Add conversation history support to chat API

- Added HISTORY_TURNS env variable
- Updated chat request data creation
- Modified server to handle history
- Added history to test cases

.env.example

@@ -43,6 +43,7 @@ MAX_ASYNC=4
 MAX_TOKENS=32768
 EMBEDDING_DIM=1024
 MAX_EMBED_TOKENS=8192
+#HISTORY_TURNS=3
 
 # Security (empty for no key)
 LIGHTRAG_API_KEY=your-secure-api-key-here

lightrag/api/lightrag_server.py

@@ -470,6 +470,13 @@ def parse_args() -> argparse.Namespace:
         help="Enable automatic scanning when the program starts",
     )
 
+    parser.add_argument(
+        "--history-turns",
+        type=int,
+        default=get_env_value("HISTORY_TURNS", None, int),
+        help="Number of conversation history turns to include (default: from env or None)",
+    )
+
     args = parser.parse_args()
 
     return args
@@ -1576,8 +1583,9 @@ def create_app(args):
             if not messages:
                 raise HTTPException(status_code=400, detail="No messages provided")
 
-            # Get the last message as query
+            # Get the last message as query and previous messages as history
             query = messages[-1].content
+            conversation_history = messages[:-1]  # 所有之前的消息作为历史记录
 
             # Check for query prefix
             cleaned_query, mode = parse_query_mode(query)
@@ -1585,9 +1593,19 @@ def create_app(args):
             start_time = time.time_ns()
             prompt_tokens = estimate_tokens(cleaned_query)
 
-            query_param = QueryParam(
+            # 构建 query_param
-                mode=mode, stream=request.stream, only_need_context=False
+            param_dict = {
-            )
+                "mode": mode,
+                "stream": request.stream,
+                "only_need_context": False,
+                "conversation_history": conversation_history,
+            }
+
+            # 如果设置了 history_turns，添加到参数中
+            if args.history_turns is not None:
+                param_dict["history_turns"] = args.history_turns
+
+            query_param = QueryParam(**param_dict)
 
             if request.stream:
                 from fastapi.responses import StreamingResponse

test_lightrag_ollama_chat.py

@@ -189,19 +189,32 @@ def get_base_url(endpoint: str = "chat") -> str:
 
 
 def create_chat_request_data(
-    content: str, stream: bool = False, model: str = None
+    content: str,
+    stream: bool = False,
+    model: str = None,
+    conversation_history: List[Dict[str, str]] = None,
+    history_turns: int = None,
 ) -> Dict[str, Any]:
     """Create chat request data
     Args:
         content: User message content
         stream: Whether to use streaming response
         model: Model name
+        conversation_history: List of previous conversation messages
+        history_turns: Number of history turns to include
     Returns:
         Dictionary containing complete chat request data
     """
+    messages = conversation_history or []
+    if history_turns is not None and conversation_history:
+        messages = messages[
+            -2 * history_turns :
+        ]  # Each turn has 2 messages (user + assistant)
+    messages.append({"role": "user", "content": content})
+
     return {
         "model": model or CONFIG["server"]["model"],
-        "messages": [{"role": "user", "content": content}],
+        "messages": messages,
         "stream": stream,
     }
 
@@ -259,11 +272,25 @@ def run_test(func: Callable, name: str) -> None:
 def test_non_stream_chat() -> None:
     """Test non-streaming call to /api/chat endpoint"""
     url = get_base_url()
-    data = create_chat_request_data(
-        CONFIG["test_cases"]["basic"]["query"], stream=False
-    )
 
-    # Send request
+    # Example conversation history
+    conversation_history = [
+        {"role": "user", "content": "你好"},
+        {"role": "assistant", "content": "你好!我是一个AI助手,很高兴为你服务。"},
+        {"role": "user", "content": "西游记里有几个主要人物?"},
+        {
+            "role": "assistant",
+            "content": "西游记的主要人物有唐僧、孙悟空、猪八戒、沙和尚这四位主角。",
+        },
+    ]
+
+    # Send request with conversation history and history turns
+    data = create_chat_request_data(
+        CONFIG["test_cases"]["basic"]["query"],
+        stream=False,
+        conversation_history=conversation_history,
+        history_turns=2,  # Only include last 2 turns
+    )
     response = make_request(url, data)
 
     # Print response
@@ -297,9 +324,25 @@ def test_stream_chat() -> None:
     The last message will contain performance statistics, with done set to true.
     """
     url = get_base_url()
-    data = create_chat_request_data(CONFIG["test_cases"]["basic"]["query"], stream=True)
 
-    # Send request and get streaming response
+    # Example conversation history
+    conversation_history = [
+        {"role": "user", "content": "你好"},
+        {"role": "assistant", "content": "你好!我是一个AI助手,很高兴为你服务。"},
+        {"role": "user", "content": "西游记里有几个主要人物?"},
+        {
+            "role": "assistant",
+            "content": "西游记的主要人物有唐僧、孙悟空、猪八戒、沙和尚这四位主角。",
+        },
+    ]
+
+    # Send request with conversation history and history turns
+    data = create_chat_request_data(
+        CONFIG["test_cases"]["basic"]["query"],
+        stream=True,
+        conversation_history=conversation_history,
+        history_turns=2,  # Only include last 2 turns
+    )
     response = make_request(url, data, stream=True)
 
     if OutputControl.is_verbose():