From 650cb709615ceacdb8a9a47a2c3c83c353d4ce19 Mon Sep 17 00:00:00 2001
From: "Rodrigo Rodriguez (Pragmatismo)" <me@rodrigorodriguez.com>
Date: Mon, 13 Apr 2026 20:46:28 -0300
Subject: [PATCH] debug: add WebSocket message tracing

---
 src/core/bot/mod.rs            | 12 ++++++++++--
 src/drive/drive_monitor/mod.rs | 21 +++++++++++++++++++++
 2 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/src/core/bot/mod.rs b/src/core/bot/mod.rs
index f4662186..5b964a5a 100644
--- a/src/core/bot/mod.rs
+++ b/src/core/bot/mod.rs
@@ -10,7 +10,7 @@ use tool_executor::ToolExecutor;
 use crate::core::config::ConfigManager;
 
 #[cfg(feature = "drive")]
-use crate::drive::drive_monitor::DriveMonitor;
+use crate::drive::drive_monitor::{DriveMonitor, set_llm_streaming};
 #[cfg(feature = "llm")]
 use crate::llm::llm_models;
 #[cfg(feature = "llm")]
@@ -811,13 +811,20 @@ impl BotOrchestrator {
         // Clone messages for the async task
         let messages_clone = messages.clone();
 
+        // Set flag to prevent DriveMonitor PDF downloads during LLM streaming
+        #[cfg(feature = "drive")]
+        set_llm_streaming(true);
+
+        let stream_tx_clone = stream_tx.clone();
         tokio::spawn(async move {
             if let Err(e) = llm
-                .generate_stream("", &messages_clone, stream_tx, &model_clone, &key_clone, tools_for_llm.as_ref())
+                .generate_stream("", &messages_clone, stream_tx_clone, &model_clone, &key_clone, tools_for_llm.as_ref())
                 .await
             {
                 error!("LLM streaming error: {}", e);
             }
+            #[cfg(feature = "drive")]
+            set_llm_streaming(false);
         });
 
         let mut full_response = String::new();
@@ -1570,6 +1577,7 @@ let mut send_task = tokio::spawn(async move {
         while let Some(Ok(msg)) = receiver.next().await {
             match msg {
                 Message::Text(text) => {
+                    debug!("WebSocket received text: {}", text);
                     if let Ok(user_msg) = serde_json::from_str::<UserMessage>(&text) {
                         let orchestrator = BotOrchestrator::new(state_clone.clone());
                         if let Some(tx_clone) = state_clone
diff --git a/src/drive/drive_monitor/mod.rs b/src/drive/drive_monitor/mod.rs
index a674274b..bde15028 100644
--- a/src/drive/drive_monitor/mod.rs
+++ b/src/drive/drive_monitor/mod.rs
@@ -23,6 +23,19 @@ use tokio::time::Duration;
 use serde::{Deserialize, Serialize};
 use tokio::fs as tokio_fs;
 
+#[cfg(any(feature = "research", feature = "llm"))]
+static LLM_STREAMING: std::sync::atomic::AtomicBool = std::sync::atomic::AtomicBool::new(false);
+
+#[cfg(any(feature = "research", feature = "llm"))]
+pub fn set_llm_streaming(streaming: bool) {
+    LLM_STREAMING.store(streaming, Ordering::SeqCst);
+}
+
+#[cfg(any(feature = "research", feature = "llm"))]
+pub fn is_llm_streaming() -> bool {
+    LLM_STREAMING.load(Ordering::SeqCst)
+}
+
 const MAX_BACKOFF_SECS: u64 = 300;
 const INITIAL_BACKOFF_SECS: u64 = 30;
 const RETRY_BACKOFF_SECS: i64 = 3600;
@@ -1672,6 +1685,14 @@ let file_state = FileState {
                 files_processed += 1;
                 debug!("[GBKB] Queue size: {}/10", files_to_process.len());
 
+                // Skip downloads if LLM is actively streaming to prevent deadlock
+                #[cfg(any(feature = "research", feature = "llm"))]
+                if is_llm_streaming() {
+                    debug!("[GBKB] Skipping download - LLM is streaming, will retry later");
+                    files_to_process.clear();
+                    break;
+                }
+
                 if files_to_process.len() >= 10 {
                     debug!("[GBKB] Downloading batch of {} files", files_to_process.len());
                     for file_path in std::mem::take(&mut files_to_process) {