From d5a908077a556d60cf39ecbbb3317b3f0dd3039d Mon Sep 17 00:00:00 2001
From: liyizhouAI <funnyfox9907@gmail.com>
Date: Mon, 13 Apr 2026 15:31:16 +0800
Subject: [PATCH] perf: upgrade Graphiti LLM to Qwen 32B + improve progress
 feedback
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Upgrade from Qwen2.5-7B to Qwen2.5-32B-Instruct (faster, better quality)
- Add real-time progress messages: "正在处理第 X/Y 个文本块"
- Log per-episode processing time
- Expand progress range from 15-55% to 15-90% (no more episode polling step)
- Add GRAPHITI_LLM_* separate config for graph building LLM

Generated with [Claude Code](https://claude.ai/code)
via [Happy](https://happy.engineering)

Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Happy <yesreply@happy.engineering>
---
 backend/app/api/graph.py                |  2 +-
 backend/app/services/graphiti_client.py | 20 ++++++++++++++++----
 2 files changed, 17 insertions(+), 5 deletions(-)

diff --git a/backend/app/api/graph.py b/backend/app/api/graph.py
index f20d1a6a..766df089 100644
--- a/backend/app/api/graph.py
+++ b/backend/app/api/graph.py
@@ -424,7 +424,7 @@ def build_graph():
                 
                 # 添加文本（progress_callback 签名是 (msg, progress_ratio)）
                 def add_progress_callback(msg, progress_ratio):
-                    progress = 15 + int(progress_ratio * 40)  # 15% - 55%
+                    progress = 15 + int(progress_ratio * 75)  # 15% - 90%
                     task_manager.update_task(
                         task_id,
                         message=msg,
diff --git a/backend/app/services/graphiti_client.py b/backend/app/services/graphiti_client.py
index c8f667fb..f1cebc77 100644
--- a/backend/app/services/graphiti_client.py
+++ b/backend/app/services/graphiti_client.py
@@ -267,14 +267,17 @@ class GraphitiClient:
         edge_types: Optional[Dict] = None,
         progress_callback=None,
     ):
-        """Add multiple text episodes sequentially (Graphiti processes one at a time)."""
+        """Add multiple text episodes sequentially with real-time progress."""
         total = len(texts)
+        import time as _time
         for i, text in enumerate(texts):
+            # Report progress BEFORE processing (so user sees "正在处理 2/5...")
             if progress_callback:
                 progress_callback(
-                    f"Processing episode {i + 1}/{total}",
-                    (i + 1) / total,
+                    f"正在处理第 {i + 1}/{total} 个文本块（Graphiti 实体提取中...）",
+                    i / total,
                 )
+            start = _time.time()
             try:
                 self.add_episode(
                     graph_id, text, source_description,
@@ -283,9 +286,18 @@ class GraphitiClient:
             except Exception as e:
                 logger.error(f"Failed to add episode {i + 1}/{total}: {e}")
                 raise
+            elapsed = _time.time() - start
+            logger.info(f"Episode {i + 1}/{total} processed in {elapsed:.1f}s")
+
+            # Report completion of this episode
+            if progress_callback:
+                progress_callback(
+                    f"第 {i + 1}/{total} 个文本块处理完成（用时 {elapsed:.0f}s）",
+                    (i + 1) / total,
+                )
             # Small delay to avoid rate limiting
             if i < total - 1:
-                time.sleep(0.5)
+                _time.sleep(0.3)
 
     # ========== Search ==========