vtcode 0.39.0

A Rust-based terminal coding agent with modular architecture supporting multiple LLM providers
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71

diff --git a/vtcode-core/src/core/conversation_summarizer.rs b/vtcode-core/src/core/conversation_summarizer.rs
index 1234567..abcdefg 100644
--- a/vtcode-core/src/core/conversation_summarizer.rs
+++ b/vtcode-core/src/core/conversation_summarizer.rs
@@ -85,7 +85,7 @@ impl ConversationSummarizer {
             enabled: true,
             last_summary_turn: None,
             summaries: Vec::new(),
-            context_trigger_percent: 60,
+            context_trigger_percent: 85,  // Changed from 60% to 85% to match spec
             error_trigger_percent: 80,
             min_turn_gap: 5,
             summarization_threshold,

diff --git a/src/agent/runloop/unified/turn.rs b/src/agent/runloop/unified/turn.rs
index 1234567..abcdefg 100644
--- a/src/agent/runloop/unified/turn.rs
+++ b/src/agent/runloop/unified/turn.rs
@@ -1380,6 +1380,52 @@ pub(crate) async fn run_single_agent_loop_unified(
                 .collect();
             ledger.update_available_tools(tool_names);
         }
+        
+        // Check if automatic summarization should trigger
+        // This prevents context overflow and agent blocking
+        let conversation_len = working_history.len();
+        let should_summarize = if token_budget_enabled {
+            let budget = context_manager.token_budget();
+            let current_tokens = budget.current_usage().await.unwrap_or(0);
+            let max_tokens = budget.max_tokens().unwrap_or(100_000);
+            let token_percent = if max_tokens > 0 {
+                (current_tokens as f64 / max_tokens as f64 * 100.0) as usize
+            } else {
+                0
+            };
+            
+            // Trigger if: 20+ turns OR 85%+ token usage
+            conversation_len >= 20 || token_percent >= 85
+        } else {
+            conversation_len >= 20
+        };
+        
+        if should_summarize {
+            renderer.line(
+                MessageStyle::Info,
+                &format!("⚡ Context optimization: {} messages", conversation_len)
+            )?;
+            
+            // Simple compression: keep system + recent 15 messages
+            let keep_recent = 15;
+            if working_history.len() > keep_recent {
+                let mut compressed = Vec::new();
+                
+                // Keep first system message if exists
+                if let Some(first) = working_history.first() {
+                    if first.role == "system" {
+                        compressed.push(first.clone());
+                    }
+                }
+                
+                // Keep recent messages
+                compressed.extend(
+                    working_history.iter().rev().take(keep_recent).rev().cloned()
+                );
+                
+                working_history = compressed;
+            }
+        }

         let mut attempt_history = working_history.clone();
         let mut retry_attempts = 0usize;