@@ -222,6 +297,7 @@ fn parse_response(response: &str) -> anyhow::Result<NormalizedFields> {
         genre: Option<String>,
         #[serde(default)]
         featured_artists: Vec<String>,
+        #[serde(rename = "release_kind")]
         release_type: Option<String>,
         confidence: Option<f64>,
         notes: Option<String>,

furumi-agent/src/merge.rs

@@ -35,12 +35,39 @@ pub async fn propose_merge(state: &Arc<AppState>, merge_id: Uuid) -> anyhow::Res
 
     let user_message = build_merge_message(&artists_data);
 
+    let schema = serde_json::json!({
+        "type": "object",
+        "properties": {
+            "canonical_artist_name": { "type": "string" },
+            "winner_artist_id":      { "type": "integer" },
+            "album_mappings": {
+                "type": "array",
+                "items": {
+                    "type": "object",
+                    "properties": {
+                        "source_album_id":     { "type": "integer" },
+                        "canonical_name":      { "type": "string" },
+                        "merge_into_album_id": { "type": ["integer", "null"] }
+                    },
+                    "required": ["source_album_id", "canonical_name", "merge_into_album_id"],
+                    "additionalProperties": false
+                }
+            },
+            "notes": { "type": "string" }
+        },
+        "required": ["canonical_artist_name", "winner_artist_id", "album_mappings", "notes"],
+        "additionalProperties": false
+    });
+
     let response = call_ollama(
         &state.config.ollama_url,
         &state.config.ollama_model,
         &state.merge_prompt,
         &user_message,
         state.config.ollama_auth.as_deref(),
+        0.5,
+        4096,
+        Some(("artist_merge", schema)),
     ).await?;
 
     let proposal = parse_merge_response(&response)?;