Refa: make RAGFlow more asynchronous 2 (#11664)

### What problem does this PR solve?

Make RAGFlow more asynchronous 2. #11551, #11579, #11619.

### Type of change

- [x] New Feature (non-breaking change which adds functionality)
- [x] Refactoring
- [x] Performance Improvement

agent/canvas.py

@@ -415,13 +415,18 @@ class Canvas(Graph):
 
             loop = asyncio.get_running_loop()
             tasks = []
+            def _run_async_in_thread(coro_func, **call_kwargs):
+                return asyncio.run(coro_func(**call_kwargs))
+
             i = f
             while i < t:
                 cpn = self.get_component_obj(self.path[i])
                 task_fn = None
+                call_kwargs = None
 
                 if cpn.component_name.lower() in ["begin", "userfillup"]:
-                    task_fn = partial(cpn.invoke, inputs=kwargs.get("inputs", {}))
+                    call_kwargs = {"inputs": kwargs.get("inputs", {})}
+                    task_fn = cpn.invoke
                     i += 1
                 else:
                     for _, ele in cpn.get_input_elements().items():
@@ -430,13 +435,18 @@ class Canvas(Graph):
                             t -= 1
                             break
                     else:
-                        task_fn = partial(cpn.invoke, **cpn.get_input())
+                        call_kwargs = cpn.get_input()
+                        task_fn = cpn.invoke
                         i += 1
 
                 if task_fn is None:
                     continue
 
-                tasks.append(loop.run_in_executor(self._thread_pool, task_fn))
+                invoke_async = getattr(cpn, "invoke_async", None)
+                if invoke_async and asyncio.iscoroutinefunction(invoke_async):
+                    tasks.append(loop.run_in_executor(self._thread_pool, partial(_run_async_in_thread, invoke_async, **(call_kwargs or {}))))
+                else:
+                    tasks.append(loop.run_in_executor(self._thread_pool, partial(task_fn, **(call_kwargs or {}))))
 
             if tasks:
                 await asyncio.gather(*tasks)

agent/component/agent_with_tools.py

@@ -13,6 +13,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import asyncio
 import json
 import logging
 import os
@@ -239,6 +240,86 @@ class Agent(LLM, ToolBase):
             self.set_output("use_tools", use_tools)
         return ans
 
+    async def invoke_async(self, **kwargs):
+        """
+        Async entry: reuse existing logic but offload heavy sync parts via async wrappers to reduce blocking.
+        """
+        if self.check_if_canceled("Agent processing"):
+            return
+
+        if kwargs.get("user_prompt"):
+            usr_pmt = ""
+            if kwargs.get("reasoning"):
+                usr_pmt += "\nREASONING:\n{}\n".format(kwargs["reasoning"])
+            if kwargs.get("context"):
+                usr_pmt += "\nCONTEXT:\n{}\n".format(kwargs["context"])
+            if usr_pmt:
+                usr_pmt += "\nQUERY:\n{}\n".format(str(kwargs["user_prompt"]))
+            else:
+                usr_pmt = str(kwargs["user_prompt"])
+            self._param.prompts = [{"role": "user", "content": usr_pmt}]
+
+        if not self.tools:
+            if self.check_if_canceled("Agent processing"):
+                return
+            return await asyncio.to_thread(LLM._invoke, self, **kwargs)
+
+        prompt, msg, user_defined_prompt = self._prepare_prompt_variables()
+        output_schema = self._get_output_schema()
+        schema_prompt = ""
+        if output_schema:
+            schema = json.dumps(output_schema, ensure_ascii=False, indent=2)
+            schema_prompt = structured_output_prompt(schema)
+
+        downstreams = self._canvas.get_component(self._id)["downstream"] if self._canvas.get_component(self._id) else []
+        ex = self.exception_handler()
+        if any([self._canvas.get_component_obj(cid).component_name.lower()=="message" for cid in downstreams]) and not (ex and ex["goto"]) and not output_schema:
+            self.set_output("content", partial(self.stream_output_with_tools_async, prompt, msg, user_defined_prompt))
+            return
+
+        _, msg = message_fit_in([{"role": "system", "content": prompt}, *msg], int(self.chat_mdl.max_length * 0.97))
+        use_tools = []
+        ans = ""
+        async for delta_ans, tk in self._react_with_tools_streamly_async(prompt, msg, use_tools, user_defined_prompt, schema_prompt=schema_prompt):
+            if self.check_if_canceled("Agent processing"):
+                return
+            ans += delta_ans
+
+        if ans.find("**ERROR**") >= 0:
+            logging.error(f"Agent._chat got error. response: {ans}")
+            if self.get_exception_default_value():
+                self.set_output("content", self.get_exception_default_value())
+            else:
+                self.set_output("_ERROR", ans)
+            return
+
+        if output_schema:
+            error = ""
+            for _ in range(self._param.max_retries + 1):
+                try:
+                    def clean_formated_answer(ans: str) -> str:
+                        ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
+                        ans = re.sub(r"^.*```json", "", ans, flags=re.DOTALL)
+                        return re.sub(r"```\n*$", "", ans, flags=re.DOTALL)
+                    obj = json_repair.loads(clean_formated_answer(ans))
+                    self.set_output("structured", obj)
+                    if use_tools:
+                        self.set_output("use_tools", use_tools)
+                    return obj
+                except Exception:
+                    error = "The answer cannot be parsed as JSON"
+                    ans = self._force_format_to_schema(ans, schema_prompt)
+                    if ans.find("**ERROR**") >= 0:
+                        continue
+
+            self.set_output("_ERROR", error)
+            return
+
+        self.set_output("content", ans)
+        if use_tools:
+            self.set_output("use_tools", use_tools)
+        return ans
+
     def stream_output_with_tools(self, prompt, msg, user_defined_prompt={}):
         _, msg = message_fit_in([{"role": "system", "content": prompt}, *msg], int(self.chat_mdl.max_length * 0.97))
         answer_without_toolcall = ""
@@ -261,6 +342,54 @@ class Agent(LLM, ToolBase):
         if use_tools:
             self.set_output("use_tools", use_tools)
 
+    async def stream_output_with_tools_async(self, prompt, msg, user_defined_prompt={}):
+        _, msg = message_fit_in([{"role": "system", "content": prompt}, *msg], int(self.chat_mdl.max_length * 0.97))
+        answer_without_toolcall = ""
+        use_tools = []
+        async for delta_ans, _ in self._react_with_tools_streamly_async(prompt, msg, use_tools, user_defined_prompt):
+            if self.check_if_canceled("Agent streaming"):
+                return
+
+            if delta_ans.find("**ERROR**") >= 0:
+                if self.get_exception_default_value():
+                    self.set_output("content", self.get_exception_default_value())
+                    yield self.get_exception_default_value()
+                else:
+                    self.set_output("_ERROR", delta_ans)
+                    return
+            answer_without_toolcall += delta_ans
+            yield delta_ans
+
+        self.set_output("content", answer_without_toolcall)
+        if use_tools:
+            self.set_output("use_tools", use_tools)
+
+    async def _react_with_tools_streamly_async(self, prompt, history: list[dict], use_tools, user_defined_prompt={}, schema_prompt: str = ""):
+        """
+        Async wrapper that offloads synchronous flow to a thread, yielding results without blocking the event loop.
+        """
+        loop = asyncio.get_running_loop()
+        queue: asyncio.Queue = asyncio.Queue()
+
+        def worker():
+            try:
+                for delta_ans, tk in self._react_with_tools_streamly(prompt, history, use_tools, user_defined_prompt, schema_prompt=schema_prompt):
+                    asyncio.run_coroutine_threadsafe(queue.put((delta_ans, tk)), loop)
+            except Exception as e:
+                asyncio.run_coroutine_threadsafe(queue.put(e), loop)
+            finally:
+                asyncio.run_coroutine_threadsafe(queue.put(StopAsyncIteration), loop)
+
+        await asyncio.to_thread(worker)
+
+        while True:
+            item = await queue.get()
+            if item is StopAsyncIteration:
+                break
+            if isinstance(item, Exception):
+                raise item
+            yield item
+
     def _gen_citations(self, text):
         retrievals = self._canvas.get_reference()
         retrievals = {"chunks": list(retrievals["chunks"].values()), "doc_aggs": list(retrievals["doc_aggs"].values())}
@@ -433,4 +562,3 @@ Respond immediately with your final comprehensive answer.
         for k in self._param.inputs.keys():
             self._param.inputs[k]["value"] = None
         self._param.debug_inputs = {}
-

agent/component/llm.py

@@ -13,12 +13,14 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 #
+import asyncio
 import json
 import logging
 import os
 import re
+import threading
 from copy import deepcopy
-from typing import Any, Generator
+from typing import Any, Generator, AsyncGenerator
 import json_repair
 from functools import partial
 from common.constants import LLMType
@@ -171,6 +173,13 @@ class LLM(ComponentBase):
             return self.chat_mdl.chat(msg[0]["content"], msg[1:], self._param.gen_conf(), **kwargs)
         return self.chat_mdl.chat(msg[0]["content"], msg[1:], self._param.gen_conf(), images=self.imgs, **kwargs)
 
+    async def _generate_async(self, msg: list[dict], **kwargs) -> str:
+        if not self.imgs and hasattr(self.chat_mdl, "async_chat"):
+            return await self.chat_mdl.async_chat(msg[0]["content"], msg[1:], self._param.gen_conf(), **kwargs)
+        if self.imgs and hasattr(self.chat_mdl, "async_chat"):
+            return await self.chat_mdl.async_chat(msg[0]["content"], msg[1:], self._param.gen_conf(), images=self.imgs, **kwargs)
+        return await asyncio.to_thread(self._generate, msg, **kwargs)
+
     def _generate_streamly(self, msg:list[dict], **kwargs) -> Generator[str, None, None]:
         ans = ""
         last_idx = 0
@@ -205,6 +214,70 @@ class LLM(ComponentBase):
             for txt in self.chat_mdl.chat_streamly(msg[0]["content"], msg[1:], self._param.gen_conf(), images=self.imgs, **kwargs):
                 yield delta(txt)
 
+    async def _generate_streamly_async(self, msg: list[dict], **kwargs) -> AsyncGenerator[str, None]:
+        # Prefer async chat_streamly if available
+        async def delta_wrapper(txt_iter):
+            ans = ""
+            last_idx = 0
+            endswith_think = False
+
+            def delta(txt):
+                nonlocal ans, last_idx, endswith_think
+                delta_ans = txt[last_idx:]
+                ans = txt
+
+                if delta_ans.find("<think>") == 0:
+                    last_idx += len("<think>")
+                    return "<think>"
+                elif delta_ans.find("<think>") > 0:
+                    delta_ans = txt[last_idx:last_idx + delta_ans.find("<think>")]
+                    last_idx += delta_ans.find("<think>")
+                    return delta_ans
+                elif delta_ans.endswith("</think>"):
+                    endswith_think = True
+                elif endswith_think:
+                    endswith_think = False
+                    return "</think>"
+
+                last_idx = len(ans)
+                if ans.endswith("</think>"):
+                    last_idx -= len("</think>")
+                return re.sub(r"(<think>|</think>)", "", delta_ans)
+
+            async for t in txt_iter:
+                yield delta(t)
+
+        if not self.imgs and hasattr(self.chat_mdl, "async_chat_streamly"):
+            async for t in delta_wrapper(self.chat_mdl.async_chat_streamly(msg[0]["content"], msg[1:], self._param.gen_conf(), **kwargs)):
+                yield t
+            return
+        if self.imgs and hasattr(self.chat_mdl, "async_chat_streamly"):
+            async for t in delta_wrapper(self.chat_mdl.async_chat_streamly(msg[0]["content"], msg[1:], self._param.gen_conf(), images=self.imgs, **kwargs)):
+                yield t
+            return
+
+        # Fallback: run sync stream in thread, bridge results
+        loop = asyncio.get_running_loop()
+        queue: asyncio.Queue = asyncio.Queue()
+
+        def worker():
+            try:
+                for item in self._generate_streamly(msg, **kwargs):
+                    loop.call_soon_threadsafe(queue.put_nowait, item)
+            except Exception as e:
+                loop.call_soon_threadsafe(queue.put_nowait, e)
+            finally:
+                loop.call_soon_threadsafe(queue.put_nowait, StopAsyncIteration)
+
+        threading.Thread(target=worker, daemon=True).start()
+        while True:
+            item = await queue.get()
+            if item is StopAsyncIteration:
+                break
+            if isinstance(item, Exception):
+                raise item
+            yield item
+
     async def _stream_output_async(self, prompt, msg):
         _, msg = message_fit_in([{"role": "system", "content": prompt}, *msg], int(self.chat_mdl.max_length * 0.97))
         answer = ""

agent/tools/base.py

@@ -17,6 +17,7 @@ import logging
 import re
 import time
 from copy import deepcopy
+import asyncio
 from functools import partial
 from typing import TypedDict, List, Any
 from agent.component.base import ComponentParamBase, ComponentBase
@@ -50,10 +51,14 @@ class LLMToolPluginCallSession(ToolCallSession):
     def tool_call(self, name: str, arguments: dict[str, Any]) -> Any:
         assert name in self.tools_map, f"LLM tool {name} does not exist"
         st = timer()
-        if isinstance(self.tools_map[name], MCPToolCallSession):
-            resp = self.tools_map[name].tool_call(name, arguments, 60)
+        tool_obj = self.tools_map[name]
+        if isinstance(tool_obj, MCPToolCallSession):
+            resp = tool_obj.tool_call(name, arguments, 60)
         else:
-            resp = self.tools_map[name].invoke(**arguments)
+            if hasattr(tool_obj, "invoke_async") and asyncio.iscoroutinefunction(tool_obj.invoke_async):
+                resp = asyncio.run(tool_obj.invoke_async(**arguments))
+            else:
+                resp = asyncio.run(asyncio.to_thread(tool_obj.invoke, **arguments))
 
         self.callback(name, arguments, resp, elapsed_time=timer()-st)
         return resp
@@ -139,6 +144,30 @@ class ToolBase(ComponentBase):
         self.set_output("_elapsed_time", time.perf_counter() - self.output("_created_time"))
         return res
 
+    async def invoke_async(self, **kwargs):
+        """
+        Async wrapper for tool invocation.
+        If `_invoke` is a coroutine, await it directly; otherwise run in a thread to avoid blocking.
+        Mirrors the exception handling of `invoke`.
+        """
+        if self.check_if_canceled("Tool processing"):
+            return
+
+        self.set_output("_created_time", time.perf_counter())
+        try:
+            if asyncio.iscoroutinefunction(self._invoke):
+                res = await self._invoke(**kwargs)
+            else:
+                res = await asyncio.to_thread(self._invoke, **kwargs)
+        except Exception as e:
+            self._param.outputs["_ERROR"] = {"value": str(e)}
+            logging.exception(e)
+            res = str(e)
+        self._param.debug_inputs = []
+
+        self.set_output("_elapsed_time", time.perf_counter() - self.output("_created_time"))
+        return res
+
     def _retrieve_chunks(self, res_list: list, get_title, get_url, get_content, get_score=None):
         chunks = []
         aggs = []