Add contact name filter to SMS search tool + misc improvements

- sms search tool: accept contact name, trim/validate, skip when
  contact_id is set, pass to API client
- sms_client: new contact field in SmsSearchParams, URL-encode on wire
- Tool description clarifies contact_id takes precedence when both given
- Add parse_title_body helper for LLM response parsing
- llamacpp backend improvements

src/ai/insight_chat.rs

@@ -817,7 +817,8 @@ impl InsightChatService {
 
         let mut amended_insight_id: Option<i32> = None;
         if req.amend {
-            let title = self.generate_title(&backend, &final_content).await?;
+            let (title, body) = crate::ai::insight_generator::parse_title_body(&final_content);
+            let final_content = body;
 
             // Amended rows intentionally do not inherit the parent's
             // `fewshot_source_ids`. The parent's few-shot influence is still
@@ -1001,7 +1002,7 @@ impl InsightChatService {
             final_content,
         } = outcome;
 
-        let title = self.generate_title(&backend, &final_content).await?;
+        let (title, body) = crate::ai::insight_generator::parse_title_body(&final_content);
 
         let json = serde_json::to_string(&messages)
             .map_err(|e| anyhow!("failed to serialize chat history: {}", e))?;
@@ -1009,7 +1010,7 @@ impl InsightChatService {
             library_id: req.library_id,
             file_path: normalized.clone(),
             title,
-            summary: final_content,
+            summary: body,
             generated_at: Utc::now().timestamp(),
             model_version: model_used.clone(),
             is_current: true,
@@ -1045,32 +1046,6 @@ impl InsightChatService {
         Ok(())
     }
 
-    /// Generate a short title via the same chat backend so voice stays
-    /// consistent with the body. Mirrors generate_agentic_insight_for_photo's
-    /// titling pass.
-    async fn generate_title(
-        &self,
-        backend: &ResolvedBackend,
-        final_content: &str,
-    ) -> Result<String> {
-        let title_prompt = format!(
-            "Create a short title (maximum 8 words) for the following journal entry:\n\n{}\n\n\
-             Capture the key moment or theme. Return ONLY the title, nothing else.",
-            final_content
-        );
-        let title_raw = backend
-            .chat()
-            .generate(
-                &title_prompt,
-                Some(
-                    "You are my long term memory assistant. Use only the information provided. Do not invent details.",
-                ),
-                None,
-            )
-            .await?;
-        Ok(title_raw.trim().trim_matches('"').to_string())
-    }
-
     /// Drive the agentic loop with streaming SSE events. Shared between
     /// bootstrap and continuation. Mutates `messages` in place (response
     /// turns + tool results are appended) and returns counters + the