Fix: Metadata bugs. (#12111)

### What problem does this PR solve?

Fix: Metadata bugs.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

---------

Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>

.github/workflows/tests.yml

@@ -209,7 +209,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api
+          source .venv/bin/activate && pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api | tee es_sdk_test.log
 
       - name: Run frontend api tests against Elasticsearch
         run: |
@@ -218,7 +218,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py
+          source .venv/bin/activate && pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py | tee es_api_test.log
           
       - name: Run http api tests against Elasticsearch
         run: |
@@ -227,7 +227,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api
+          source .venv/bin/activate && pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api | tee es_http_api_test.log
 
       - name: Stop ragflow:nightly
         if: always()  # always run this step even if previous steps failed
@@ -247,7 +247,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api > infinity_sdk_test.log
+          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_sdk_api | tee infinity_sdk_test.log
 
       - name: Run frontend api tests against Infinity
         run: |
@@ -256,7 +256,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py > infinity_api_test.log
+          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short sdk/python/test/test_frontend_api/get_email.py sdk/python/test/test_frontend_api/test_dataset.py | tee infinity_api_test.log
 
       - name: Run http api tests against Infinity
         run: |
@@ -265,7 +265,7 @@ jobs:
             echo "Waiting for service to be available..."
             sleep 5
           done
-          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api > infinity_http_api_test.log
+          source .venv/bin/activate && DOC_ENGINE=infinity pytest -s --tb=short --level=${HTTP_API_TEST_LEVEL} test/testcases/test_http_api | tee infinity_http_api_test.log
 
       - name: Stop ragflow:nightly
         if: always()  # always run this step even if previous steps failed

agent/canvas.py

@@ -540,6 +540,8 @@ class Canvas(Graph):
                         cite = re.search(r"\[ID:[ 0-9]+\]",  cpn_obj.output("content"))
 
                     message_end = {}
+                    if cpn_obj.get_param("status"):
+                        message_end["status"] = cpn_obj.get_param("status")
                     if isinstance(cpn_obj.output("attachment"), dict):
                         message_end["attachment"] = cpn_obj.output("attachment")
                     if cite:

agent/component/agent_with_tools.py

@@ -29,8 +29,8 @@ from api.db.services.llm_service import LLMBundle
 from api.db.services.tenant_llm_service import TenantLLMService
 from api.db.services.mcp_server_service import MCPServerService
 from common.connection_utils import timeout
-from rag.prompts.generator import next_step_async, COMPLETE_TASK, analyze_task_async, \
-    citation_prompt, reflect_async, kb_prompt, citation_plus, full_question, message_fit_in, structured_output_prompt
+from rag.prompts.generator import next_step_async, COMPLETE_TASK, \
+    citation_prompt, kb_prompt, citation_plus, full_question, message_fit_in, structured_output_prompt
 from common.mcp_tool_call_conn import MCPToolCallSession, mcp_tool_metadata_to_openai_tool
 from agent.component.llm import LLMParam, LLM
 
@@ -84,9 +84,10 @@ class Agent(LLM, ToolBase):
     def __init__(self, canvas, id, param: LLMParam):
         LLM.__init__(self, canvas, id, param)
         self.tools = {}
-        for cpn in self._param.tools:
+        for idx, cpn in enumerate(self._param.tools):
             cpn = self._load_tool_obj(cpn)
-            self.tools[cpn.get_meta()["function"]["name"]] = cpn
+            name = cpn.get_meta()["function"]["name"]
+            self.tools[f"{name}_{idx}"] = cpn
 
         self.chat_mdl = LLMBundle(self._canvas.get_tenant_id(), TenantLLMService.llm_id2llm_type(self._param.llm_id), self._param.llm_id,
                                   max_retries=self._param.max_retries,
@@ -202,7 +203,7 @@ class Agent(LLM, ToolBase):
         _, 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):
+        async for delta_ans, _tk in self._react_with_tools_streamly_async_simple(prompt, msg, use_tools, user_defined_prompt,schema_prompt=schema_prompt):
             if self.check_if_canceled("Agent processing"):
                 return
             ans += delta_ans
@@ -246,7 +247,7 @@ class Agent(LLM, ToolBase):
         _, 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):
+        async for delta_ans, _ in self._react_with_tools_streamly_async_simple(prompt, msg, use_tools, user_defined_prompt):
             if self.check_if_canceled("Agent streaming"):
                 return
 
@@ -264,7 +265,7 @@ class Agent(LLM, ToolBase):
         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 def _react_with_tools_streamly_async_simple(self, prompt, history: list[dict], use_tools, user_defined_prompt={}, schema_prompt: str = ""):
         token_count = 0
         tool_metas = self.tool_meta
         hist = deepcopy(history)
@@ -276,6 +277,28 @@ class Agent(LLM, ToolBase):
         else:
             user_request = history[-1]["content"]
 
+        def build_task_desc(prompt: str, user_request: str, tool_metas: list[dict], user_defined_prompt: dict | None = None) -> str:
+            """Build a minimal task_desc by concatenating prompt, query, and tool schemas."""
+            user_defined_prompt = user_defined_prompt or {}
+
+            tools_json = json.dumps(tool_metas, ensure_ascii=False, indent=2)
+
+            task_desc = (
+                "### Agent Prompt\n"
+                f"{prompt}\n\n"
+                "### User Request\n"
+                f"{user_request}\n\n"
+                "### Tools (schemas)\n"
+                f"{tools_json}\n"
+            )
+
+            if user_defined_prompt:
+                udp_json = json.dumps(user_defined_prompt, ensure_ascii=False, indent=2)
+                task_desc += "\n### User Defined Prompts\n" + udp_json + "\n"
+
+            return task_desc
+
+
         async def use_tool_async(name, args):
             nonlocal hist, use_tools, last_calling
             logging.info(f"{last_calling=} == {name=}")
@@ -286,9 +309,6 @@ class Agent(LLM, ToolBase):
                 "arguments": args,
                 "results": tool_response
             })
-            # self.callback("add_memory", {}, "...")
-            #self.add_memory(hist[-2]["content"], hist[-1]["content"], name, args, str(tool_response), user_defined_prompt)
-
             return name, tool_response
 
         async def complete():
@@ -326,6 +346,21 @@ class Agent(LLM, ToolBase):
 
             self.callback("gen_citations", {}, txt, elapsed_time=timer()-st)
 
+        def build_observation(tool_call_res: list[tuple]) -> str:
+            """
+            Build a Observation from tool call results.
+            No LLM involved.
+            """
+            if not tool_call_res:
+                return ""
+
+            lines = ["Observation:"]
+            for name, result in tool_call_res:
+                lines.append(f"[{name} result]")
+                lines.append(str(result))
+
+            return "\n".join(lines)
+
         def append_user_content(hist, content):
             if hist[-1]["role"] == "user":
                 hist[-1]["content"] += content
@@ -333,7 +368,7 @@ class Agent(LLM, ToolBase):
                 hist.append({"role": "user", "content": content})
 
         st = timer()
-        task_desc = await analyze_task_async(self.chat_mdl, prompt, user_request, tool_metas, user_defined_prompt)
+        task_desc = build_task_desc(prompt, user_request, tool_metas, user_defined_prompt)
         self.callback("analyze_task", {}, task_desc, elapsed_time=timer()-st)
         for _ in range(self._param.max_rounds + 1):
             if self.check_if_canceled("Agent streaming"):
@@ -364,7 +399,7 @@ class Agent(LLM, ToolBase):
 
                 results = await asyncio.gather(*tool_tasks) if tool_tasks else []
                 st = timer()
-                reflection = await reflect_async(self.chat_mdl, hist, results, user_defined_prompt)
+                reflection = build_observation(results)
                 append_user_content(hist, reflection)
                 self.callback("reflection", {}, str(reflection), elapsed_time=timer()-st)
 
@@ -393,6 +428,135 @@ Respond immediately with your final comprehensive answer.
         async for txt, tkcnt in complete():
             yield txt, tkcnt
 
+#     async def _react_with_tools_streamly_async(self, prompt, history: list[dict], use_tools, user_defined_prompt={}, schema_prompt: str = ""):
+#         token_count = 0
+#         tool_metas = self.tool_meta
+#         hist = deepcopy(history)
+#         last_calling = ""
+#         if len(hist) > 3:
+#             st = timer()
+#             user_request = await full_question(messages=history, chat_mdl=self.chat_mdl)
+#             self.callback("Multi-turn conversation optimization", {}, user_request, elapsed_time=timer()-st)
+#         else:
+#             user_request = history[-1]["content"]
+
+#         async def use_tool_async(name, args):
+#             nonlocal hist, use_tools, last_calling
+#             logging.info(f"{last_calling=} == {name=}")
+#             last_calling = name
+#             tool_response = await self.toolcall_session.tool_call_async(name, args)
+#             use_tools.append({
+#                 "name": name,
+#                 "arguments": args,
+#                 "results": tool_response
+#             })
+#             # self.callback("add_memory", {}, "...")
+#             #self.add_memory(hist[-2]["content"], hist[-1]["content"], name, args, str(tool_response), user_defined_prompt)
+
+#             return name, tool_response
+
+#         async def complete():
+#             nonlocal hist
+#             need2cite = self._param.cite and self._canvas.get_reference()["chunks"] and self._id.find("-->") < 0
+#             if schema_prompt:
+#                 need2cite = False
+#             cited = False
+#             if hist and hist[0]["role"] == "system":
+#                 if schema_prompt:
+#                     hist[0]["content"] += "\n" + schema_prompt
+#                 if need2cite and len(hist) < 7:
+#                     hist[0]["content"] += citation_prompt()
+#                     cited = True
+#             yield "", token_count
+
+#             _hist = hist
+#             if len(hist) > 12:
+#                 _hist = [hist[0], hist[1], *hist[-10:]]
+#             entire_txt = ""
+#             async for delta_ans in self._generate_streamly(_hist):
+#                 if not need2cite or cited:
+#                     yield delta_ans, 0
+#                 entire_txt += delta_ans
+#             if not need2cite or cited:
+#                 return
+
+#             st = timer()
+#             txt = ""
+#             async for delta_ans in self._gen_citations_async(entire_txt):
+#                 if self.check_if_canceled("Agent streaming"):
+#                     return
+#                 yield delta_ans, 0
+#                 txt += delta_ans
+
+#             self.callback("gen_citations", {}, txt, elapsed_time=timer()-st)
+
+#         def append_user_content(hist, content):
+#             if hist[-1]["role"] == "user":
+#                 hist[-1]["content"] += content
+#             else:
+#                 hist.append({"role": "user", "content": content})
+
+#         st = timer()
+#         task_desc = await analyze_task_async(self.chat_mdl, prompt, user_request, tool_metas, user_defined_prompt)
+#         self.callback("analyze_task", {}, task_desc, elapsed_time=timer()-st)
+#         for _ in range(self._param.max_rounds + 1):
+#             if self.check_if_canceled("Agent streaming"):
+#                 return
+#             response, tk = await next_step_async(self.chat_mdl, hist, tool_metas, task_desc, user_defined_prompt)
+#             # self.callback("next_step", {}, str(response)[:256]+"...")
+#             token_count += tk or 0
+#             hist.append({"role": "assistant", "content": response})
+#             try:
+#                 functions = json_repair.loads(re.sub(r"```.*", "", response))
+#                 if not isinstance(functions, list):
+#                     raise TypeError(f"List should be returned, but `{functions}`")
+#                 for f in functions:
+#                     if not isinstance(f, dict):
+#                         raise TypeError(f"An object type should be returned, but `{f}`")
+
+#                 tool_tasks = []
+#                 for func in functions:
+#                     name = func["name"]
+#                     args = func["arguments"]
+#                     if name == COMPLETE_TASK:
+#                         append_user_content(hist, f"Respond with a formal answer. FORGET(DO NOT mention) about `{COMPLETE_TASK}`. The language for the response MUST be as the same as the first user request.\n")
+#                         async for txt, tkcnt in complete():
+#                             yield txt, tkcnt
+#                         return
+
+#                     tool_tasks.append(asyncio.create_task(use_tool_async(name, args)))
+
+#                 results = await asyncio.gather(*tool_tasks) if tool_tasks else []
+#                 st = timer()
+#                 reflection = await reflect_async(self.chat_mdl, hist, results, user_defined_prompt)
+#                 append_user_content(hist, reflection)
+#                 self.callback("reflection", {}, str(reflection), elapsed_time=timer()-st)
+
+#             except Exception as e:
+#                 logging.exception(msg=f"Wrong JSON argument format in LLM ReAct response: {e}")
+#                 e = f"\nTool call error, please correct the input parameter of response format and call it again.\n *** Exception ***\n{e}"
+#                 append_user_content(hist, str(e))
+
+#         logging.warning( f"Exceed max rounds: {self._param.max_rounds}")
+#         final_instruction = f"""
+# {user_request}
+# IMPORTANT: You have reached the conversation limit. Based on ALL the information and research you have gathered so far, please provide a DIRECT and COMPREHENSIVE final answer to the original request.
+# Instructions:
+# 1. SYNTHESIZE all information collected during this conversation
+# 2. Provide a COMPLETE response using existing data - do not suggest additional research
+# 3. Structure your response as a FINAL DELIVERABLE, not a plan
+# 4. If information is incomplete, state what you found and provide the best analysis possible with available data
+# 5. DO NOT mention conversation limits or suggest further steps
+# 6. Focus on delivering VALUE with the information already gathered
+# Respond immediately with your final comprehensive answer.
+#         """
+#         if self.check_if_canceled("Agent final instruction"):
+#             return
+#         append_user_content(hist, final_instruction)
+
+#         async for txt, tkcnt in complete():
+#             yield txt, tkcnt
+
     async def _gen_citations_async(self, text):
         retrievals = self._canvas.get_reference()
         retrievals = {"chunks": list(retrievals["chunks"].values()), "doc_aggs": list(retrievals["doc_aggs"].values())}

api/apps/chunk_app.py

@@ -76,6 +76,7 @@ async def list_chunk():
                 "image_id": sres.field[id].get("img_id", ""),
                 "available_int": int(sres.field[id].get("available_int", 1)),
                 "positions": sres.field[id].get("position_int", []),
+                "doc_type_kwd": sres.field[id].get("doc_type_kwd")
             }
             assert isinstance(d["positions"], list)
             assert len(d["positions"]) == 0 or (isinstance(d["positions"][0], list) and len(d["positions"][0]) == 5)
@@ -176,10 +177,9 @@ async def set():
             settings.docStoreConn.update({"id": req["chunk_id"]}, _d, search.index_name(tenant_id), doc.kb_id)
 
             # update image
-            image_id = req.get("img_id")
-            bkt, name = image_id.split("-")
             image_base64 = req.get("image_base64", None)
             if image_base64:
+                bkt, name = req.get("img_id", "-").split("-")
                 image_binary = base64.b64decode(image_base64)
                 settings.STORAGE_IMPL.put(bkt, name, image_binary)
             return get_json_result(data=True)

api/apps/kb_app.py

@@ -150,6 +150,21 @@ async def update():
         return server_error_response(e)
 
 
+@manager.route('/update_metadata_setting', methods=['post'])  # noqa: F821
+@login_required
+@validate_request("kb_id", "metadata")
+async def update_metadata_setting():
+    req = await get_request_json()
+    e, kb = KnowledgebaseService.get_by_id(req["kb_id"])
+    if not e:
+        return get_data_error_result(
+            message="Database error (Knowledgebase rename)!")
+    kb = kb.to_dict()
+    kb["parser_config"]["metadata"] = req["metadata"]
+    KnowledgebaseService.update_by_id(kb["id"], kb)
+    return get_json_result(data=kb)
+
+
 @manager.route('/detail', methods=['GET'])  # noqa: F821
 @login_required
 def detail():

api/apps/sdk/agents.py

@@ -326,7 +326,6 @@ async def webhook(agent_id: str):
         secret = jwt_cfg.get("secret")
         if not secret:
             raise Exception("JWT secret not configured")
-        required_claims = jwt_cfg.get("required_claims", [])
 
         auth_header = request.headers.get("Authorization", "")
         if not auth_header.startswith("Bearer "):
@@ -750,7 +749,7 @@ async def webhook(agent_id: str):
         async def sse():
             nonlocal canvas
             contents: list[str] = []
-
+            status = 200
             try:
                 async for ans in canvas.run(
                     query="",
@@ -765,6 +764,8 @@ async def webhook(agent_id: str):
                             content = "</think>"
                         if content:
                             contents.append(content)
+                    if ans["event"] == "message_end":
+                        status = int(ans["data"].get("status", status))
                     if is_test:
                         append_webhook_trace(
                             agent_id,
@@ -782,7 +783,11 @@ async def webhook(agent_id: str):
                         }
                     )
                 final_content = "".join(contents)
-                yield json.dumps(final_content, ensure_ascii=False)
+                return {
+                    "message": final_content,
+                    "success": True,
+                    "code":  status,
+                }
 
             except Exception as e:
                 if is_test:
@@ -804,10 +809,14 @@ async def webhook(agent_id: str):
                             "success": False,
                         }
                     )
-                yield json.dumps({"code": 500, "message": str(e)}, ensure_ascii=False)
+                return {"code": 400, "message": str(e),"success":False}
 
-        resp = Response(sse(), mimetype="application/json")
-        return resp
+        result = await sse()
+        return Response(
+            json.dumps(result),
+            status=result["code"],
+            mimetype="application/json",
+        )
 
 
 @manager.route("/webhook_trace/<agent_id>", methods=["GET"])  # noqa: F821

api/db/services/knowledgebase_service.py

@@ -411,6 +411,8 @@ class KnowledgebaseService(CommonService):
         ok, _t = TenantService.get_by_id(tenant_id)
         if not ok:
             return False, get_data_error_result(message="Tenant not found.")
+        if kwargs.get("parser_config") and isinstance(kwargs["parser_config"], dict) and not kwargs["parser_config"].get("llm_id"):
+            kwargs["parser_config"]["llm_id"] = _t.llm_id
 
         # Build payload
         kb_id = get_uuid()

conf/infinity_mapping.json

@@ -31,6 +31,7 @@
 	"entity_type_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"source_id": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"n_hop_with_weight": {"type": "varchar", "default": ""},
+	"mom_with_weight": {"type": "varchar", "default": ""},
 	"removed_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"doc_type_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},
 	"toc_kwd": {"type": "varchar", "default": "", "analyzer": "whitespace-#"},

conf/llm_factories.json

@@ -762,6 +762,13 @@
             "status": "1",
             "rank": "940",
             "llm": [
+                {
+                    "llm_name": "glm-4.7",
+                    "tags": "LLM,CHAT,128K",
+                    "max_tokens": 128000,
+                    "model_type": "chat",
+                    "is_tools": true
+                },
                 {
                     "llm_name": "glm-4.5",
                     "tags": "LLM,CHAT,128K",

deepdoc/parser/figure_parser.py

@@ -38,8 +38,8 @@ def vision_figure_parser_figure_data_wrapper(figures_data_without_positions):
 
 
 def vision_figure_parser_docx_wrapper(sections, tbls, callback=None,**kwargs):
-    if not tbls:
-        return []
+    if not sections:
+        return tbls
     try:
         vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT)
         callback(0.7, "Visual model detected. Attempting to enhance figure extraction...")

download_deps.py

@@ -23,8 +23,8 @@ def get_urls(use_china_mirrors=False) -> list[Union[str, list[str]]]:
         return [
             "http://mirrors.tuna.tsinghua.edu.cn/ubuntu/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2_amd64.deb",
             "http://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2_arm64.deb",
-            "https://repo.huaweicloud.com/repository/maven/org/apache/tika/tika-server-standard/3.0.0/tika-server-standard-3.0.0.jar",
-            "https://repo.huaweicloud.com/repository/maven/org/apache/tika/tika-server-standard/3.0.0/tika-server-standard-3.0.0.jar.md5",
+            "https://repo.huaweicloud.com/repository/maven/org/apache/tika/tika-server-standard/3.2.3/tika-server-standard-3.2.3.jar",
+            "https://repo.huaweicloud.com/repository/maven/org/apache/tika/tika-server-standard/3.2.3/tika-server-standard-3.2.3.jar.md5",
             "https://openaipublic.blob.core.windows.net/encodings/cl100k_base.tiktoken",
             ["https://registry.npmmirror.com/-/binary/chrome-for-testing/121.0.6167.85/linux64/chrome-linux64.zip", "chrome-linux64-121-0-6167-85"],
             ["https://registry.npmmirror.com/-/binary/chrome-for-testing/121.0.6167.85/linux64/chromedriver-linux64.zip", "chromedriver-linux64-121-0-6167-85"],
@@ -34,8 +34,8 @@ def get_urls(use_china_mirrors=False) -> list[Union[str, list[str]]]:
         return [
             "http://archive.ubuntu.com/ubuntu/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2_amd64.deb",
             "http://ports.ubuntu.com/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2_arm64.deb",
-            "https://repo1.maven.org/maven2/org/apache/tika/tika-server-standard/3.0.0/tika-server-standard-3.0.0.jar",
-            "https://repo1.maven.org/maven2/org/apache/tika/tika-server-standard/3.0.0/tika-server-standard-3.0.0.jar.md5",
+            "https://repo1.maven.org/maven2/org/apache/tika/tika-server-standard/3.2.3/tika-server-standard-3.2.3.jar",
+            "https://repo1.maven.org/maven2/org/apache/tika/tika-server-standard/3.2.3/tika-server-standard-3.2.3.jar.md5",
             "https://openaipublic.blob.core.windows.net/encodings/cl100k_base.tiktoken",
             ["https://storage.googleapis.com/chrome-for-testing-public/121.0.6167.85/linux64/chrome-linux64.zip", "chrome-linux64-121-0-6167-85"],
             ["https://storage.googleapis.com/chrome-for-testing-public/121.0.6167.85/linux64/chromedriver-linux64.zip", "chromedriver-linux64-121-0-6167-85"],

rag/app/book.py

@@ -91,7 +91,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
             filename, binary=binary, from_page=from_page, to_page=to_page)
         remove_contents_table(sections, eng=is_english(
             random_choices([t for t, _ in sections], k=200)))
-        tbls=vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
         # tbls = [((None, lns), None) for lns in tbls]
         sections=[(item[0],item[1] if item[1] is not None else "") for item in sections if not isinstance(item[1], Image.Image)]
         callback(0.8, "Finish parsing.")
@@ -147,9 +147,16 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
 
     elif re.search(r"\.doc$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
-        with BytesIO(binary) as binary:
-            binary = BytesIO(binary)
-            doc_parsed = parser.from_buffer(binary)
+        try:
+            from tika import parser as tika_parser
+        except Exception as e:
+            callback(0.8, f"tika not available: {e}. Unsupported .doc parsing.")
+            logging.warning(f"tika not available: {e}. Unsupported .doc parsing for {filename}.")
+            return []
+
+        binary = BytesIO(binary)
+        doc_parsed = tika_parser.from_buffer(binary)
+        if doc_parsed.get('content', None) is not None:
             sections = doc_parsed['content'].split('\n')
             sections = [(line, "") for line in sections if line]
             remove_contents_table(sections, eng=is_english(

rag/app/manual.py

@@ -312,7 +312,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
             tk_cnt = num_tokens_from_string(txt)
             if sec_id > -1:
                 last_sid = sec_id
-        tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
         res = tokenize_table(tbls, doc, eng)
         res.extend(tokenize_chunks(chunks, doc, eng, pdf_parser))
         table_ctx = max(0, int(parser_config.get("table_context_size", 0) or 0))
@@ -325,7 +325,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
         docx_parser = Docx()
         ti_list, tbls = docx_parser(filename, binary,
                                     from_page=0, to_page=10000, callback=callback)
-        tbls=vision_figure_parser_docx_wrapper(sections=ti_list,tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_docx_wrapper(sections=ti_list,tbls=tbls,callback=callback,**kwargs)
         res = tokenize_table(tbls, doc, eng)
         for text, image in ti_list:
             d = copy.deepcopy(doc)

rag/app/one.py

@@ -76,7 +76,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
     if re.search(r"\.docx$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
         sections, tbls = naive.Docx()(filename, binary)
-        tbls=vision_figure_parser_docx_wrapper(sections=sections,tbls=tbls,callback=callback,**kwargs)
+        tbls = vision_figure_parser_docx_wrapper(sections=sections, tbls=tbls, callback=callback, **kwargs)
         sections = [s for s, _ in sections if s]
         for (_, html), _ in tbls:
             sections.append(html)
@@ -142,10 +142,18 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
 
     elif re.search(r"\.doc$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
+        try:
+            from tika import parser as tika_parser
+        except Exception as e:
+            callback(0.8, f"tika not available: {e}. Unsupported .doc parsing.")
+            logging.warning(f"tika not available: {e}. Unsupported .doc parsing for {filename}.")
+            return []
+
         binary = BytesIO(binary)
-        doc_parsed = parser.from_buffer(binary)
-        sections = doc_parsed['content'].split('\n')
-        sections = [s for s in sections if s]
+        doc_parsed = tika_parser.from_buffer(binary)
+        if doc_parsed.get('content', None) is not None:
+            sections = doc_parsed['content'].split('\n')
+            sections = [s for s in sections if s]
         callback(0.8, "Finish parsing.")
 
     else:

rag/flow/parser/parser.py

@@ -650,7 +650,7 @@ class Parser(ProcessBase):
             tmpf.flush()
             tmp_path = os.path.abspath(tmpf.name)
 
-            seq2txt_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.SPEECH2TEXT)
+            seq2txt_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.SPEECH2TEXT, llm_name=conf["llm_id"])
             txt = seq2txt_mdl.transcription(tmp_path)
 
             self.set_output("text", txt)

rag/prompts/generator.py

@@ -324,8 +324,9 @@ def tool_schema(tools_description: list[dict], complete_task=False):
                 }
             }
         }
-    for tool in tools_description:
-        desc[tool["function"]["name"]] = tool
+    for idx, tool in enumerate(tools_description):
+        name = tool["function"]["name"]
+        desc[f"{name}_{idx}"] = tool
 
     return "\n\n".join([f"## {i+1}. {fnm}\n{json.dumps(des, ensure_ascii=False, indent=4)}" for i, (fnm, des) in enumerate(desc.items())])
 

rag/svr/task_executor.py

@@ -374,7 +374,7 @@ async def build_chunks(task, progress_callback):
         chat_mdl = LLMBundle(task["tenant_id"], LLMType.CHAT, llm_name=task["llm_id"], lang=task["language"])
 
         async def gen_metadata_task(chat_mdl, d):
-            cached = get_llm_cache(chat_mdl.llm_name, d["content_with_weight"], "metadata")
+            cached = get_llm_cache(chat_mdl.llm_name, d["content_with_weight"], "metadata", {})
             if not cached:
                 async with chat_limiter:
                     cached = await gen_metadata(chat_mdl,
@@ -852,7 +852,7 @@ async def do_handle_task(task):
     task_tenant_id = task["tenant_id"]
     task_embedding_id = task["embd_id"]
     task_language = task["language"]
-    task_llm_id = task["llm_id"]
+    task_llm_id = task["parser_config"].get("llm_id") or task["llm_id"]
     task_dataset_id = task["kb_id"]
     task_doc_id = task["doc_id"]
     task_document_name = task["name"]
@@ -1024,33 +1024,65 @@ async def do_handle_task(task):
 
     chunk_count = len(set([chunk["id"] for chunk in chunks]))
     start_ts = timer()
-    e = await insert_es(task_id, task_tenant_id, task_dataset_id, chunks, progress_callback)
-    if not e:
-        return
 
-    logging.info("Indexing doc({}), page({}-{}), chunks({}), elapsed: {:.2f}".format(task_document_name, task_from_page,
-                                                                                     task_to_page, len(chunks),
-                                                                                     timer() - start_ts))
+    async def _maybe_insert_es(_chunks):
+        if has_canceled(task_id):
+            return True
+        e = await insert_es(task_id, task_tenant_id, task_dataset_id, _chunks, progress_callback)
+        return bool(e)
+    
+    try:
+        if not await _maybe_insert_es(chunks):
+            return
 
-    DocumentService.increment_chunk_num(task_doc_id, task_dataset_id, token_count, chunk_count, 0)
+        logging.info(
+            "Indexing doc({}), page({}-{}), chunks({}), elapsed: {:.2f}".format(
+                task_document_name, task_from_page, task_to_page, len(chunks), timer() - start_ts
+            )
+        )
 
-    time_cost = timer() - start_ts
-    progress_callback(msg="Indexing done ({:.2f}s).".format(time_cost))
-    if toc_thread:
-        d = toc_thread.result()
-        if d:
-            e = await insert_es(task_id, task_tenant_id, task_dataset_id, [d], progress_callback)
-            if not e:
-                return
-            DocumentService.increment_chunk_num(task_doc_id, task_dataset_id, 0, 1, 0)
+        DocumentService.increment_chunk_num(task_doc_id, task_dataset_id, token_count, chunk_count, 0)
 
-    task_time_cost = timer() - task_start_ts
-    progress_callback(prog=1.0, msg="Task done ({:.2f}s)".format(task_time_cost))
-    logging.info(
-        "Chunk doc({}), page({}-{}), chunks({}), token({}), elapsed:{:.2f}".format(task_document_name, task_from_page,
-                                                                                   task_to_page, len(chunks),
-                                                                                   token_count, task_time_cost))
+        progress_callback(msg="Indexing done ({:.2f}s).".format(timer() - start_ts))
 
+        if toc_thread:
+            d = toc_thread.result()
+            if d:
+                if not await _maybe_insert_es([d]):
+                    return
+                DocumentService.increment_chunk_num(task_doc_id, task_dataset_id, 0, 1, 0)
+
+        if has_canceled(task_id):
+            progress_callback(-1, msg="Task has been canceled.")
+            return
+
+        task_time_cost = timer() - task_start_ts
+        progress_callback(prog=1.0, msg="Task done ({:.2f}s)".format(task_time_cost))
+        logging.info(
+            "Chunk doc({}), page({}-{}), chunks({}), token({}), elapsed:{:.2f}".format(
+                task_document_name, task_from_page, task_to_page, len(chunks), token_count, task_time_cost
+            )
+        )
+
+    finally:
+        if has_canceled(task_id):
+            try:
+                exists = await asyncio.to_thread(
+                    settings.docStoreConn.indexExist,
+                    search.index_name(task_tenant_id),
+                    task_dataset_id,
+                )
+                if exists:
+                    await asyncio.to_thread(
+                        settings.docStoreConn.delete,
+                        {"doc_id": task_doc_id},
+                        search.index_name(task_tenant_id),
+                        task_dataset_id,
+                    )
+            except Exception:
+                logging.exception(
+                    f"Remove doc({task_doc_id}) from docStore failed when task({task_id}) canceled."
+                )
 
 async def handle_task():
 

web/src/assets/svg/home-icon/memory-bri.svg

@@ -0,0 +1,10 @@
+<svg width="24" height="24" viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg">
+<path d="M22 12H2M22 12V18C22 18.5304 21.7893 19.0391 21.4142 19.4142C21.0391 19.7893 20.5304 20 20 20H4C3.46957 20 2.96086 19.7893 2.58579 19.4142C2.21071 19.0391 2 18.5304 2 18V12M22 12L18.55 5.11C18.3844 4.77679 18.1292 4.49637 17.813 4.30028C17.4967 4.10419 17.1321 4.0002 16.76 4H7.24C6.86792 4.0002 6.50326 4.10419 6.18704 4.30028C5.87083 4.49637 5.61558 4.77679 5.45 5.11L2 12" stroke="url(#paint0_linear_1415_84974)" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/>
+<path d="M6 16H6.01M10 16H10.01" stroke="#00BEB4" stroke-width="1.5" stroke-linecap="round" stroke-linejoin="round"/>
+<defs>
+<linearGradient id="paint0_linear_1415_84974" x1="12.5556" y1="4" x2="12.5556" y2="20" gradientUnits="userSpaceOnUse">
+<stop stop-color="#161618"/>
+<stop offset="1" stop-color="#666666"/>
+</linearGradient>
+</defs>
+</svg>

web/src/components/list-filter-bar/filter-field.tsx

@@ -118,7 +118,7 @@ export const FilterField = memo(
                     setShowAll(!showAll);
                   }}
                 >
-                  <FormLabel className="text-text-primary">
+                  <FormLabel className="text-text-secondary text-sm">
                     {item.label}
                   </FormLabel>
                   {showAll ? (

web/src/components/list-filter-bar/filter-popover.tsx

@@ -33,6 +33,7 @@ export type CheckboxFormMultipleProps = {
   onChange?: FilterChange;
   onOpenChange?: (open: boolean) => void;
   setOpen(open: boolean): void;
+  filterGroup?: Record<string, string[]>;
 };
 
 function CheckboxFormMultiple({
@@ -40,6 +41,7 @@ function CheckboxFormMultiple({
   value,
   onChange,
   setOpen,
+  filterGroup,
 }: CheckboxFormMultipleProps) {
   const [resolvedFilters, setResolvedFilters] =
     useState<FilterCollection[]>(filters);
@@ -120,6 +122,20 @@ function CheckboxFormMultiple({
     }
   }, [form, value, resolvedFilters, fieldsDict]);
 
+  const filterList = useMemo(() => {
+    const filterSet = filterGroup
+      ? Object.values(filterGroup).reduce<Set<string>>((pre, cur) => {
+          cur.forEach((item) => pre.add(item));
+          return pre;
+        }, new Set())
+      : new Set();
+    return [...filterSet];
+  }, [filterGroup]);
+
+  const notInfilterGroup = useMemo(() => {
+    return filters.filter((x) => !filterList.includes(x.field));
+  }, [filterList, filters]);
+
   return (
     <Form {...form}>
       <form
@@ -127,34 +143,70 @@ function CheckboxFormMultiple({
         className="space-y-8 px-5 py-2.5"
         onReset={() => form.reset()}
       >
-        {filters.map((x) => (
-          <FormField
-            key={x.field}
-            control={form.control}
-            name={x.field}
-            render={() => (
-              <FormItem className="space-y-4">
-                <div>
-                  <FormLabel className="text-text-primary">{x.label}</FormLabel>
+        <div className="space-y-4">
+          {filterGroup &&
+            Object.keys(filterGroup).map((key) => {
+              const filterKeys = filterGroup[key];
+              const thisFilters = filters.filter((x) =>
+                filterKeys.includes(x.field),
+              );
+              return (
+                <div
+                  key={key}
+                  className="flex flex-col space-y-4 border-b border-border-button pb-4"
+                >
+                  <div className="text-text-primary text-sm">{key}</div>
+                  <div className="flex flex-col space-y-4">
+                    {thisFilters.map((x) => (
+                      <FilterField
+                        key={x.field}
+                        item={{ ...x, id: x.field }}
+                        parent={{
+                          ...x,
+                          id: x.field,
+                          field: ``,
+                        }}
+                      />
+                    ))}
+                  </div>
                 </div>
-                {x.list.map((item) => {
-                  return (
-                    <FilterField
-                      key={item.id}
-                      item={{ ...item }}
-                      parent={{
-                        ...x,
-                        id: x.field,
-                        // field: `${x.field}${item.field ? '.' + item.field : ''}`,
-                      }}
-                    />
-                  );
-                })}
-                <FormMessage />
-              </FormItem>
-            )}
-          />
-        ))}
+              );
+            })}
+          {notInfilterGroup &&
+            notInfilterGroup.map((x) => {
+              return (
+                <FormField
+                  key={x.field}
+                  control={form.control}
+                  name={x.field}
+                  render={() => (
+                    <FormItem className="space-y-4">
+                      <div>
+                        <FormLabel className="text-text-primary text-sm">
+                          {x.label}
+                        </FormLabel>
+                      </div>
+                      {x.list.map((item) => {
+                        return (
+                          <FilterField
+                            key={item.id}
+                            item={{ ...item }}
+                            parent={{
+                              ...x,
+                              id: x.field,
+                              // field: `${x.field}${item.field ? '.' + item.field : ''}`,
+                            }}
+                          />
+                        );
+                      })}
+                      <FormMessage />
+                    </FormItem>
+                  )}
+                />
+              );
+            })}
+        </div>
+
         <div className="flex justify-end gap-5">
           <Button
             type="button"
@@ -185,6 +237,7 @@ export function FilterPopover({
   onChange,
   onOpenChange,
   filters,
+  filterGroup,
 }: PropsWithChildren & Omit<CheckboxFormMultipleProps, 'setOpen'>) {
   const [open, setOpen] = useState(false);
   const onOpenChangeFun = useCallback(
@@ -203,6 +256,7 @@ export function FilterPopover({
           value={value}
           filters={filters}
           setOpen={setOpen}
+          filterGroup={filterGroup}
         ></CheckboxFormMultiple>
       </PopoverContent>
     </Popover>

web/src/components/list-filter-bar/index.tsx

@@ -58,9 +58,11 @@ export default function ListFilterBar({
   filters,
   className,
   icon,
+  filterGroup,
 }: PropsWithChildren<IProps & Omit<CheckboxFormMultipleProps, 'setOpen'>> & {
   className?: string;
   icon?: ReactNode;
+  filterGroup?: Record<string, string[]>;
 }) {
   const filterCount = useMemo(() => {
     return typeof value === 'object' && value !== null
@@ -99,6 +101,7 @@ export default function ListFilterBar({
             value={value}
             onChange={onChange}
             filters={filters}
+            filterGroup={filterGroup}
             onOpenChange={onOpenChange}
           >
             <FilterButton count={filterCount}></FilterButton>

web/src/components/memories-form-field.tsx

@@ -0,0 +1,33 @@
+import { useFetchAllMemoryList } from '@/hooks/use-memory-request';
+import { useTranslation } from 'react-i18next';
+import { RAGFlowFormItem } from './ragflow-form';
+import { MultiSelect } from './ui/multi-select';
+
+type MemoriesFormFieldProps = {
+  label: string;
+};
+
+export function MemoriesFormField({ label }: MemoriesFormFieldProps) {
+  const { t } = useTranslation();
+  const memoryList = useFetchAllMemoryList();
+
+  const options = memoryList.data?.map((memory) => ({
+    label: memory.name,
+    value: memory.id,
+  }));
+
+  return (
+    <RAGFlowFormItem name="memory_ids" label={label}>
+      {(field) => (
+        <MultiSelect
+          options={options || []}
+          placeholder={t('common.pleaseSelect')}
+          maxCount={100}
+          onValueChange={field.onChange}
+          defaultValue={field.value}
+          modalPopover
+        />
+      )}
+    </RAGFlowFormItem>
+  );
+}

web/src/components/parse-configuration/graph-rag-form-fields.tsx

@@ -1,6 +1,7 @@
 import { DocumentParserType } from '@/constants/knowledge';
 import { useTranslate } from '@/hooks/common-hooks';
 import { cn } from '@/lib/utils';
+import { LLMModelItem } from '@/pages/dataset/dataset-setting/configuration/common-item';
 import {
   GenerateLogButton,
   GenerateType,
@@ -136,6 +137,10 @@ const GraphRagItems = ({
 
   return (
     <FormContainer className={cn({ 'mb-4': marginBottom }, className)}>
+      <LLMModelItem
+        label={t('globalIndexModel')}
+        name={'parser_config.llm_id'}
+      />
       <UseGraphRagFormField
         data={data}
         onDelete={onDelete}

web/src/hooks/use-document-request.ts

@@ -122,9 +122,9 @@ export const useFetchDocumentList = () => {
           page: pagination.current,
         },
         {
-          suffix: filterValue.type,
-          run_status: filterValue.run,
-          metadata: filterValue.metadata,
+          suffix: filterValue.type as string[],
+          run_status: filterValue.run as string[],
+          metadata: filterValue.metadata as Record<string, string[]>,
         },
       );
       if (ret.data.code === 0) {

web/src/hooks/use-memory-request.ts

@@ -0,0 +1,30 @@
+import { IMemory } from '@/interfaces/database/memory';
+import memoryService from '@/services/memory-service';
+import { useQuery } from '@tanstack/react-query';
+
+export const enum MemoryApiAction {
+  FetchMemoryList = 'fetchMemoryList',
+}
+
+export const useFetchAllMemoryList = () => {
+  const { data, isLoading, isError, refetch } = useQuery<IMemory[], Error>({
+    queryKey: [MemoryApiAction.FetchMemoryList],
+    queryFn: async () => {
+      const { data: response } = await memoryService.getMemoryList(
+        {
+          params: { page_size: 100000000, page: 1 },
+          data: {},
+        },
+        true,
+      );
+      return response.data.memory_list ?? [];
+    },
+  });
+
+  return {
+    data,
+    isLoading,
+    isError,
+    refetch,
+  };
+};

web/src/interfaces/database/memory.ts

@@ -0,0 +1,11 @@
+export interface IMemory {
+  avatar: null;
+  description: null;
+  id: string;
+  memory_type: string[];
+  name: string;
+  owner_name: string;
+  permissions: string;
+  storage_type: string;
+  tenant_id: string;
+}

web/src/interfaces/request/knowledge.ts

@@ -33,4 +33,5 @@ export interface IFetchKnowledgeListRequestParams {
 export interface IFetchDocumentListRequestBody {
   suffix?: string[];
   run_status?: string[];
+  metadata?: Record<string, string[]>;
 }

web/src/locales/en.ts

@@ -176,6 +176,7 @@ Procedural Memory: Learned skills, habits, and automated procedures.`,
     },
     knowledgeDetails: {
       metadata: {
+        fieldSetting: 'Field settings',
         changesAffectNewParses: 'Changes affect new parses only.',
         editMetadataForDataset: 'View and edit metadata for ',
         restrictDefinedValues: 'Restrict to defined values',
@@ -190,6 +191,8 @@ Procedural Memory: Learned skills, habits, and automated procedures.`,
         fieldName: 'Field name',
         editMetadata: 'Edit metadata',
       },
+      metadataField: 'Metadata field',
+      systemAttribute: 'System attribute',
       localUpload: 'Local upload',
       fileSize: 'File size',
       fileType: 'File type',
@@ -363,7 +366,11 @@ Procedural Memory: Learned skills, habits, and automated procedures.`,
       reRankModelWaring: 'Re-rank model is very time consuming.',
     },
     knowledgeConfiguration: {
+      globalIndexModelTip:
+        'Used to generate Knowledge graphs, RAPTOR, auto-metadata, auto-keyword and auto-question. Model performance will affects generation quality.',
+      globalIndexModel: 'Indexing model',
       settings: 'Settings',
+      autoMetadataTip: `Automatically generate metadata. Applies to new files during parsing. Existing files require re-parsing to update (chunks remain preserved). Be aware that extra tokens will be consumed by the indexing model specified in 'Configuration'.`,
       imageContextWindow: 'Image context window',
       imageContextWindowTip:
         'Captures N tokens of text above and below the image to provide richer background context for the image chunk.',
@@ -2120,9 +2127,11 @@ Important structured information may include: names, dates, locations, events, k
         queryParameters: 'Query parameters',
         headerParameters: 'Header parameters',
         requestBodyParameters: 'Request body parameters',
-        streaming: 'Accepted response',
-        immediately: 'Final response',
+        immediately: 'Accepted response',
+        streaming: 'Final response',
       },
+      saveToMemory: 'Save to memory',
+      memory: 'Memory',
     },
     llmTools: {
       bad_calculator: {

web/src/locales/zh.ts

@@ -93,6 +93,67 @@ export default {
       search: '搜索',
       welcome: '欢迎来到',
       dataset: '知识库',
+      Memories: '记忆',
+    },
+    memories: {
+      llmTooltip: '分析对话内容，提取关键信息，并生成结构化的记忆摘要。',
+      embeddingModelTooltip:
+        '将文本转换为数值向量，用于语义相似度搜索和记忆检索。',
+      embeddingModelError: '记忆类型为必填项，且"原始"类型不可删除。',
+      memoryTypeTooltip: `原始: 用户与代理之间的原始对话内容（默认必需）。
+语义记忆: 关于用户和世界的通用知识和事实。
+情景记忆: 带时间戳的特定事件和经历记录。
+程序记忆: 学习的技能、习惯和自动化程序。`,
+      editName: '编辑名称',
+      memory: '记忆',
+      createMemory: '创建记忆',
+      name: '名称',
+      memoryNamePlaceholder: '记忆名称',
+      memoryType: '记忆类型',
+      embeddingModel: '嵌入模型',
+      selectModel: '选择模型',
+      llm: '大语言模型',
+      delMemoryWarn: `删除后，此记忆中的所有消息都将被删除，代理将无法检索。`,
+    },
+    memory: {
+      messages: {
+        messageDescription:
+          '记忆检索已在高级设置中配置相似度阈值、关键词相似度权重和前N个结果。',
+        copied: '已复制！',
+        contentEmbed: '内容嵌入',
+        content: '内容',
+        delMessageWarn: `遗忘后，代理将无法检索此消息。`,
+        forgetMessage: '遗忘消息',
+        sessionId: '会话ID',
+        agent: '代理',
+        type: '类型',
+        validDate: '有效日期',
+        forgetAt: '遗忘于',
+        source: '来源',
+        enable: '启用',
+        action: '操作',
+      },
+      config: {
+        avatar: '头像',
+        description: '描述',
+        memorySize: '记忆大小',
+        advancedSettings: '高级设置',
+        permission: '权限',
+        onlyMe: '仅自己',
+        team: '团队',
+        storageType: '存储类型',
+        storageTypePlaceholder: '请选择存储类型',
+        forgetPolicy: '遗忘策略',
+        temperature: '温度',
+        systemPrompt: '系统提示词',
+        systemPromptPlaceholder: '请输入系统提示词',
+        userPrompt: '用户提示词',
+        userPromptPlaceholder: '请输入用户提示词',
+      },
+      sideBar: {
+        messages: '消息',
+        configuration: '配置',
+      },
     },
     knowledgeList: {
       welcome: '欢迎回来',
@@ -106,6 +167,22 @@ export default {
       parserRequired: '分块方法必填',
     },
     knowledgeDetails: {
+      metadata: {
+        fieldSetting: '字段设置',
+        changesAffectNewParses: '更改仅影响新解析。',
+        editMetadataForDataset: '查看和编辑元数据于 ',
+        restrictDefinedValues: '限制为已定义值',
+        metadataGenerationSettings: '元数据生成设置',
+        manageMetadataForDataset: '管理此数据集的元数据',
+        manageMetadata: '管理元数据',
+        metadata: '元数据',
+        values: '值',
+        action: '操作',
+        field: '字段',
+        description: '描述',
+        fieldName: '字段名',
+        editMetadata: '编辑元数据',
+      },
       localUpload: '本地上传',
       fileSize: '文件大小',
       fileType: '文件类型',
@@ -1863,6 +1940,8 @@ Tokenizer 会根据所选方式将内容存储为对应的数据结构。`,
         headerParameters: '请求头参数',
         requestBodyParameters: '请求体参数',
       },
+      saveToMemory: '保存到Memory',
+      memory: 'Memory',
     },
     footer: {
       profile: 'All rights reserved @ React',

web/src/pages/agent/chat/use-send-agent-message.ts

@@ -7,7 +7,6 @@ import {
 import {
   IAttachment,
   IEventList,
-  IInputEvent,
   IMessageEndData,
   IMessageEndEvent,
   IMessageEvent,
@@ -93,7 +92,7 @@ export function findMessageFromList(eventList: IEventList) {
 export function findInputFromList(eventList: IEventList) {
   const inputEvent = eventList.find(
     (x) => x.event === MessageEventType.UserInputs,
-  ) as IInputEvent;
+  );
 
   if (!inputEvent) {
     return {};

web/src/pages/agent/form/message-form/index.tsx

@@ -1,4 +1,4 @@
-import { FormContainer } from '@/components/form-container';
+import { MemoriesFormField } from '@/components/memories-form-field';
 import { BlockButton, Button } from '@/components/ui/button';
 import {
   Form,
@@ -41,6 +41,7 @@ function MessageForm({ node }: INextOperatorForm) {
     output_format: z.string().optional(),
     auto_play: z.boolean().optional(),
     status: z.number().optional(),
+    memory_ids: z.array(z.string()).optional(),
   });
 
   const form = useForm({
@@ -67,99 +68,100 @@ function MessageForm({ node }: INextOperatorForm) {
         {showWebhookResponseStatus && (
           <WebHookResponseStatusFormField name="status"></WebHookResponseStatusFormField>
         )}
-        <FormContainer>
-          <FormItem>
-            <FormLabel tooltip={t('flow.msgTip')}>{t('flow.msg')}</FormLabel>
-            <div className="space-y-4">
-              {fields.map((field, index) => (
-                <div key={field.id} className="flex items-start gap-2">
-                  <FormField
-                    control={form.control}
-                    name={`content.${index}.value`}
-                    render={({ field }) => (
-                      <FormItem className="flex-1">
-                        <FormControl>
-                          <PromptEditor
-                            {...field}
-                            placeholder={t('flow.messagePlaceholder')}
-                          ></PromptEditor>
-                        </FormControl>
-                      </FormItem>
-                    )}
-                  />
-                  {fields.length > 1 && (
-                    <Button
-                      type="button"
-                      variant={'ghost'}
-                      onClick={() => remove(index)}
-                    >
-                      <X />
-                    </Button>
+        <FormItem>
+          <FormLabel tooltip={t('flow.msgTip')}>{t('flow.msg')}</FormLabel>
+          <div className="space-y-4">
+            {fields.map((field, index) => (
+              <div key={field.id} className="flex items-start gap-2">
+                <FormField
+                  control={form.control}
+                  name={`content.${index}.value`}
+                  render={({ field }) => (
+                    <FormItem className="flex-1">
+                      <FormControl>
+                        <PromptEditor
+                          {...field}
+                          placeholder={t('flow.messagePlaceholder')}
+                        ></PromptEditor>
+                      </FormControl>
+                    </FormItem>
                   )}
-                </div>
-              ))}
+                />
+                {fields.length > 1 && (
+                  <Button
+                    type="button"
+                    variant={'ghost'}
+                    onClick={() => remove(index)}
+                  >
+                    <X />
+                  </Button>
+                )}
+              </div>
+            ))}
 
-              <BlockButton
-                type="button"
-                onClick={() => append({ value: '' })} // "" will cause the inability to add, refer to: https://github.com/orgs/react-hook-form/discussions/8485#discussioncomment-2961861
-              >
-                {t('flow.addMessage')}
-              </BlockButton>
-            </div>
-            <FormMessage />
-          </FormItem>
-        </FormContainer>
-        <FormContainer>
-          <FormItem>
-            <FormLabel tooltip={t('flow.downloadFileTypeTip')}>
-              {t('flow.downloadFileType')}
-            </FormLabel>
-            <FormField
-              control={form.control}
-              name={`output_format`}
-              render={({ field }) => (
-                <FormItem className="flex-1">
-                  <FormControl>
-                    <RAGFlowSelect
-                      options={Object.keys(ExportFileType).map(
-                        (key: string) => {
-                          return {
-                            value:
-                              ExportFileType[
-                                key as keyof typeof ExportFileType
-                              ],
-                            label: key,
-                          };
-                        },
-                      )}
-                      {...field}
-                      onValueChange={field.onChange}
-                      placeholder={t('common.selectPlaceholder')}
-                      allowClear
-                    ></RAGFlowSelect>
-                  </FormControl>
-                </FormItem>
-              )}
-            />
-          </FormItem>
-          <FormItem>
-            <FormLabel>{t('flow.autoPlay')}</FormLabel>
-            <FormField
-              control={form.control}
-              name={`auto_play`}
-              render={({ field }) => (
-                <FormItem className="flex-1">
-                  <FormControl>
-                    <Switch
-                      checked={field.value}
-                      onCheckedChange={field.onChange}
-                    />
-                  </FormControl>
-                </FormItem>
-              )}
-            />
-          </FormItem>
-        </FormContainer>
+            <BlockButton
+              type="button"
+              onClick={() => append({ value: '' })} // "" will cause the inability to add, refer to: https://github.com/orgs/react-hook-form/discussions/8485#discussioncomment-2961861
+            >
+              {t('flow.addMessage')}
+            </BlockButton>
+          </div>
+          <FormMessage />
+        </FormItem>
+        {!showWebhookResponseStatus && (
+          <>
+            <FormItem>
+              <FormLabel tooltip={t('flow.downloadFileTypeTip')}>
+                {t('flow.downloadFileType')}
+              </FormLabel>
+              <FormField
+                control={form.control}
+                name={`output_format`}
+                render={({ field }) => (
+                  <FormItem className="flex-1">
+                    <FormControl>
+                      <RAGFlowSelect
+                        options={Object.keys(ExportFileType).map(
+                          (key: string) => {
+                            return {
+                              value:
+                                ExportFileType[
+                                  key as keyof typeof ExportFileType
+                                ],
+                              label: key,
+                            };
+                          },
+                        )}
+                        {...field}
+                        onValueChange={field.onChange}
+                        placeholder={t('common.selectPlaceholder')}
+                        allowClear
+                      ></RAGFlowSelect>
+                    </FormControl>
+                  </FormItem>
+                )}
+              />
+            </FormItem>
+            <FormItem>
+              <FormLabel>{t('flow.autoPlay')}</FormLabel>
+              <FormField
+                control={form.control}
+                name={`auto_play`}
+                render={({ field }) => (
+                  <FormItem className="flex-1">
+                    <FormControl>
+                      <Switch
+                        checked={field.value}
+                        onCheckedChange={field.onChange}
+                      />
+                    </FormControl>
+                  </FormItem>
+                )}
+              />
+            </FormItem>
+          </>
+        )}
+        <MemoriesFormField label={t('flow.saveToMemory')}></MemoriesFormField>
       </FormWrapper>
     </Form>
   );

web/src/pages/agent/form/message-form/use-show-response-status.ts

@@ -15,16 +15,16 @@ export function useShowWebhookResponseStatus(form: UseFormReturn<any>) {
   const showWebhookResponseStatus = useMemo(() => {
     const formData: BeginFormSchemaType = getNode(BeginId)?.data.form;
     return (
-      formData.mode === AgentDialogueMode.Webhook &&
+      formData?.mode === AgentDialogueMode.Webhook &&
       formData.execution_mode === WebhookExecutionMode.Streaming
     );
-  }, []);
+  }, [getNode]);
 
   useEffect(() => {
     if (showWebhookResponseStatus && isEmpty(form.getValues('status'))) {
       form.setValue('status', 200, { shouldValidate: true, shouldDirty: true });
     }
-  }, []);
+  }, [form, showWebhookResponseStatus]);
 
   return showWebhookResponseStatus;
 }

web/src/pages/agent/form/retrieval-form/next.tsx

@@ -2,6 +2,7 @@ import { Collapse } from '@/components/collapse';
 import { CrossLanguageFormField } from '@/components/cross-language-form-field';
 import { FormContainer } from '@/components/form-container';
 import { KnowledgeBaseFormField } from '@/components/knowledge-base-item';
+import { MemoriesFormField } from '@/components/memories-form-field';
 import {
   MetadataFilter,
   MetadataFilterSchema,
@@ -46,6 +47,7 @@ export const RetrievalPartialSchema = {
   use_kg: z.boolean(),
   toc_enhance: z.boolean(),
   ...MetadataFilterSchema,
+  memory_ids: z.array(z.string()).optional(),
 };
 
 export const FormSchema = z.object({
@@ -109,12 +111,11 @@ function RetrievalForm({ node }: INextOperatorForm) {
   return (
     <Form {...form}>
       <FormWrapper>
-        <FormContainer>
-          <RAGFlowFormItem name="query" label={t('flow.query')}>
-            <PromptEditor></PromptEditor>
-          </RAGFlowFormItem>
-          <KnowledgeBaseFormField showVariable></KnowledgeBaseFormField>
-        </FormContainer>
+        <RAGFlowFormItem name="query" label={t('flow.query')}>
+          <PromptEditor></PromptEditor>
+        </RAGFlowFormItem>
+        <KnowledgeBaseFormField showVariable></KnowledgeBaseFormField>
+        <MemoriesFormField label={t('flow.memory')}></MemoriesFormField>
         <Collapse title={<div>{t('flow.advancedSettings')}</div>}>
           <FormContainer>
             <SimilaritySliderFormField

web/src/pages/agent/form/tool-form/retrieval-form/index.tsx

@@ -2,6 +2,7 @@ import { Collapse } from '@/components/collapse';
 import { CrossLanguageFormField } from '@/components/cross-language-form-field';
 import { FormContainer } from '@/components/form-container';
 import { KnowledgeBaseFormField } from '@/components/knowledge-base-item';
+import { MemoriesFormField } from '@/components/memories-form-field';
 import { MetadataFilter } from '@/components/metadata-filter';
 import { RerankFormFields } from '@/components/rerank';
 import { SimilaritySliderFormField } from '@/components/similarity-slider';
@@ -40,10 +41,9 @@ const RetrievalForm = () => {
   return (
     <Form {...form}>
       <FormWrapper>
-        <FormContainer>
-          <DescriptionField></DescriptionField>
-          <KnowledgeBaseFormField showVariable></KnowledgeBaseFormField>
-        </FormContainer>
+        <DescriptionField></DescriptionField>
+        <KnowledgeBaseFormField showVariable></KnowledgeBaseFormField>
+        <MemoriesFormField label={t('flow.memory')}></MemoriesFormField>
         <Collapse title={<div>{t('flow.advancedSettings')}</div>}>
           <FormContainer>
             <SimilaritySliderFormField

web/src/pages/dataset/components/metedata/hook.ts

@@ -1,7 +1,7 @@
 import message from '@/components/ui/message';
 import { useSetModalState } from '@/hooks/common-hooks';
 import { useSetDocumentMeta } from '@/hooks/use-document-request';
-import {
+import kbService, {
   getMetaDataService,
   updateMetaData,
 } from '@/services/knowledge-service';
@@ -198,20 +198,24 @@ export const useManageMetaDataModal = (
   const { setDocumentMeta } = useSetDocumentMeta();
 
   useEffect(() => {
-    if (data) {
-      setTableData(data);
-    } else {
-      setTableData([]);
+    if (type === MetadataType.Manage) {
+      if (data) {
+        setTableData(data);
+      } else {
+        setTableData([]);
+      }
     }
-  }, [data]);
+  }, [data, type]);
 
   useEffect(() => {
-    if (metaData) {
-      setTableData(metaData);
-    } else {
-      setTableData([]);
+    if (type !== MetadataType.Manage) {
+      if (metaData) {
+        setTableData(metaData);
+      } else {
+        setTableData([]);
+      }
     }
-  }, [metaData]);
+  }, [metaData, type]);
 
   const handleDeleteSingleValue = useCallback(
     (field: string, value: string) => {
@@ -255,7 +259,7 @@ export const useManageMetaDataModal = (
         data: operations,
       });
       if (res.code === 0) {
-        message.success(t('message.success'));
+        message.success(t('message.operated'));
         callback();
       }
     },
@@ -282,11 +286,18 @@ export const useManageMetaDataModal = (
   const handleSaveSettings = useCallback(
     async (callback: () => void) => {
       const data = util.tableDataToMetaDataSettingJSON(tableData);
-      callback();
+      const { data: res } = await kbService.kbUpdateMetaData({
+        kb_id: id,
+        metadata: data,
+      });
+      if (res.code === 0) {
+        message.success(t('message.operated'));
+        callback?.();
+      }
 
       return data;
     },
-    [tableData],
+    [tableData, id],
   );
 
   const handleSave = useCallback(

web/src/pages/dataset/components/metedata/manage-modal.tsx

@@ -27,7 +27,7 @@ import {
 import { Plus, Settings, Trash2 } from 'lucide-react';
 import { useCallback, useMemo, useState } from 'react';
 import { useTranslation } from 'react-i18next';
-import { useManageMetaDataModal } from './hook';
+import { MetadataType, useManageMetaDataModal } from './hook';
 import { IManageModalProps, IMetaDataTableData } from './interface';
 import { ManageValuesModal } from './manage-values-modal';
 export const ManageMetadataModal = (props: IManageModalProps) => {
@@ -54,7 +54,6 @@ export const ManageMetadataModal = (props: IManageModalProps) => {
     values: [],
   });
 
-  const [currentValueIndex, setCurrentValueIndex] = useState<number>(0);
   const [deleteDialogContent, setDeleteDialogContent] = useState({
     visible: false,
     title: '',
@@ -95,12 +94,12 @@ export const ManageMetadataModal = (props: IManageModalProps) => {
       description: '',
       values: [],
     });
-    setCurrentValueIndex(tableData.length || 0);
+    // setCurrentValueIndex(tableData.length || 0);
     showManageValuesModal();
   };
   const handleEditValueRow = useCallback(
-    (data: IMetaDataTableData, index: number) => {
-      setCurrentValueIndex(index);
+    (data: IMetaDataTableData) => {
+      // setCurrentValueIndex(index);
       setValueData(data);
       showManageValuesModal();
     },
@@ -186,7 +185,7 @@ export const ManageMetadataModal = (props: IManageModalProps) => {
               variant={'ghost'}
               className="bg-transparent px-1 py-0"
               onClick={() => {
-                handleEditValueRow(row.original, row.index);
+                handleEditValueRow(row.original);
               }}
             >
               <Settings />
@@ -244,16 +243,32 @@ export const ManageMetadataModal = (props: IManageModalProps) => {
 
   const handleSaveValues = (data: IMetaDataTableData) => {
     setTableData((prev) => {
-      if (currentValueIndex >= prev.length) {
-        return [...prev, data];
+      //If the keys are the same, they need to be merged.
+      const fieldMap = new Map<string, any>();
+
+      prev.forEach((item) => {
+        if (fieldMap.has(item.field)) {
+          const existingItem = fieldMap.get(item.field);
+          const mergedValues = [
+            ...new Set([...existingItem.values, ...item.values]),
+          ];
+          fieldMap.set(item.field, { ...existingItem, values: mergedValues });
+        } else {
+          fieldMap.set(item.field, item);
+        }
+      });
+
+      if (fieldMap.has(data.field)) {
+        const existingItem = fieldMap.get(data.field);
+        const mergedValues = [
+          ...new Set([...existingItem.values, ...data.values]),
+        ];
+        fieldMap.set(data.field, { ...existingItem, values: mergedValues });
       } else {
-        return prev.map((item, index) => {
-          if (index === currentValueIndex) {
-            return data;
-          }
-          return item;
-        });
+        fieldMap.set(data.field, data);
       }
+
+      return Array.from(fieldMap.values());
     });
   };
 
@@ -335,7 +350,13 @@ export const ManageMetadataModal = (props: IManageModalProps) => {
       </Modal>
       {manageValuesVisible && (
         <ManageValuesModal
-          title={<div>{t('knowledgeDetails.metadata.editMetadata')}</div>}
+          title={
+            <div>
+              {metadataType === MetadataType.Setting
+                ? t('knowledgeDetails.metadata.fieldSetting')
+                : t('knowledgeDetails.metadata.editMetadata')}
+            </div>
+          }
           visible={manageValuesVisible}
           hideModal={hideManageValuesModal}
           data={valueData}

web/src/pages/dataset/components/metedata/manage-values-modal.tsx

@@ -110,8 +110,8 @@ export const ManageValuesModal = (props: IManageValuesProps) => {
 
   // Handle blur event, synchronize to main state
   const handleValueBlur = useCallback(() => {
-    addUpdateValue(metaData.field, [...tempValues]);
-    handleChange('values', [...tempValues]);
+    addUpdateValue(metaData.field, [...new Set([...tempValues])]);
+    handleChange('values', [...new Set([...tempValues])]);
   }, [handleChange, tempValues, metaData, addUpdateValue]);
 
   // Handle delete operation
@@ -139,12 +139,12 @@ export const ManageValuesModal = (props: IManageValuesProps) => {
 
   // Handle adding new value
   const handleAddValue = useCallback(() => {
-    setTempValues((prev) => [...prev, '']);
+    setTempValues((prev) => [...new Set([...prev, ''])]);
 
     // Synchronize to main state
     setMetaData((prev) => ({
       ...prev,
-      values: [...prev.values, ''],
+      values: [...new Set([...prev.values, ''])],
     }));
   }, []);
 
@@ -154,7 +154,7 @@ export const ManageValuesModal = (props: IManageValuesProps) => {
       open={visible}
       onCancel={handleHideModal}
       className="!w-[460px]"
-      okText={t('common.save')}
+      okText={t('common.confirm')}
       onOk={handleSave}
       maskClosable={false}
       footer={null}

web/src/pages/dataset/dataset-setting/configuration/audio.tsx

@@ -3,11 +3,13 @@ import {
   AutoQuestionsFormField,
 } from '@/components/auto-keywords-form-field';
 import { ConfigurationFormContainer } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function AudioConfiguration() {
   return (
     <ConfigurationFormContainer>
       <>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </>

web/src/pages/dataset/dataset-setting/configuration/book.tsx

@@ -7,6 +7,7 @@ import {
   ConfigurationFormContainer,
   MainContainer,
 } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function BookConfiguration() {
   return (
@@ -16,6 +17,7 @@ export function BookConfiguration() {
       </ConfigurationFormContainer>
 
       <ConfigurationFormContainer>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </ConfigurationFormContainer>

web/src/pages/dataset/dataset-setting/configuration/common-item.tsx

@@ -3,7 +3,10 @@ import {
   FormFieldType,
   RenderField,
 } from '@/components/dynamic-form';
-import { SelectWithSearch } from '@/components/originui/select-with-search';
+import {
+  SelectWithSearch,
+  SelectWithSearchFlagOptionType,
+} from '@/components/originui/select-with-search';
 import { SliderInputFormField } from '@/components/slider-input-form-field';
 import { Button } from '@/components/ui/button';
 import {
@@ -16,7 +19,9 @@ import {
 import { Radio } from '@/components/ui/radio';
 import { Spin } from '@/components/ui/spin';
 import { Switch } from '@/components/ui/switch';
+import { LlmModelType } from '@/constants/knowledge';
 import { useTranslate } from '@/hooks/common-hooks';
+import { useComposeLlmOptionsByModelTypes } from '@/hooks/use-llm-request';
 import { cn } from '@/lib/utils';
 import { t } from 'i18next';
 import { Settings } from 'lucide-react';
@@ -41,6 +46,8 @@ import {
 interface IProps {
   line?: 1 | 2;
   isEdit?: boolean;
+  label?: string;
+  name?: string;
 }
 export function ChunkMethodItem(props: IProps) {
   const { line } = props;
@@ -368,7 +375,7 @@ export function AutoMetadata() {
     type: FormFieldType.Custom,
     horizontal: true,
     defaultValue: true,
-
+    tooltip: t('knowledgeConfiguration.autoMetadataTip'),
     render: (fieldProps: ControllerRenderProps) => (
       <div className="flex items-center justify-between">
         <Button
@@ -432,3 +439,80 @@ export function AutoMetadata() {
     </>
   );
 }
+
+export const LLMSelect = ({
+  isEdit,
+  field,
+  disabled = false,
+}: {
+  isEdit: boolean;
+  field: FieldValues;
+  name?: string;
+  disabled?: boolean;
+}) => {
+  const { t } = useTranslate('knowledgeConfiguration');
+  const modelOptions = useComposeLlmOptionsByModelTypes([
+    LlmModelType.Chat,
+    LlmModelType.Image2text,
+  ]);
+  return (
+    <SelectWithSearch
+      onChange={async (value) => {
+        field.onChange(value);
+      }}
+      disabled={disabled && !isEdit}
+      value={field.value}
+      options={modelOptions as SelectWithSearchFlagOptionType[]}
+      placeholder={t('embeddingModelPlaceholder')}
+    />
+  );
+};
+
+export function LLMModelItem({ line = 1, isEdit, label, name }: IProps) {
+  const { t } = useTranslate('knowledgeConfiguration');
+  const form = useFormContext();
+  const disabled = useHasParsedDocument(isEdit);
+  return (
+    <>
+      <FormField
+        control={form.control}
+        name={name ?? 'llm_id'}
+        render={({ field }) => (
+          <FormItem className={cn(' items-center space-y-0 ')}>
+            <div
+              className={cn('flex', {
+                ' items-center': line === 1,
+                'flex-col gap-1': line === 2,
+              })}
+            >
+              <FormLabel
+                required
+                tooltip={t('globalIndexModelTip')}
+                className={cn('text-sm  whitespace-wrap ', {
+                  'w-1/4': line === 1,
+                })}
+              >
+                {label ?? t('llmModel')}
+              </FormLabel>
+              <div
+                className={cn('text-text-secondary', { 'w-3/4': line === 1 })}
+              >
+                <FormControl>
+                  <LLMSelect
+                    isEdit={!!isEdit}
+                    field={field}
+                    disabled={disabled}
+                  ></LLMSelect>
+                </FormControl>
+              </div>
+            </div>
+            <div className="flex pt-1">
+              <div className={line === 1 ? 'w-1/4' : ''}></div>
+              <FormMessage />
+            </div>
+          </FormItem>
+        )}
+      />
+    </>
+  );
+}

web/src/pages/dataset/dataset-setting/configuration/email.tsx

@@ -3,11 +3,13 @@ import {
   AutoQuestionsFormField,
 } from '@/components/auto-keywords-form-field';
 import { ConfigurationFormContainer } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function EmailConfiguration() {
   return (
     <ConfigurationFormContainer>
       <>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </>

web/src/pages/dataset/dataset-setting/configuration/laws.tsx

@@ -7,6 +7,7 @@ import {
   ConfigurationFormContainer,
   MainContainer,
 } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function LawsConfiguration() {
   return (
@@ -16,6 +17,7 @@ export function LawsConfiguration() {
       </ConfigurationFormContainer>
 
       <ConfigurationFormContainer>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </ConfigurationFormContainer>

web/src/pages/dataset/dataset-setting/configuration/manual.tsx

@@ -7,6 +7,7 @@ import {
   ConfigurationFormContainer,
   MainContainer,
 } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function ManualConfiguration() {
   return (
@@ -16,6 +17,7 @@ export function ManualConfiguration() {
       </ConfigurationFormContainer>
 
       <ConfigurationFormContainer>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </ConfigurationFormContainer>

web/src/pages/dataset/dataset-setting/configuration/one.tsx

@@ -4,12 +4,14 @@ import {
 } from '@/components/auto-keywords-form-field';
 import { LayoutRecognizeFormField } from '@/components/layout-recognize-form-field';
 import { ConfigurationFormContainer } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function OneConfiguration() {
   return (
     <ConfigurationFormContainer>
       <LayoutRecognizeFormField></LayoutRecognizeFormField>
       <>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </>

web/src/pages/dataset/dataset-setting/configuration/paper.tsx

@@ -7,6 +7,7 @@ import {
   ConfigurationFormContainer,
   MainContainer,
 } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function PaperConfiguration() {
   return (
@@ -16,6 +17,7 @@ export function PaperConfiguration() {
       </ConfigurationFormContainer>
 
       <ConfigurationFormContainer>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </ConfigurationFormContainer>

web/src/pages/dataset/dataset-setting/configuration/picture.tsx

@@ -3,11 +3,13 @@ import {
   AutoQuestionsFormField,
 } from '@/components/auto-keywords-form-field';
 import { ConfigurationFormContainer } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function PictureConfiguration() {
   return (
     <ConfigurationFormContainer>
       <>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </>

web/src/pages/dataset/dataset-setting/configuration/presentation.tsx

@@ -7,6 +7,7 @@ import {
   ConfigurationFormContainer,
   MainContainer,
 } from '../configuration-form-container';
+import { AutoMetadata } from './common-item';
 
 export function PresentationConfiguration() {
   return (
@@ -16,6 +17,7 @@ export function PresentationConfiguration() {
       </ConfigurationFormContainer>
 
       <ConfigurationFormContainer>
+        <AutoMetadata />
         <AutoKeywordsFormField></AutoKeywordsFormField>
         <AutoQuestionsFormField></AutoQuestionsFormField>
       </ConfigurationFormContainer>

web/src/pages/dataset/dataset-setting/form-schema.ts

@@ -96,6 +96,7 @@ export const formSchema = z
           )
           .optional(),
         enable_metadata: z.boolean().optional(),
+        llm_id: z.string().optional(),
       })
       .optional(),
     pagerank: z.number(),

web/src/pages/dataset/dataset-setting/index.tsx

@@ -93,6 +93,7 @@ export default function DatasetSettings() {
         },
         metadata: [],
         enable_metadata: false,
+        llm_id: '',
       },
       pipeline_id: '',
       parseType: 1,

web/src/pages/dataset/dataset/index.tsx

@@ -16,7 +16,7 @@ import { useFetchKnowledgeBaseConfiguration } from '@/hooks/use-knowledge-reques
 import { Pen, Upload } from 'lucide-react';
 import { useMemo } from 'react';
 import { useTranslation } from 'react-i18next';
-import { useManageMetadata } from '../components/metedata/hook';
+import { MetadataType, useManageMetadata } from '../components/metedata/hook';
 import { ManageMetadataModal } from '../components/metedata/manage-modal';
 import { DatasetTable } from './dataset-table';
 import Generate from './generate-button/generate';
@@ -53,7 +53,7 @@ export default function Dataset() {
   const { data: dataSetData } = useFetchKnowledgeBaseConfiguration({
     refreshCount,
   });
-  const { filters, onOpenChange } = useSelectDatasetFilters();
+  const { filters, onOpenChange, filterGroup } = useSelectDatasetFilters();
 
   const {
     createLoading,
@@ -90,6 +90,7 @@ export default function Dataset() {
           onSearchChange={handleInputChange}
           searchString={searchString}
           value={filterValue}
+          filterGroup={filterGroup}
           onChange={handleFilterSubmit}
           onOpenChange={onOpenChange}
           filters={filters}
@@ -105,7 +106,25 @@ export default function Dataset() {
             <Button
               variant={'ghost'}
               className="border border-border-button"
-              onClick={() => showManageMetadataModal()}
+              onClick={() =>
+                showManageMetadataModal({
+                  type: MetadataType.Manage,
+                  isCanAdd: false,
+                  isEditField: true,
+                  title: (
+                    <div className="flex flex-col gap-2">
+                      <div className="text-base font-normal">
+                        {t('knowledgeDetails.metadata.manageMetadata')}
+                      </div>
+                      <div className="text-sm text-text-secondary">
+                        {t(
+                          'knowledgeDetails.metadata.manageMetadataForDataset',
+                        )}
+                      </div>
+                    </div>
+                  ),
+                })
+              }
             >
               <div className="flex gap-1 items-center">
                 <Pen size={14} />
@@ -179,6 +198,7 @@ export default function Dataset() {
             // selectedRowKeys={selectedRowKeys}
             tableData={tableData}
             isCanAdd={metadataConfig.isCanAdd}
+            isEditField={metadataConfig.isEditField}
             isDeleteSingleValue={metadataConfig.isDeleteSingleValue}
             type={metadataConfig.type}
             otherData={metadataConfig.record}

web/src/pages/dataset/dataset/use-select-filters.ts

@@ -49,9 +49,13 @@ export function useSelectDatasetFilters() {
     return [
       { field: 'type', label: 'File Type', list: fileTypes },
       { field: 'run', label: 'Status', list: fileStatus },
-      { field: 'metadata', label: 'metadata', list: metaDataList },
+      { field: 'metadata', label: 'Metadata field', list: metaDataList },
     ] as FilterCollection[];
   }, [fileStatus, fileTypes, metaDataList]);
 
-  return { filters, onOpenChange };
+  const filterGroup = {
+    [t('systemAttribute')]: ['type', 'run'],
+    // [t('metadataField')]: ['metadata'],
+  };
+  return { filters, onOpenChange, filterGroup };
 }

web/src/pages/memories/add-or-edit-modal.tsx

@@ -1,5 +1,4 @@
 import { DynamicForm } from '@/components/dynamic-form';
-import { useModelOptions } from '@/components/llm-setting-items/llm-form-field';
 import { HomeIcon } from '@/components/svg-icon';
 import { Modal } from '@/components/ui/modal/modal';
 import { memo, useMemo } from 'react';
@@ -18,7 +17,7 @@ type IProps = {
 export const AddOrEditModal = memo((props: IProps) => {
   const { open, onClose, onSubmit, initialMemory, isCreate } = props;
   const { t } = useTranslation();
-  const { modelOptions } = useModelOptions();
+  // const { modelOptions } = useModelOptions();
 
   const fields = useMemo(() => {
     if (!isCreate) {
@@ -26,21 +25,22 @@ export const AddOrEditModal = memo((props: IProps) => {
         (field: any) => field.name === 'name',
       );
     } else {
-      const tempFields = createMemoryFields(t).map((field: any) => {
-        if (field.name === 'llm_id') {
-          return {
-            ...field,
-            options: modelOptions,
-          };
-        } else {
-          return {
-            ...field,
-          };
-        }
-      });
-      return tempFields;
+      // const tempFields = createMemoryFields(t).map((field: any) => {
+      //   if (field.name === 'llm_id') {
+      //     return {
+      //       ...field,
+      //       options: modelOptions,
+      //     };
+      //   } else {
+      //     return {
+      //       ...field,
+      //     };
+      //   }
+      // });
+      // return tempFields;
+      return createMemoryFields(t);
     }
-  }, [modelOptions, isCreate]);
+  }, [isCreate, t]);
 
   return (
     <Modal

web/src/pages/memories/constants/index.tsx

@@ -1,5 +1,8 @@
 import { FormFieldConfig, FormFieldType } from '@/components/dynamic-form';
-import { EmbeddingSelect } from '@/pages/dataset/dataset-setting/configuration/common-item';
+import {
+  EmbeddingSelect,
+  LLMSelect,
+} from '@/pages/dataset/dataset-setting/configuration/common-item';
 import { TFunction } from 'i18next';
 export enum MemoryType {
   Raw = 'raw',
@@ -52,6 +55,7 @@ export const createMemoryFields = (t: TFunction) =>
       required: true,
       type: FormFieldType.Select,
       tooltip: t('memories.llmTooltip'),
+      render: (field) => <LLMSelect field={field} isEdit={false} />,
     },
   ] as FormFieldConfig[];
 

web/src/services/knowledge-service.ts

@@ -47,6 +47,7 @@ const {
   runRaptor,
   traceRaptor,
   check_embedding,
+  kbUpdateMetaData,
 } = api;
 
 const methods = {
@@ -215,6 +216,10 @@ const methods = {
     url: check_embedding,
     method: 'post',
   },
+  kbUpdateMetaData: {
+    url: kbUpdateMetaData,
+    method: 'post',
+  },
   // getMetaData: {
   //   url: getMetaData,
   //   method: 'get',

web/src/services/memory-service.ts

@@ -1,6 +1,6 @@
 import api from '@/utils/api';
+import request from '@/utils/next-request';
 import { registerNextServer } from '@/utils/register-server';
-import request from '@/utils/request';
 
 const {
   createMemory,

web/src/utils/api.ts

@@ -79,6 +79,7 @@ export default {
   pipelineRerun: `${api_host}/canvas/rerun`,
   getMetaData: `${api_host}/document/metadata/summary`,
   updateMetaData: `${api_host}/document/metadata/update`,
+  kbUpdateMetaData: `${api_host}/kb/update_metadata_setting`,
 
   // tags
   listTag: (knowledgeId: string) => `${api_host}/kb/${knowledgeId}/tags`,