feat(automation): improve prompt compaction with async LLM summarization

- Added initial 30s delay to compact prompt scheduler
- Implemented async LLM summarization for conversation history
- Reduced lock contention by minimizing critical sections
- Added fallback to original text if summarization fails
- Updated README with guidance for failed requirements
- Added new `summarize` method to LLMProvider trait
- Improved session manager query with proper DSL usage

The changes optimize the prompt compaction process by:
1. Reducing lock contention through better resource management
2. Adding LLM-based summarization for better conversation compression
3. Making the system more resilient with proper error handling
4. Improving documentation for development practices

prompts/dev/platform/README.md

@@ -13,5 +13,6 @@ When initial attempts fail, sequentially try these LLMs:
 - **One requirement at a time** with sequential commits
 - **On error**: Stop and consult Claude for guidance
 - **Change progression**: Start with DeepSeek, conclude with gpt-oss-120b
+- If a big req. fail, specify a @code file that has similar pattern or sample from official docs.
 - **Final validation**: Use prompt "cargo check" with gpt-oss-120b
 - Be humble, one requirement, one commit. But sometimes, freedom of caos is welcome - when no deadlines are set.

src/automation/compact_prompt.rs

@@ -5,9 +5,12 @@ use diesel::prelude::*;
 use log::{error, info};
 use std::sync::Arc;
 use tokio::time::{interval, Duration};
+use uuid::Uuid;
 
 pub fn start_compact_prompt_scheduler(state: Arc<AppState>) {
     tokio::spawn(async move {
+        // Initial 30 second delay before first run
+        tokio::time::sleep(Duration::from_secs(30)).await;
         let mut interval = interval(Duration::from_secs(60));
         loop {
             interval.tick().await;
@@ -60,15 +63,22 @@ async fn compact_prompt_for_bot(
         return Ok(());
     }
 
-    let mut session_manager = state.session_manager.lock().await;
+    // Get sessions without holding lock
-    let sessions = session_manager.get_user_sessions(uuid::Uuid::nil())?;
+    let sessions = {
+        let mut session_manager = state.session_manager.lock().await;
+        session_manager.get_user_sessions(Uuid::nil())?
+    };
 
     for session in sessions {
         if session.bot_id != automation.bot_id {
             continue;
         }
 
-        let history = session_manager.get_conversation_history(session.id, session.user_id)?;
+        // Get history without holding lock
+        let history = {
+            let mut session_manager = state.session_manager.lock().await;
+            session_manager.get_conversation_history(session.id, session.user_id)?
+        };
 
         if history.len() > compact_threshold {
             info!(
@@ -82,8 +92,29 @@ async fn compact_prompt_for_bot(
                 compacted.push_str(&format!("{}: {}\n", role, content));
             }
 
-            let summarized = format!("SUMMARY: {}", compacted);
+            // Clone needed references for async task
-            session_manager.save_message(session.id, session.user_id, 3, &summarized, 1)?;
+            let llm_provider = state.llm_provider.clone();
+            let compacted_clone = compacted.clone();
+            
+            // Run LLM summarization
+            let summarized = match llm_provider.generate(&compacted_clone, &serde_json::Value::Null).await {
+                Ok(summary) => format!("SUMMARY: {}", summary),
+                Err(e) => {
+                    error!("Failed to summarize conversation: {}", e);
+                    format!("SUMMARY: {}", compacted) // Fallback
+                }
+            };
+            info!(
+                "Prompt compacted {}: {} messages",
+                session.id,
+                history.len()
+            );
+
+            // Save with minimal lock time
+            {
+                let mut session_manager = state.session_manager.lock().await;
+                session_manager.save_message(session.id, session.user_id, 3, &summarized, 1)?;
+            }
         }
     }
 

src/llm/mod.rs

@@ -21,6 +21,14 @@ pub trait LLMProvider: Send + Sync {
         tx: mpsc::Sender<String>,
     ) -> Result<(), Box<dyn std::error::Error + Send + Sync>>;
 
+    async fn summarize(
+        &self,
+        text: &str,
+    ) -> Result<String, Box<dyn std::error::Error + Send + Sync>> {
+        let prompt = format!("Summarize the following conversation while preserving key details:\n\n{}", text);
+        self.generate(&prompt, &serde_json::json!({"max_tokens": 500}))
+            .await
+    }
 
     async fn cancel_job(
         &self,

src/session/mod.rs

@@ -331,11 +331,19 @@ impl SessionManager {
         &mut self,
         uid: Uuid,
     ) -> Result<Vec<UserSession>, Box<dyn Error + Send + Sync>> {
-        use crate::shared::models::user_sessions;
+        use crate::shared::models::user_sessions::dsl::*;
-        let sessions = user_sessions::table
+
-            .filter(user_sessions::user_id.eq(uid))
+        let sessions = if uid == Uuid::nil() {
-            .order(user_sessions::created_at.desc())
+            user_sessions
-            .load::<UserSession>(&mut self.conn)?;
+                .order(created_at.desc())
+                .load::<UserSession>(&mut self.conn)?
+        } else {
+            user_sessions
+                .filter(user_id.eq(uid))
+                .order(created_at.desc())
+                .load::<UserSession>(&mut self.conn)?
+        };
+
         Ok(sessions)
     }