Feat: enriches Notion connector (#11414)

### What problem does this PR solve?

Enriches rich text (links, mentions, equations), flags to-do blocks with
[x]/[ ], captures block-level equations, builds table HTML, downloads
attachments.

### Type of change

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

agent/tools/retrieval.py

@@ -132,12 +132,12 @@ class Retrieval(ToolBase, ABC):
             metas = DocumentService.get_meta_by_kbs(kb_ids)
             if self._param.meta_data_filter.get("method") == "auto":
                 chat_mdl = LLMBundle(self._canvas.get_tenant_id(), LLMType.CHAT)
-                filters = gen_meta_filter(chat_mdl, metas, query)
-                doc_ids.extend(meta_filter(metas, filters))
+                filters: dict = gen_meta_filter(chat_mdl, metas, query)
+                doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
                 if not doc_ids:
                     doc_ids = None
             elif self._param.meta_data_filter.get("method") == "manual":
-                filters=self._param.meta_data_filter["manual"]
+                filters = self._param.meta_data_filter["manual"]
                 for flt in filters:
                     pat = re.compile(self.variable_ref_patt)
                     s = flt["value"]
@@ -165,9 +165,9 @@ class Retrieval(ToolBase, ABC):
 
                     out_parts.append(s[last:])
                     flt["value"] = "".join(out_parts)
-                doc_ids.extend(meta_filter(metas, filters))
-                if not doc_ids:
-                    doc_ids = None
+                doc_ids.extend(meta_filter(metas, filters, self._param.meta_data_filter.get("logic", "and")))
+                if filters and not doc_ids:
+                    doc_ids = ["-999"]
 
         if self._param.cross_languages:
             query = cross_languages(kbs[0].tenant_id, None, query, self._param.cross_languages)

api/apps/chunk_app.py

@@ -305,14 +305,14 @@ async def retrieval_test():
         metas = DocumentService.get_meta_by_kbs(kb_ids)
         if meta_data_filter.get("method") == "auto":
             chat_mdl = LLMBundle(current_user.id, LLMType.CHAT, llm_name=search_config.get("chat_id", ""))
-            filters = gen_meta_filter(chat_mdl, metas, question)
-            doc_ids.extend(meta_filter(metas, filters))
+            filters: dict = gen_meta_filter(chat_mdl, metas, question)
+            doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
             if not doc_ids:
                 doc_ids = None
         elif meta_data_filter.get("method") == "manual":
-            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
-            if not doc_ids:
-                doc_ids = None
+            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
+            if meta_data_filter["manual"] and not doc_ids:
+                doc_ids = ["-999"]
 
     try:
         tenants = UserTenantService.query(user_id=current_user.id)

api/apps/sdk/agents.py

@@ -159,10 +159,10 @@ async def webhook(tenant_id: str, agent_id: str):
             data=False, message=str(e),
             code=RetCode.EXCEPTION_ERROR)
 
-    def sse():
+    async def sse():
         nonlocal canvas
         try:
-            for ans in canvas.run(query=req.get("query", ""), files=req.get("files", []), user_id=req.get("user_id", tenant_id), webhook_payload=req):
+            async for ans in canvas.run(query=req.get("query", ""), files=req.get("files", []), user_id=req.get("user_id", tenant_id), webhook_payload=req):
                 yield "data:" + json.dumps(ans, ensure_ascii=False) + "\n\n"
 
             cvs.dsl = json.loads(str(canvas))

api/apps/sdk/dify_retrieval.py

@@ -120,7 +120,7 @@ async def retrieval(tenant_id):
     retrieval_setting = req.get("retrieval_setting", {})
     similarity_threshold = float(retrieval_setting.get("score_threshold", 0.0))
     top = int(retrieval_setting.get("top_k", 1024))
-    metadata_condition = req.get("metadata_condition", {})
+    metadata_condition = req.get("metadata_condition", {}) or {}
     metas = DocumentService.get_meta_by_kbs([kb_id])
 
     doc_ids = []
@@ -132,7 +132,7 @@ async def retrieval(tenant_id):
 
         embd_mdl = LLMBundle(kb.tenant_id, LLMType.EMBEDDING.value, llm_name=kb.embd_id)
         if metadata_condition:
-            doc_ids.extend(meta_filter(metas, convert_conditions(metadata_condition)))
+            doc_ids.extend(meta_filter(metas, convert_conditions(metadata_condition), metadata_condition.get("logic", "and")))
         if not doc_ids and metadata_condition:
             doc_ids = ["-999"]
         ranks = settings.retriever.retrieval(

api/apps/sdk/doc.py

@@ -1442,9 +1442,11 @@ async def retrieval_test(tenant_id):
         if doc_id not in doc_ids_list:
             return get_error_data_result(f"The datasets don't own the document {doc_id}")
     if not doc_ids:
-        metadata_condition = req.get("metadata_condition", {})
+        metadata_condition = req.get("metadata_condition", {}) or {}
         metas = DocumentService.get_meta_by_kbs(kb_ids)
-        doc_ids = meta_filter(metas, convert_conditions(metadata_condition))
+        doc_ids = meta_filter(metas, convert_conditions(metadata_condition), metadata_condition.get("logic", "and"))
+        if metadata_condition and not doc_ids:
+            doc_ids = ["-999"]
     similarity_threshold = float(req.get("similarity_threshold", 0.2))
     vector_similarity_weight = float(req.get("vector_similarity_weight", 0.3))
     top = int(req.get("top_k", 1024))

api/apps/sdk/session.py

@@ -428,17 +428,15 @@ async def agents_completion_openai_compatibility(tenant_id, agent_id):
         return resp
     else:
         # For non-streaming, just return the response directly
-        response = next(
-            completion_openai(
+        async for response in completion_openai(
                 tenant_id,
                 agent_id,
                 question,
                 session_id=req.pop("session_id", req.get("id", "")) or req.get("metadata", {}).get("id", ""),
                 stream=False,
                 **req,
-            )
-        )
-        return jsonify(response)
+            ):
+            return jsonify(response)
 
 
 @manager.route("/agents/<agent_id>/completions", methods=["POST"])  # noqa: F821
@@ -448,8 +446,8 @@ async def agent_completions(tenant_id, agent_id):
 
     if req.get("stream", True):
 
-        def generate():
-            for answer in agent_completion(tenant_id=tenant_id, agent_id=agent_id, **req):
+        async def generate():
+            async for answer in agent_completion(tenant_id=tenant_id, agent_id=agent_id, **req):
                 if isinstance(answer, str):
                     try:
                         ans = json.loads(answer[5:])  # remove "data:"
@@ -473,7 +471,7 @@ async def agent_completions(tenant_id, agent_id):
     full_content = ""
     reference = {}
     final_ans = ""
-    for answer in agent_completion(tenant_id=tenant_id, agent_id=agent_id, **req):
+    async for answer in agent_completion(tenant_id=tenant_id, agent_id=agent_id, **req):
         try:
             ans = json.loads(answer[5:])
 
@@ -875,7 +873,7 @@ async def agent_bot_completions(agent_id):
         resp.headers.add_header("Content-Type", "text/event-stream; charset=utf-8")
         return resp
 
-    for answer in agent_completion(objs[0].tenant_id, agent_id, **req):
+    async for answer in agent_completion(objs[0].tenant_id, agent_id, **req):
         return get_result(data=answer)
 
 
@@ -977,14 +975,14 @@ async def retrieval_test_embedded():
         metas = DocumentService.get_meta_by_kbs(kb_ids)
         if meta_data_filter.get("method") == "auto":
             chat_mdl = LLMBundle(tenant_id, LLMType.CHAT, llm_name=search_config.get("chat_id", ""))
-            filters = gen_meta_filter(chat_mdl, metas, question)
-            doc_ids.extend(meta_filter(metas, filters))
+            filters: dict = gen_meta_filter(chat_mdl, metas, question)
+            doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
             if not doc_ids:
                 doc_ids = None
         elif meta_data_filter.get("method") == "manual":
-            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
-            if not doc_ids:
-                doc_ids = None
+            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
+            if meta_data_filter["manual"] and not doc_ids:
+                doc_ids = ["-999"]
 
     try:
         tenants = UserTenantService.query(user_id=tenant_id)

api/db/services/canvas_service.py

@@ -177,7 +177,7 @@ class UserCanvasService(CommonService):
         return True
 
 
-def completion(tenant_id, agent_id, session_id=None, **kwargs):
+async def completion(tenant_id, agent_id, session_id=None, **kwargs):
     query = kwargs.get("query", "") or kwargs.get("question", "")
     files = kwargs.get("files", [])
     inputs = kwargs.get("inputs", {})
@@ -219,10 +219,14 @@ def completion(tenant_id, agent_id, session_id=None, **kwargs):
         "id": message_id
     })
     txt = ""
-    for ans in canvas.run(query=query, files=files, user_id=user_id, inputs=inputs):
+    async for ans in canvas.run(query=query, files=files, user_id=user_id, inputs=inputs):
         ans["session_id"] = session_id
         if ans["event"] == "message":
             txt += ans["data"]["content"]
+            if ans["data"].get("start_to_think", False):
+                txt += "<think>"
+            elif ans["data"].get("end_to_think", False):
+                txt += "</think>"
         yield "data:" + json.dumps(ans, ensure_ascii=False) + "\n\n"
 
     conv.message.append({"role": "assistant", "content": txt, "created_at": time.time(), "id": message_id})
@@ -233,7 +237,7 @@ def completion(tenant_id, agent_id, session_id=None, **kwargs):
     API4ConversationService.append_message(conv["id"], conv)
 
 
-def completion_openai(tenant_id, agent_id, question, session_id=None, stream=True, **kwargs):
+async def completion_openai(tenant_id, agent_id, question, session_id=None, stream=True, **kwargs):
     tiktoken_encoder = tiktoken.get_encoding("cl100k_base")
     prompt_tokens = len(tiktoken_encoder.encode(str(question)))
     user_id = kwargs.get("user_id", "")
@@ -241,7 +245,7 @@ def completion_openai(tenant_id, agent_id, question, session_id=None, stream=Tru
     if stream:
         completion_tokens = 0
         try:
-            for ans in completion(
+            async for ans in completion(
                 tenant_id=tenant_id,
                 agent_id=agent_id,
                 session_id=session_id,
@@ -300,7 +304,7 @@ def completion_openai(tenant_id, agent_id, question, session_id=None, stream=Tru
         try:
             all_content = ""
             reference = {}
-            for ans in completion(
+            async for ans in completion(
                 tenant_id=tenant_id,
                 agent_id=agent_id,
                 session_id=session_id,

api/db/services/connector_service.py

@@ -15,6 +15,7 @@
 #
 import logging
 from datetime import datetime
+import os
 from typing import Tuple, List
 
 from anthropic import BaseModel
@@ -103,7 +104,8 @@ class SyncLogsService(CommonService):
             Knowledgebase.avatar.alias("kb_avatar"),
             Connector2Kb.auto_parse,
             cls.model.from_beginning.alias("reindex"),
-            cls.model.status
+            cls.model.status,
+            cls.model.update_time
         ]
         if not connector_id:
             fields.append(Connector.config)
@@ -116,7 +118,11 @@ class SyncLogsService(CommonService):
         if connector_id:
             query = query.where(cls.model.connector_id == connector_id)
         else:
-            interval_expr = SQL("INTERVAL `t2`.`refresh_freq` MINUTE")
+            database_type = os.getenv("DB_TYPE", "mysql")
+            if "postgres" in database_type.lower():
+                interval_expr = SQL("make_interval(mins => t2.refresh_freq)")
+            else:
+                interval_expr = SQL("INTERVAL `t2`.`refresh_freq` MINUTE")
             query = query.where(
                 Connector.input_type == InputType.POLL,
                 Connector.status == TaskStatus.SCHEDULE,

api/db/services/dialog_service.py

@@ -287,7 +287,7 @@ def convert_conditions(metadata_condition):
     ]
 
 
-def meta_filter(metas: dict, filters: list[dict]):
+def meta_filter(metas: dict, filters: list[dict], logic: str = "and"):
     doc_ids = set([])
 
     def filter_out(v2docs, operator, value):
@@ -331,7 +331,10 @@ def meta_filter(metas: dict, filters: list[dict]):
             if not doc_ids:
                 doc_ids = set(ids)
             else:
-                doc_ids = doc_ids & set(ids)
+                if logic == "and":
+                    doc_ids = doc_ids & set(ids)
+                else:
+                    doc_ids = doc_ids | set(ids)
             if not doc_ids:
                 return []
     return list(doc_ids)
@@ -407,14 +410,15 @@ def chat(dialog, messages, stream=True, **kwargs):
     if dialog.meta_data_filter:
         metas = DocumentService.get_meta_by_kbs(dialog.kb_ids)
         if dialog.meta_data_filter.get("method") == "auto":
-            filters = gen_meta_filter(chat_mdl, metas, questions[-1])
-            attachments.extend(meta_filter(metas, filters))
+            filters: dict = gen_meta_filter(chat_mdl, metas, questions[-1])
+            attachments.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
             if not attachments:
                 attachments = None
         elif dialog.meta_data_filter.get("method") == "manual":
-            attachments.extend(meta_filter(metas, dialog.meta_data_filter["manual"]))
-            if not attachments:
-                attachments = None
+            conds = dialog.meta_data_filter["manual"]
+            attachments.extend(meta_filter(metas, conds, dialog.meta_data_filter.get("logic", "and")))
+            if conds and not attachments:
+                attachments = ["-999"]
 
     if prompt_config.get("keyword", False):
         questions[-1] += keyword_extraction(chat_mdl, questions[-1])
@@ -778,14 +782,14 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
     if meta_data_filter:
         metas = DocumentService.get_meta_by_kbs(kb_ids)
         if meta_data_filter.get("method") == "auto":
-            filters = gen_meta_filter(chat_mdl, metas, question)
-            doc_ids.extend(meta_filter(metas, filters))
+            filters: dict = gen_meta_filter(chat_mdl, metas, question)
+            doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
             if not doc_ids:
                 doc_ids = None
         elif meta_data_filter.get("method") == "manual":
-            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
-            if not doc_ids:
-                doc_ids = None
+            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
+            if meta_data_filter["manual"] and not doc_ids:
+                doc_ids = ["-999"]
 
     kbinfos = retriever.retrieval(
         question=question,
@@ -853,14 +857,14 @@ def gen_mindmap(question, kb_ids, tenant_id, search_config={}):
     if meta_data_filter:
         metas = DocumentService.get_meta_by_kbs(kb_ids)
         if meta_data_filter.get("method") == "auto":
-            filters = gen_meta_filter(chat_mdl, metas, question)
-            doc_ids.extend(meta_filter(metas, filters))
+            filters: dict = gen_meta_filter(chat_mdl, metas, question)
+            doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
             if not doc_ids:
                 doc_ids = None
         elif meta_data_filter.get("method") == "manual":
-            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
-            if not doc_ids:
-                doc_ids = None
+            doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
+            if meta_data_filter["manual"] and not doc_ids:
+                doc_ids = ["-999"]
 
     ranks = settings.retriever.retrieval(
         question=question,

common/data_source/notion_connector.py

@@ -1,38 +1,45 @@
+import html
 import logging
 from collections.abc import Generator
+from datetime import datetime, timezone
+from pathlib import Path
 from typing import Any, Optional
+from urllib.parse import urlparse
+
 from retry import retry
 
 from common.data_source.config import (
     INDEX_BATCH_SIZE,
-    DocumentSource, NOTION_CONNECTOR_DISABLE_RECURSIVE_PAGE_LOOKUP
+    NOTION_CONNECTOR_DISABLE_RECURSIVE_PAGE_LOOKUP,
+    DocumentSource,
+)
+from common.data_source.exceptions import (
+    ConnectorMissingCredentialError,
+    ConnectorValidationError,
+    CredentialExpiredError,
+    InsufficientPermissionsError,
+    UnexpectedValidationError,
 )
 from common.data_source.interfaces import (
     LoadConnector,
     PollConnector,
-    SecondsSinceUnixEpoch
+    SecondsSinceUnixEpoch,
 )
 from common.data_source.models import (
     Document,
-    TextSection, GenerateDocumentsOutput
-)
-from common.data_source.exceptions import (
-    ConnectorValidationError,
-    CredentialExpiredError,
-    InsufficientPermissionsError,
-    UnexpectedValidationError, ConnectorMissingCredentialError
-)
-from common.data_source.models import (
-    NotionPage,
+    GenerateDocumentsOutput,
     NotionBlock,
-    NotionSearchResponse
+    NotionPage,
+    NotionSearchResponse,
+    TextSection,
 )
 from common.data_source.utils import (
-    rl_requests,
     batch_generator,
+    datetime_from_string,
     fetch_notion_data,
+    filter_pages_by_time,
     properties_to_str,
-    filter_pages_by_time, datetime_from_string
+    rl_requests,
 )
 
 
@@ -61,11 +68,9 @@ class NotionConnector(LoadConnector, PollConnector):
         self.recursive_index_enabled = recursive_index_enabled or bool(root_page_id)
 
     @retry(tries=3, delay=1, backoff=2)
-    def _fetch_child_blocks(
-        self, block_id: str, cursor: Optional[str] = None
-    ) -> dict[str, Any] | None:
+    def _fetch_child_blocks(self, block_id: str, cursor: Optional[str] = None) -> dict[str, Any] | None:
         """Fetch all child blocks via the Notion API."""
-        logging.debug(f"Fetching children of block with ID '{block_id}'")
+        logging.debug(f"[Notion]: Fetching children of block with ID {block_id}")
         block_url = f"https://api.notion.com/v1/blocks/{block_id}/children"
         query_params = {"start_cursor": cursor} if cursor else None
 
@@ -79,49 +84,42 @@ class NotionConnector(LoadConnector, PollConnector):
             response.raise_for_status()
             return response.json()
         except Exception as e:
-            if hasattr(e, 'response') and e.response.status_code == 404:
-                logging.error(
-                    f"Unable to access block with ID '{block_id}'. "
-                    f"This is likely due to the block not being shared with the integration."
-                )
+            if hasattr(e, "response") and e.response.status_code == 404:
+                logging.error(f"[Notion]: Unable to access block with ID {block_id}. This is likely due to the block not being shared with the integration.")
                 return None
             else:
-                logging.exception(f"Error fetching blocks: {e}")
+                logging.exception(f"[Notion]: Error fetching blocks: {e}")
                 raise
 
     @retry(tries=3, delay=1, backoff=2)
     def _fetch_page(self, page_id: str) -> NotionPage:
         """Fetch a page from its ID via the Notion API."""
-        logging.debug(f"Fetching page for ID '{page_id}'")
+        logging.debug(f"[Notion]: Fetching page for ID {page_id}")
         page_url = f"https://api.notion.com/v1/pages/{page_id}"
 
         try:
             data = fetch_notion_data(page_url, self.headers, "GET")
             return NotionPage(**data)
         except Exception as e:
-            logging.warning(f"Failed to fetch page, trying database for ID '{page_id}': {e}")
+            logging.warning(f"[Notion]: Failed to fetch page, trying database for ID {page_id}: {e}")
             return self._fetch_database_as_page(page_id)
 
     @retry(tries=3, delay=1, backoff=2)
     def _fetch_database_as_page(self, database_id: str) -> NotionPage:
         """Attempt to fetch a database as a page."""
-        logging.debug(f"Fetching database for ID '{database_id}' as a page")
+        logging.debug(f"[Notion]: Fetching database for ID {database_id} as a page")
         database_url = f"https://api.notion.com/v1/databases/{database_id}"
 
         data = fetch_notion_data(database_url, self.headers, "GET")
         database_name = data.get("title")
-        database_name = (
-            database_name[0].get("text", {}).get("content") if database_name else None
-        )
+        database_name = database_name[0].get("text", {}).get("content") if database_name else None
 
         return NotionPage(**data, database_name=database_name)
 
     @retry(tries=3, delay=1, backoff=2)
-    def _fetch_database(
-        self, database_id: str, cursor: Optional[str] = None
-    ) -> dict[str, Any]:
+    def _fetch_database(self, database_id: str, cursor: Optional[str] = None) -> dict[str, Any]:
         """Fetch a database from its ID via the Notion API."""
-        logging.debug(f"Fetching database for ID '{database_id}'")
+        logging.debug(f"[Notion]: Fetching database for ID {database_id}")
         block_url = f"https://api.notion.com/v1/databases/{database_id}/query"
         body = {"start_cursor": cursor} if cursor else None
 
@@ -129,17 +127,12 @@ class NotionConnector(LoadConnector, PollConnector):
             data = fetch_notion_data(block_url, self.headers, "POST", body)
             return data
         except Exception as e:
-            if hasattr(e, 'response') and e.response.status_code in [404, 400]:
-                logging.error(
-                    f"Unable to access database with ID '{database_id}'. "
-                    f"This is likely due to the database not being shared with the integration."
-                )
+            if hasattr(e, "response") and e.response.status_code in [404, 400]:
+                logging.error(f"[Notion]: Unable to access database with ID {database_id}. This is likely due to the database not being shared with the integration.")
                 return {"results": [], "next_cursor": None}
             raise
 
-    def _read_pages_from_database(
-        self, database_id: str
-    ) -> tuple[list[NotionBlock], list[str]]:
+    def _read_pages_from_database(self, database_id: str) -> tuple[list[NotionBlock], list[str]]:
         """Returns a list of top level blocks and all page IDs in the database."""
         result_blocks: list[NotionBlock] = []
         result_pages: list[str] = []
@@ -158,10 +151,10 @@ class NotionConnector(LoadConnector, PollConnector):
 
                 if self.recursive_index_enabled:
                     if obj_type == "page":
-                        logging.debug(f"Found page with ID '{obj_id}' in database '{database_id}'")
+                        logging.debug(f"[Notion]: Found page with ID {obj_id} in database {database_id}")
                         result_pages.append(result["id"])
                     elif obj_type == "database":
-                        logging.debug(f"Found database with ID '{obj_id}' in database '{database_id}'")
+                        logging.debug(f"[Notion]: Found database with ID {obj_id} in database {database_id}")
                         _, child_pages = self._read_pages_from_database(obj_id)
                         result_pages.extend(child_pages)
 
@@ -172,44 +165,229 @@ class NotionConnector(LoadConnector, PollConnector):
 
         return result_blocks, result_pages
 
-    def _read_blocks(self, base_block_id: str) -> tuple[list[NotionBlock], list[str]]:
-        """Reads all child blocks for the specified block, returns blocks and child page ids."""
+    def _extract_rich_text(self, rich_text_array: list[dict[str, Any]]) -> str:
+        collected_text: list[str] = []
+        for rich_text in rich_text_array:
+            content = ""
+            r_type = rich_text.get("type")
+
+            if r_type == "equation":
+                expr = rich_text.get("equation", {}).get("expression")
+                if expr:
+                    content = expr
+            elif r_type == "mention":
+                mention = rich_text.get("mention", {}) or {}
+                mention_type = mention.get("type")
+                mention_value = mention.get(mention_type, {}) if mention_type else {}
+                if mention_type == "date":
+                    start = mention_value.get("start")
+                    end = mention_value.get("end")
+                    if start and end:
+                        content = f"{start} - {end}"
+                    elif start:
+                        content = start
+                elif mention_type in {"page", "database"}:
+                    content = mention_value.get("id", rich_text.get("plain_text", ""))
+                elif mention_type == "link_preview":
+                    content = mention_value.get("url", rich_text.get("plain_text", ""))
+                else:
+                    content = rich_text.get("plain_text", "") or str(mention_value)
+            else:
+                if rich_text.get("plain_text"):
+                    content = rich_text["plain_text"]
+                elif "text" in rich_text and rich_text["text"].get("content"):
+                    content = rich_text["text"]["content"]
+
+            href = rich_text.get("href")
+            if content and href:
+                content = f"{content} ({href})"
+
+            if content:
+                collected_text.append(content)
+
+        return "".join(collected_text).strip()
+
+    def _build_table_html(self, table_block_id: str) -> str | None:
+        rows: list[str] = []
+        cursor = None
+        while True:
+            data = self._fetch_child_blocks(table_block_id, cursor)
+            if data is None:
+                break
+
+            for result in data["results"]:
+                if result.get("type") != "table_row":
+                    continue
+                cells_html: list[str] = []
+                for cell in result["table_row"].get("cells", []):
+                    cell_text = self._extract_rich_text(cell)
+                    cell_html = html.escape(cell_text) if cell_text else ""
+                    cells_html.append(f"<td>{cell_html}</td>")
+                rows.append(f"<tr>{''.join(cells_html)}</tr>")
+
+            if data.get("next_cursor") is None:
+                break
+            cursor = data["next_cursor"]
+
+        if not rows:
+            return None
+        return "<table>\n" + "\n".join(rows) + "\n</table>"
+
+    def _download_file(self, url: str) -> bytes | None:
+        try:
+            response = rl_requests.get(url, timeout=60)
+            response.raise_for_status()
+            return response.content
+        except Exception as exc:
+            logging.warning(f"[Notion]: Failed to download Notion file from {url}: {exc}")
+            return None
+
+    def _extract_file_metadata(self, result_obj: dict[str, Any], block_id: str) -> tuple[str | None, str, str | None]:
+        file_source_type = result_obj.get("type")
+        file_source = result_obj.get(file_source_type, {}) if file_source_type else {}
+        url = file_source.get("url")
+
+        name = result_obj.get("name") or file_source.get("name")
+        if url and not name:
+            parsed_name = Path(urlparse(url).path).name
+            name = parsed_name or f"notion_file_{block_id}"
+        elif not name:
+            name = f"notion_file_{block_id}"
+
+        caption = self._extract_rich_text(result_obj.get("caption", [])) if "caption" in result_obj else None
+
+        return url, name, caption
+
+    def _build_attachment_document(
+        self,
+        block_id: str,
+        url: str,
+        name: str,
+        caption: Optional[str],
+        page_last_edited_time: Optional[str],
+    ) -> Document | None:
+        file_bytes = self._download_file(url)
+        if file_bytes is None:
+            return None
+
+        extension = Path(name).suffix or Path(urlparse(url).path).suffix or ".bin"
+        if extension and not extension.startswith("."):
+            extension = f".{extension}"
+        if not extension:
+            extension = ".bin"
+
+        updated_at = datetime_from_string(page_last_edited_time) if page_last_edited_time else datetime.now(timezone.utc)
+        semantic_identifier = caption or name or f"Notion file {block_id}"
+
+        return Document(
+            id=block_id,
+            blob=file_bytes,
+            source=DocumentSource.NOTION,
+            semantic_identifier=semantic_identifier,
+            extension=extension,
+            size_bytes=len(file_bytes),
+            doc_updated_at=updated_at,
+        )
+
+    def _read_blocks(self, base_block_id: str, page_last_edited_time: Optional[str] = None) -> tuple[list[NotionBlock], list[str], list[Document]]:
         result_blocks: list[NotionBlock] = []
         child_pages: list[str] = []
+        attachments: list[Document] = []
         cursor = None
 
         while True:
             data = self._fetch_child_blocks(base_block_id, cursor)
 
             if data is None:
-                return result_blocks, child_pages
+                return result_blocks, child_pages, attachments
 
             for result in data["results"]:
-                logging.debug(f"Found child block for block with ID '{base_block_id}': {result}")
+                logging.debug(f"[Notion]: Found child block for block with ID {base_block_id}: {result}")
                 result_block_id = result["id"]
                 result_type = result["type"]
                 result_obj = result[result_type]
 
                 if result_type in ["ai_block", "unsupported", "external_object_instance_page"]:
-                    logging.warning(f"Skipping unsupported block type '{result_type}'")
+                    logging.warning(f"[Notion]: Skipping unsupported block type {result_type}")
+                    continue
+
+                if result_type == "table":
+                    table_html = self._build_table_html(result_block_id)
+                    if table_html:
+                        result_blocks.append(
+                            NotionBlock(
+                                id=result_block_id,
+                                text=table_html,
+                                prefix="\n\n",
+                            )
+                        )
+                    continue
+
+                if result_type == "equation":
+                    expr = result_obj.get("expression")
+                    if expr:
+                        result_blocks.append(
+                            NotionBlock(
+                                id=result_block_id,
+                                text=expr,
+                                prefix="\n",
+                            )
+                        )
                     continue
 
                 cur_result_text_arr = []
                 if "rich_text" in result_obj:
-                    for rich_text in result_obj["rich_text"]:
-                        if "text" in rich_text:
-                            text = rich_text["text"]["content"]
-                            cur_result_text_arr.append(text)
+                    text = self._extract_rich_text(result_obj["rich_text"])
+                    if text:
+                        cur_result_text_arr.append(text)
+
+                if result_type == "bulleted_list_item":
+                    if cur_result_text_arr:
+                        cur_result_text_arr[0] = f"- {cur_result_text_arr[0]}"
+                    else:
+                        cur_result_text_arr = ["- "]
+
+                if result_type == "numbered_list_item":
+                    if cur_result_text_arr:
+                        cur_result_text_arr[0] = f"1. {cur_result_text_arr[0]}"
+                    else:
+                        cur_result_text_arr = ["1. "]
+
+                if result_type == "to_do":
+                    checked = result_obj.get("checked")
+                    checkbox_prefix = "[x]" if checked else "[ ]"
+                    if cur_result_text_arr:
+                        cur_result_text_arr = [f"{checkbox_prefix} {cur_result_text_arr[0]}"] + cur_result_text_arr[1:]
+                    else:
+                        cur_result_text_arr = [checkbox_prefix]
+
+                if result_type in {"file", "image", "pdf", "video", "audio"}:
+                    file_url, file_name, caption = self._extract_file_metadata(result_obj, result_block_id)
+                    if file_url:
+                        attachment_doc = self._build_attachment_document(
+                            block_id=result_block_id,
+                            url=file_url,
+                            name=file_name,
+                            caption=caption,
+                            page_last_edited_time=page_last_edited_time,
+                        )
+                        if attachment_doc:
+                            attachments.append(attachment_doc)
+
+                        attachment_label = caption or file_name
+                        if attachment_label:
+                            cur_result_text_arr.append(f"{result_type.capitalize()}: {attachment_label}")
 
                 if result["has_children"]:
                     if result_type == "child_page":
                         child_pages.append(result_block_id)
                     else:
-                        logging.debug(f"Entering sub-block: {result_block_id}")
-                        subblocks, subblock_child_pages = self._read_blocks(result_block_id)
-                        logging.debug(f"Finished sub-block: {result_block_id}")
+                        logging.debug(f"[Notion]: Entering sub-block: {result_block_id}")
+                        subblocks, subblock_child_pages, subblock_attachments = self._read_blocks(result_block_id, page_last_edited_time)
+                        logging.debug(f"[Notion]: Finished sub-block: {result_block_id}")
                         result_blocks.extend(subblocks)
                         child_pages.extend(subblock_child_pages)
+                        attachments.extend(subblock_attachments)
 
                 if result_type == "child_database":
                     inner_blocks, inner_child_pages = self._read_pages_from_database(result_block_id)
@@ -231,7 +409,7 @@ class NotionConnector(LoadConnector, PollConnector):
 
             cursor = data["next_cursor"]
 
-        return result_blocks, child_pages
+        return result_blocks, child_pages, attachments
 
     def _read_page_title(self, page: NotionPage) -> Optional[str]:
         """Extracts the title from a Notion page."""
@@ -245,9 +423,7 @@ class NotionConnector(LoadConnector, PollConnector):
 
         return None
 
-    def _read_pages(
-        self, pages: list[NotionPage]
-    ) -> Generator[Document, None, None]:
+    def _read_pages(self, pages: list[NotionPage], start: SecondsSinceUnixEpoch | None = None, end: SecondsSinceUnixEpoch | None = None) -> Generator[Document, None, None]:
         """Reads pages for rich text content and generates Documents."""
         all_child_page_ids: list[str] = []
 
@@ -255,11 +431,17 @@ class NotionConnector(LoadConnector, PollConnector):
             if isinstance(page, dict):
                 page = NotionPage(**page)
             if page.id in self.indexed_pages:
-                logging.debug(f"Already indexed page with ID '{page.id}'. Skipping.")
+                logging.debug(f"[Notion]: Already indexed page with ID {page.id}. Skipping.")
                 continue
 
-            logging.info(f"Reading page with ID '{page.id}', with url {page.url}")
-            page_blocks, child_page_ids = self._read_blocks(page.id)
+            if start is not None and end is not None:
+                page_ts = datetime_from_string(page.last_edited_time).timestamp()
+                if not (page_ts > start and page_ts <= end):
+                    logging.debug(f"[Notion]: Skipping page {page.id} outside polling window.")
+                    continue
+
+            logging.info(f"[Notion]: Reading page with ID {page.id}, with url {page.url}")
+            page_blocks, child_page_ids, attachment_docs = self._read_blocks(page.id, page.last_edited_time)
             all_child_page_ids.extend(child_page_ids)
             self.indexed_pages.add(page.id)
 
@@ -268,14 +450,12 @@ class NotionConnector(LoadConnector, PollConnector):
 
             if not page_blocks:
                 if not raw_page_title:
-                    logging.warning(f"No blocks OR title found for page with ID '{page.id}'. Skipping.")
+                    logging.warning(f"[Notion]: No blocks OR title found for page with ID {page.id}. Skipping.")
                     continue
 
                 text = page_title
                 if page.properties:
-                    text += "\n\n" + "\n".join(
-                        [f"{key}: {value}" for key, value in page.properties.items()]
-                    )
+                    text += "\n\n" + "\n".join([f"{key}: {value}" for key, value in page.properties.items()])
                 sections = [TextSection(link=page.url, text=text)]
             else:
                 sections = [
@@ -286,45 +466,39 @@ class NotionConnector(LoadConnector, PollConnector):
                     for block in page_blocks
                 ]
 
-            blob = ("\n".join([sec.text for sec in sections])).encode("utf-8")
+            joined_text = "\n".join(sec.text for sec in sections)
+            blob = joined_text.encode("utf-8")
             yield Document(
-                id=page.id,
-                blob=blob,
-                source=DocumentSource.NOTION,
-                semantic_identifier=page_title,
-                extension=".txt",
-                size_bytes=len(blob),
-                doc_updated_at=datetime_from_string(page.last_edited_time)
+                id=page.id, blob=blob, source=DocumentSource.NOTION, semantic_identifier=page_title, extension=".txt", size_bytes=len(blob), doc_updated_at=datetime_from_string(page.last_edited_time)
             )
 
+            for attachment_doc in attachment_docs:
+                yield attachment_doc
+
         if self.recursive_index_enabled and all_child_page_ids:
             for child_page_batch_ids in batch_generator(all_child_page_ids, INDEX_BATCH_SIZE):
-                child_page_batch = [
-                    self._fetch_page(page_id)
-                    for page_id in child_page_batch_ids
-                    if page_id not in self.indexed_pages
-                ]
-                yield from self._read_pages(child_page_batch)
+                child_page_batch = [self._fetch_page(page_id) for page_id in child_page_batch_ids if page_id not in self.indexed_pages]
+                yield from self._read_pages(child_page_batch, start, end)
 
     @retry(tries=3, delay=1, backoff=2)
     def _search_notion(self, query_dict: dict[str, Any]) -> NotionSearchResponse:
         """Search for pages from a Notion database."""
-        logging.debug(f"Searching for pages in Notion with query_dict: {query_dict}")
+        logging.debug(f"[Notion]: Searching for pages in Notion with query_dict: {query_dict}")
         data = fetch_notion_data("https://api.notion.com/v1/search", self.headers, "POST", query_dict)
         return NotionSearchResponse(**data)
 
-    def _recursive_load(self) -> Generator[list[Document], None, None]:
+    def _recursive_load(self, start: SecondsSinceUnixEpoch | None = None, end: SecondsSinceUnixEpoch | None = None) -> Generator[list[Document], None, None]:
         """Recursively load pages starting from root page ID."""
         if self.root_page_id is None or not self.recursive_index_enabled:
             raise RuntimeError("Recursive page lookup is not enabled")
 
-        logging.info(f"Recursively loading pages from Notion based on root page with ID: {self.root_page_id}")
+        logging.info(f"[Notion]: Recursively loading pages from Notion based on root page with ID: {self.root_page_id}")
         pages = [self._fetch_page(page_id=self.root_page_id)]
-        yield from batch_generator(self._read_pages(pages), self.batch_size)
+        yield from batch_generator(self._read_pages(pages, start, end), self.batch_size)
 
     def load_credentials(self, credentials: dict[str, Any]) -> dict[str, Any] | None:
         """Applies integration token to headers."""
-        self.headers["Authorization"] = f'Bearer {credentials["notion_integration_token"]}'
+        self.headers["Authorization"] = f"Bearer {credentials['notion_integration_token']}"
         return None
 
     def load_from_state(self) -> GenerateDocumentsOutput:
@@ -348,12 +522,10 @@ class NotionConnector(LoadConnector, PollConnector):
             else:
                 break
 
-    def poll_source(
-        self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch
-    ) -> GenerateDocumentsOutput:
+    def poll_source(self, start: SecondsSinceUnixEpoch, end: SecondsSinceUnixEpoch) -> GenerateDocumentsOutput:
         """Poll Notion for updated pages within a time period."""
         if self.recursive_index_enabled and self.root_page_id:
-            yield from self._recursive_load()
+            yield from self._recursive_load(start, end)
             return
 
         query_dict = {
@@ -367,7 +539,7 @@ class NotionConnector(LoadConnector, PollConnector):
             pages = filter_pages_by_time(db_res.results, start, end, "last_edited_time")
 
             if pages:
-                yield from batch_generator(self._read_pages(pages), self.batch_size)
+                yield from batch_generator(self._read_pages(pages, start, end), self.batch_size)
                 if db_res.has_more:
                     query_dict["start_cursor"] = db_res.next_cursor
                 else:

common/settings.py

@@ -27,6 +27,7 @@ from common.constants import SVR_QUEUE_NAME, Storage
 import rag.utils
 import rag.utils.es_conn
 import rag.utils.infinity_conn
+import rag.utils.ob_conn
 import rag.utils.opensearch_conn
 from rag.utils.azure_sas_conn import RAGFlowAzureSasBlob
 from rag.utils.azure_spn_conn import RAGFlowAzureSpnBlob
@@ -103,6 +104,7 @@ INFINITY = {}
 AZURE = {}
 S3 = {}
 MINIO = {}
+OB = {}
 OSS = {}
 OS = {}
 
@@ -227,7 +229,7 @@ def init_settings():
     FEISHU_OAUTH = get_base_config("oauth", {}).get("feishu")
     OAUTH_CONFIG = get_base_config("oauth", {})
 
-    global DOC_ENGINE, docStoreConn, ES, OS, INFINITY
+    global DOC_ENGINE, docStoreConn, ES, OB, OS, INFINITY
     DOC_ENGINE = os.environ.get("DOC_ENGINE", "elasticsearch")
     # DOC_ENGINE = os.environ.get('DOC_ENGINE', "opensearch")
     lower_case_doc_engine = DOC_ENGINE.lower()
@@ -240,6 +242,9 @@ def init_settings():
     elif lower_case_doc_engine == "opensearch":
         OS = get_base_config("os", {})
         docStoreConn = rag.utils.opensearch_conn.OSConnection()
+    elif lower_case_doc_engine == "oceanbase":
+        OB = get_base_config("oceanbase", {})
+        docStoreConn = rag.utils.ob_conn.OBConnection()
     else:
         raise Exception(f"Not supported doc engine: {DOC_ENGINE}")
 

common/token_utils.py

@@ -35,6 +35,12 @@ def num_tokens_from_string(string: str) -> int:
         return 0
 
 def total_token_count_from_response(resp):
+    """
+    Extract token count from LLM response in various formats.
+
+    Handles None responses and different response structures from various LLM providers.
+    Returns 0 if token count cannot be determined.
+    """
     if resp is None:
         return 0
 
@@ -50,19 +56,19 @@ def total_token_count_from_response(resp):
         except Exception:
             pass
 
-    if 'usage' in resp and 'total_tokens' in resp['usage']:
+    if isinstance(resp, dict) and 'usage' in resp and 'total_tokens' in resp['usage']:
         try:
             return resp["usage"]["total_tokens"]
         except Exception:
             pass
 
-    if 'usage' in resp and 'input_tokens' in resp['usage'] and 'output_tokens' in resp['usage']:
+    if isinstance(resp, dict) and 'usage' in resp and 'input_tokens' in resp['usage'] and 'output_tokens' in resp['usage']:
         try:
             return resp["usage"]["input_tokens"] + resp["usage"]["output_tokens"]
         except Exception:
             pass
 
-    if 'meta' in resp and 'tokens' in resp['meta'] and 'input_tokens' in resp['meta']['tokens'] and 'output_tokens' in resp['meta']['tokens']:
+    if isinstance(resp, dict) and 'meta' in resp and 'tokens' in resp['meta'] and 'input_tokens' in resp['meta']['tokens'] and 'output_tokens' in resp['meta']['tokens']:
         try:
             return resp["meta"]["tokens"]["input_tokens"] + resp["meta"]["tokens"]["output_tokens"]
         except Exception:

conf/service_conf.yaml

@@ -28,6 +28,14 @@ os:
 infinity:
   uri: 'localhost:23817'
   db_name: 'default_db'
+oceanbase:
+  scheme: 'oceanbase' # set 'mysql' to create connection using mysql config
+  config:
+    db_name: 'test'
+    user: 'root@ragflow'
+    password: 'infini_rag_flow'
+    host: 'localhost'
+    port: 2881
 redis:
   db: 1
   password: 'infini_rag_flow'
@@ -139,5 +147,3 @@ user_default_llm:
 #  secret_id: 'tencent_secret_id'
 #  secret_key: 'tencent_secret_key'
 #  region: 'tencent_region'
-#  table_result_type: '1'
-#  markdown_image_response_type: '1'

deepdoc/parser/docling_parser.py

@@ -187,7 +187,7 @@ class DoclingParser(RAGFlowPdfParser):
                         bbox = _BBox(int(pn), bb[0], bb[1], bb[2], bb[3])
                 yield (DoclingContentType.EQUATION.value, text, bbox)
 
-    def _transfer_to_sections(self, doc) -> list[tuple[str, str]]:
+    def _transfer_to_sections(self, doc, parse_method: str) -> list[tuple[str, str]]:
         sections: list[tuple[str, str]] = []
         for typ, payload, bbox in self._iter_doc_items(doc):
             if typ == DoclingContentType.TEXT.value:
@@ -200,7 +200,12 @@ class DoclingParser(RAGFlowPdfParser):
                 continue
             
             tag = self._make_line_tag(bbox) if isinstance(bbox,_BBox) else ""
-            sections.append((section, tag))
+            if parse_method == "manual":
+                sections.append((section, typ, tag))
+            elif parse_method == "paper":
+                sections.append((section + tag, typ))
+            else:
+                sections.append((section, tag))
         return sections
 
     def cropout_docling_table(self, page_no: int, bbox: tuple[float, float, float, float], zoomin: int = 1):
@@ -282,7 +287,8 @@ class DoclingParser(RAGFlowPdfParser):
         output_dir: Optional[str] = None, 
         lang: Optional[str] = None,        
         method: str = "auto",             
-        delete_output: bool = True,       
+        delete_output: bool = True,
+        parse_method: str = "raw"     
     ):
 
         if not self.check_installation():
@@ -318,7 +324,7 @@ class DoclingParser(RAGFlowPdfParser):
         if callback:
             callback(0.7, f"[Docling] Parsed doc: {getattr(doc, 'num_pages', 'n/a')} pages")
 
-        sections = self._transfer_to_sections(doc)
+        sections = self._transfer_to_sections(doc, parse_method=parse_method)
         tables = self._transfer_to_tables(doc)
 
         if callback:

deepdoc/parser/mineru_parser.py

@@ -476,7 +476,7 @@ class MinerUParser(RAGFlowPdfParser):
                     item[key] = str((subdir / item[key]).resolve())
         return data
 
-    def _transfer_to_sections(self, outputs: list[dict[str, Any]]):
+    def _transfer_to_sections(self, outputs: list[dict[str, Any]], parse_method: str = None):
         sections = []
         for output in outputs:
             match output["type"]:
@@ -497,7 +497,11 @@ class MinerUParser(RAGFlowPdfParser):
                 case MinerUContentType.DISCARDED:
                     pass
 
-            if section:
+            if section and parse_method == "manual":
+                sections.append((section, output["type"], self._line_tag(output)))
+            elif section and parse_method == "paper":
+                sections.append((section + self._line_tag(output), output["type"]))
+            else:
                 sections.append((section, self._line_tag(output)))
         return sections
 
@@ -516,6 +520,7 @@ class MinerUParser(RAGFlowPdfParser):
         method: str = "auto",
         server_url: Optional[str] = None,
         delete_output: bool = True,
+        parse_method: str = "raw"
     ) -> tuple:
         import shutil
 
@@ -565,7 +570,8 @@ class MinerUParser(RAGFlowPdfParser):
             self.logger.info(f"[MinerU] Parsed {len(outputs)} blocks from PDF.")
             if callback:
                 callback(0.75, f"[MinerU] Parsed {len(outputs)} blocks from PDF.")
-            return self._transfer_to_sections(outputs), self._transfer_to_tables(outputs)
+                
+            return self._transfer_to_sections(outputs, parse_method), self._transfer_to_tables(outputs)
         finally:
             if temp_pdf and temp_pdf.exists():
                 try:

deepdoc/parser/pdf_parser.py

@@ -33,6 +33,8 @@ import xgboost as xgb
 from huggingface_hub import snapshot_download
 from PIL import Image
 from pypdf import PdfReader as pdf2_read
+from sklearn.cluster import KMeans
+from sklearn.metrics import silhouette_score
 
 from common.file_utils import get_project_base_directory
 from common.misc_utils import pip_install_torch
@@ -353,7 +355,6 @@ class RAGFlowPdfParser:
     def _assign_column(self, boxes, zoomin=3):
         if not boxes:
             return boxes
-
         if all("col_id" in b for b in boxes):
             return boxes
 
@@ -361,61 +362,80 @@ class RAGFlowPdfParser:
         for b in boxes:
             by_page[b["page_number"]].append(b)
 
-        page_info = {}  # pg -> dict(page_w, left_edge, cand_cols)
-        counter = Counter()
+        page_cols = {}
 
         for pg, bxs in by_page.items():
             if not bxs:
-                page_info[pg] = {"page_w": 1.0, "left_edge": 0.0, "cand": 1}
-                counter[1] += 1
+                page_cols[pg] = 1
                 continue
 
-            if hasattr(self, "page_images") and self.page_images and len(self.page_images) >= pg:
-                page_w = self.page_images[pg - 1].size[0] / max(1, zoomin)
-                left_edge = 0.0
-            else:
-                xs0 = [box["x0"] for box in bxs]
-                xs1 = [box["x1"] for box in bxs]
-                left_edge = float(min(xs0))
-                page_w = max(1.0, float(max(xs1) - left_edge))
+            x0s_raw = np.array([b["x0"] for b in bxs], dtype=float)
 
-            widths = [max(1.0, (box["x1"] - box["x0"])) for box in bxs]
-            median_w = float(np.median(widths)) if widths else 1.0
+            min_x0 = np.min(x0s_raw)
+            max_x1 = np.max([b["x1"] for b in bxs])
+            width = max_x1 - min_x0
 
-            raw_cols = int(page_w / max(1.0, median_w))
+            INDENT_TOL = width * 0.12
+            x0s = []
+            for x in x0s_raw:
+                if abs(x - min_x0) < INDENT_TOL:
+                    x0s.append([min_x0])
+                else:
+                    x0s.append([x])
+            x0s = np.array(x0s, dtype=float)
+            
+            max_try = min(4, len(bxs))
+            if max_try < 2:
+                max_try = 1
+            best_k = 1
+            best_score = -1
 
-            # cand = raw_cols if (raw_cols >= 2 and median_w < page_w / raw_cols * 0.8) else 1
-            cand = raw_cols
+            for k in range(1, max_try + 1):
+                km = KMeans(n_clusters=k, n_init="auto")
+                labels = km.fit_predict(x0s)
 
-            page_info[pg] = {"page_w": page_w, "left_edge": left_edge, "cand": cand}
-            counter[cand] += 1
+                centers = np.sort(km.cluster_centers_.flatten())
+                if len(centers) > 1:
+                    try:
+                        score = silhouette_score(x0s, labels)
+                    except ValueError:
+                        continue
+                else:
+                    score = 0
+                print(f"{k=},{score=}",flush=True)
+                if score > best_score:
+                    best_score = score
+                    best_k = k
 
-            logging.info(f"[Page {pg}] median_w={median_w:.2f}, page_w={page_w:.2f}, raw_cols={raw_cols}, cand={cand}")
+            page_cols[pg] = best_k
+            logging.info(f"[Page {pg}] best_score={best_score:.2f}, best_k={best_k}")
 
-        global_cols = counter.most_common(1)[0][0]
+
+        global_cols = Counter(page_cols.values()).most_common(1)[0][0]
         logging.info(f"Global column_num decided by majority: {global_cols}")
 
+
         for pg, bxs in by_page.items():
             if not bxs:
                 continue
+            k = page_cols[pg] 
+            if len(bxs) < k:
+                k = 1
+            x0s = np.array([[b["x0"]] for b in bxs], dtype=float)
+            km = KMeans(n_clusters=k, n_init="auto")
+            labels = km.fit_predict(x0s)
 
-            page_w = page_info[pg]["page_w"]
-            left_edge = page_info[pg]["left_edge"]
+            centers = km.cluster_centers_.flatten()
+            order = np.argsort(centers)
 
-            if global_cols == 1:
-                for box in bxs:
-                    box["col_id"] = 0
-                continue
+            remap = {orig: new for new, orig in enumerate(order)}
 
-            for box in bxs:
-                w = box["x1"] - box["x0"]
-                if w >= 0.8 * page_w:
-                    box["col_id"] = 0
-                    continue
-                cx = 0.5 * (box["x0"] + box["x1"])
-                norm_cx = (cx - left_edge) / page_w
-                norm_cx = max(0.0, min(norm_cx, 0.999999))
-                box["col_id"] = int(min(global_cols - 1, norm_cx * global_cols))
+            for b, lb in zip(bxs, labels):
+                b["col_id"] = remap[lb]
+            
+            grouped = defaultdict(list)
+            for b in bxs:
+                grouped[b["col_id"]].append(b)
 
         return boxes
 
@@ -1303,7 +1323,10 @@ class RAGFlowPdfParser:
 
         positions = []
         for ii, (pns, left, right, top, bottom) in enumerate(poss):
-            right = left + max_width
+            if 0 < ii < len(poss) - 1:
+                right = max(left + 10, right)
+            else:
+                right = left + max_width
             bottom *= ZM
             for pn in pns[1:]:
                 if 0 <= pn - 1 < page_count:

deepdoc/parser/tcadp_parser.py

@@ -192,12 +192,16 @@ class TencentCloudAPIClient:
 
 
 class TCADPParser(RAGFlowPdfParser):
-    def __init__(self, secret_id: str = None, secret_key: str = None, region: str = "ap-guangzhou"):
+    def __init__(self, secret_id: str = None, secret_key: str = None, region: str = "ap-guangzhou", 
+                 table_result_type: str = None, markdown_image_response_type: str = None):
         super().__init__()
         
         # First initialize logger
         self.logger = logging.getLogger(self.__class__.__name__)
         
+        # Log received parameters
+        self.logger.info(f"[TCADP] Initializing with parameters - table_result_type: {table_result_type}, markdown_image_response_type: {markdown_image_response_type}")
+        
         # Priority: read configuration from RAGFlow configuration system (service_conf.yaml)
         try:
             tcadp_parser = get_base_config("tcadp_config", {})
@@ -205,14 +209,30 @@ class TCADPParser(RAGFlowPdfParser):
                 self.secret_id = secret_id or tcadp_parser.get("secret_id")
                 self.secret_key = secret_key or tcadp_parser.get("secret_key")
                 self.region = region or tcadp_parser.get("region", "ap-guangzhou")
-                self.table_result_type = tcadp_parser.get("table_result_type", "1")
-                self.markdown_image_response_type = tcadp_parser.get("markdown_image_response_type", "1")
-                self.logger.info("[TCADP] Configuration read from service_conf.yaml")
+                # Set table_result_type and markdown_image_response_type from config or parameters
+                self.table_result_type = table_result_type if table_result_type is not None else tcadp_parser.get("table_result_type", "1")
+                self.markdown_image_response_type = markdown_image_response_type if markdown_image_response_type is not None else tcadp_parser.get("markdown_image_response_type", "1")
+                
             else:
                 self.logger.error("[TCADP] Please configure tcadp_config in service_conf.yaml first")
+                # If config file is empty, use provided parameters or defaults
+                self.secret_id = secret_id
+                self.secret_key = secret_key
+                self.region = region or "ap-guangzhou"
+                self.table_result_type = table_result_type if table_result_type is not None else "1"
+                self.markdown_image_response_type = markdown_image_response_type if markdown_image_response_type is not None else "1"
 
         except ImportError:
             self.logger.info("[TCADP] Configuration module import failed")
+            # If config file is not available, use provided parameters or defaults
+            self.secret_id = secret_id
+            self.secret_key = secret_key
+            self.region = region or "ap-guangzhou"
+            self.table_result_type = table_result_type if table_result_type is not None else "1"
+            self.markdown_image_response_type = markdown_image_response_type if markdown_image_response_type is not None else "1"
+
+        # Log final values
+        self.logger.info(f"[TCADP] Final values - table_result_type: {self.table_result_type}, markdown_image_response_type: {self.markdown_image_response_type}")
 
         if not self.secret_id or not self.secret_key:
             raise ValueError("[TCADP] Please set Tencent Cloud API keys, configure tcadp_config in service_conf.yaml")
@@ -400,6 +420,8 @@ class TCADPParser(RAGFlowPdfParser):
                         "TableResultType": self.table_result_type,
                         "MarkdownImageResponseType": self.markdown_image_response_type
                     }
+                    
+                    self.logger.info(f"[TCADP] API request config - TableResultType: {self.table_result_type}, MarkdownImageResponseType: {self.markdown_image_response_type}")
 
                     result = client.reconstruct_document_sse(
                         file_type=file_type, 

docker/.env

@@ -7,6 +7,7 @@
 # Available options:
 # - `elasticsearch` (default)
 # - `infinity` (https://github.com/infiniflow/infinity)
+# - `oceanbase` (https://github.com/oceanbase/oceanbase)
 # - `opensearch` (https://github.com/opensearch-project/OpenSearch)
 DOC_ENGINE=${DOC_ENGINE:-elasticsearch}
 
@@ -62,6 +63,27 @@ INFINITY_THRIFT_PORT=23817
 INFINITY_HTTP_PORT=23820
 INFINITY_PSQL_PORT=5432
 
+# The hostname where the OceanBase service is exposed
+OCEANBASE_HOST=oceanbase
+# The port used to expose the OceanBase service
+OCEANBASE_PORT=2881
+# The username for OceanBase
+OCEANBASE_USER=root@ragflow
+# The password for OceanBase
+OCEANBASE_PASSWORD=infini_rag_flow
+# The doc database of the OceanBase service to use
+OCEANBASE_DOC_DBNAME=ragflow_doc
+
+# OceanBase container configuration
+OB_CLUSTER_NAME=${OB_CLUSTER_NAME:-ragflow}
+OB_TENANT_NAME=${OB_TENANT_NAME:-ragflow}
+OB_SYS_PASSWORD=${OCEANBASE_PASSWORD:-infini_rag_flow}
+OB_TENANT_PASSWORD=${OCEANBASE_PASSWORD:-infini_rag_flow}
+OB_MEMORY_LIMIT=${OB_MEMORY_LIMIT:-10G}
+OB_SYSTEM_MEMORY=${OB_SYSTEM_MEMORY:-2G}
+OB_DATAFILE_SIZE=${OB_DATAFILE_SIZE:-20G}
+OB_LOG_DISK_SIZE=${OB_LOG_DISK_SIZE:-20G}
+
 # The password for MySQL.
 MYSQL_PASSWORD=infini_rag_flow
 # The hostname where the MySQL service is exposed
@@ -208,9 +230,16 @@ REGISTER_ENABLED=1
 # SANDBOX_MAX_MEMORY=256m # b, k, m, g
 # SANDBOX_TIMEOUT=10s # s, m, 1m30s
 
-# Enable DocLing and Mineru
+# Enable DocLing
 USE_DOCLING=false
+
+# Enable Mineru
 USE_MINERU=false
+MINERU_EXECUTABLE="$HOME/uv_tools/.venv/bin/mineru"
+MINERU_DELETE_OUTPUT=0   # keep output directory
+MINERU_BACKEND=pipeline  # or another backend you prefer
+
+
 
 # pptx support
 DOTNET_SYSTEM_GLOBALIZATION_INVARIANT=1

docker/README.md

@@ -138,6 +138,15 @@ The [.env](./.env) file contains important environment variables for Docker.
   - `password`: The password for MinIO.
   - `host`: The MinIO serving IP *and* port inside the Docker container. Defaults to `minio:9000`.
 
+- `oceanbase`
+  - `scheme`: The connection scheme. Set to `mysql` to use mysql config, or other values to use config below.
+  - `config`:
+    - `db_name`: The OceanBase database name.
+    - `user`: The username for OceanBase.
+    - `password`: The password for OceanBase.
+    - `host`: The hostname of the OceanBase service.
+    - `port`: The port of OceanBase.
+
 - `oss`
   - `access_key`: The access key ID used to authenticate requests to the OSS service.
   - `secret_key`: The secret access key used to authenticate requests to the OSS service.

docker/docker-compose-base.yml

@@ -72,7 +72,7 @@ services:
   infinity:
     profiles:
       - infinity
-    image: infiniflow/infinity:v0.6.5
+    image: infiniflow/infinity:v0.6.6
     volumes:
       - infinity_data:/var/infinity
       - ./infinity_conf.toml:/infinity_conf.toml
@@ -96,6 +96,31 @@ services:
       retries: 120
     restart: on-failure
 
+  oceanbase:
+    profiles:
+      - oceanbase
+    image: oceanbase/oceanbase-ce:4.4.1.0-100000032025101610
+    volumes:
+      - ./oceanbase/data:/root/ob
+      - ./oceanbase/conf:/root/.obd/cluster
+      - ./oceanbase/init.d:/root/boot/init.d
+    ports:
+      - ${OCEANBASE_PORT:-2881}:2881
+    env_file: .env
+    environment:
+      - MODE=normal
+      - OB_SERVER_IP=127.0.0.1
+    mem_limit: ${MEM_LIMIT}
+    healthcheck:
+      test: [ 'CMD-SHELL', 'obclient -h127.0.0.1 -P2881 -uroot@${OB_TENANT_NAME:-ragflow} -p${OB_TENANT_PASSWORD:-infini_rag_flow} -e "CREATE DATABASE IF NOT EXISTS ${OCEANBASE_DOC_DBNAME:-ragflow_doc};"' ]
+      interval: 10s
+      retries: 30
+      start_period: 30s
+      timeout: 10s
+    networks:
+      - ragflow
+    restart: on-failure
+
   sandbox-executor-manager:
     profiles:
       - sandbox
@@ -154,7 +179,7 @@ services:
 
   minio:
     image: quay.io/minio/minio:RELEASE.2025-06-13T11-33-47Z
-    command: server --console-address ":9001" /data
+    command: ["server", "--console-address", ":9001", "/data"]
     ports:
       - ${MINIO_PORT}:9000
       - ${MINIO_CONSOLE_PORT}:9001
@@ -176,7 +201,7 @@ services:
   redis:
     # swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/valkey/valkey:8
     image: valkey/valkey:8
-    command: redis-server --requirepass ${REDIS_PASSWORD} --maxmemory 128mb --maxmemory-policy allkeys-lru
+    command: ["redis-server", "--requirepass", "${REDIS_PASSWORD}", "--maxmemory", "128mb", "--maxmemory-policy", "allkeys-lru"]
     env_file: .env
     ports:
       - ${REDIS_PORT}:6379
@@ -256,6 +281,8 @@ volumes:
     driver: local
   infinity_data:
     driver: local
+  ob_data:
+    driver: local
   mysql_data:
     driver: local
   minio_data:

docker/infinity_conf.toml

@@ -1,5 +1,5 @@
 [general]
-version                  = "0.6.5"
+version                  = "0.6.6"
 time_zone                = "utc-8"
 
 [network]

docker/oceanbase/init.d/vec_memory.sql

@@ -0,0 +1 @@
+ALTER SYSTEM SET ob_vector_memory_limit_percentage = 30;

docker/service_conf.yaml.template

@@ -28,6 +28,14 @@ os:
 infinity:
   uri: '${INFINITY_HOST:-infinity}:23817'
   db_name: 'default_db'
+oceanbase:
+  scheme: 'oceanbase' # set 'mysql' to create connection using mysql config
+  config:
+    db_name: '${OCEANBASE_DOC_DBNAME:-test}'
+    user: '${OCEANBASE_USER:-root@ragflow}'
+    password: '${OCEANBASE_PASSWORD:-infini_rag_flow}'
+    host: '${OCEANBASE_HOST:-oceanbase}'
+    port: ${OCEANBASE_PORT:-2881}
 redis:
   db: 1
   password: '${REDIS_PASSWORD:-infini_rag_flow}'
@@ -142,5 +150,3 @@ user_default_llm:
 #   secret_id: '${TENCENT_SECRET_ID}'
 #   secret_key: '${TENCENT_SECRET_KEY}'
 #   region: '${TENCENT_REGION}'
-#   table_result_type: '1'
-#   markdown_image_response_type: '1'

docs/references/http_api_reference.md

@@ -2085,6 +2085,7 @@ curl --request POST \
           "dataset_ids": ["b2a62730759d11ef987d0242ac120004"],
           "document_ids": ["77df9ef4759a11ef8bdd0242ac120004"],
           "metadata_condition": {
+            "logic": "and",
             "conditions": [
               {
                 "name": "author",

helm/values.yaml

@@ -96,7 +96,7 @@ ragflow:
 infinity:
   image:
     repository: infiniflow/infinity
-    tag: v0.6.5
+    tag: v0.6.6
     pullPolicy: IfNotPresent
     pullSecrets: []
   storage:

pyproject.toml

@@ -49,7 +49,7 @@ dependencies = [
     "html-text==0.6.2",
     "httpx[socks]>=0.28.1,<0.29.0",
     "huggingface-hub>=0.25.0,<0.26.0",
-    "infinity-sdk==0.6.5",
+    "infinity-sdk==0.6.6",
     "infinity-emb>=0.0.66,<0.0.67",
     "itsdangerous==2.1.2",
     "json-repair==0.35.0",
@@ -149,6 +149,7 @@ dependencies = [
     "captcha>=0.7.1",
     "pip>=25.2",
     "pypandoc>=1.16",
+    "pyobvector==0.2.18",
 ]
 
 [dependency-groups]

rag/app/manual.py

@@ -213,6 +213,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
             lang = lang,
             callback = callback,
             pdf_cls = Pdf,
+            parse_method = "manual",
             **kwargs
         )
 
@@ -225,7 +226,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
             elif len(section) != 3:
                 raise ValueError(f"Unexpected section length: {len(section)} (value={section!r})")
 
-            txt, sec_id, poss = section
+            txt, layoutno, poss = section
             if isinstance(poss, str):
                 poss = pdf_parser.extract_positions(poss)
                 first = poss[0]          # tuple: ([pn], x1, x2, y1, y2)
@@ -235,7 +236,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
                     pn = pn[0]           # [pn] -> pn
                     poss[0] = (pn, *first[1:])
 
-            return (txt, sec_id, poss)
+            return (txt, layoutno, poss)
         
 
         sections = [_normalize_section(sec) for sec in sections]

rag/app/naive.py

@@ -59,6 +59,7 @@ def by_mineru(filename, binary=None, from_page=0, to_page=100000, lang="Chinese"
     mineru_executable = os.environ.get("MINERU_EXECUTABLE", "mineru")
     mineru_api = os.environ.get("MINERU_APISERVER", "http://host.docker.internal:9987")
     pdf_parser = MinerUParser(mineru_path=mineru_executable, mineru_api=mineru_api)
+    parse_method = kwargs.get("parse_method", "raw")
 
     if not pdf_parser.check_installation():
         callback(-1, "MinerU not found.")
@@ -72,12 +73,14 @@ def by_mineru(filename, binary=None, from_page=0, to_page=100000, lang="Chinese"
         backend=os.environ.get("MINERU_BACKEND", "pipeline"),
         server_url=os.environ.get("MINERU_SERVER_URL", ""),
         delete_output=bool(int(os.environ.get("MINERU_DELETE_OUTPUT", 1))),
+        parse_method=parse_method
     )
     return sections, tables, pdf_parser
 
 
 def by_docling(filename, binary=None, from_page=0, to_page=100000, lang="Chinese", callback=None, pdf_cls = None ,**kwargs):
     pdf_parser = DoclingParser()
+    parse_method = kwargs.get("parse_method", "raw")
 
     if not pdf_parser.check_installation():
         callback(-1, "Docling not found.")
@@ -89,6 +92,7 @@ def by_docling(filename, binary=None, from_page=0, to_page=100000, lang="Chinese
         callback=callback,
         output_dir=os.environ.get("MINERU_OUTPUT_DIR", ""),
         delete_output=bool(int(os.environ.get("MINERU_DELETE_OUTPUT", 1))),
+        parse_method=parse_method
     )
     return sections, tables, pdf_parser
 
@@ -116,7 +120,7 @@ def by_plaintext(filename, binary=None, from_page=0, to_page=100000, callback=No
     else:
         vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT, llm_name=kwargs.get("layout_recognizer", ""), lang=kwargs.get("lang", "Chinese"))
         pdf_parser = VisionParser(vision_model=vision_model, **kwargs)
-    
+
     sections, tables = pdf_parser(
         filename if not binary else binary,
         from_page=from_page,
@@ -504,7 +508,7 @@ class Markdown(MarkdownParser):
 
         return images if images else None
 
-    def __call__(self, filename, binary=None, separate_tables=True,delimiter=None):
+    def __call__(self, filename, binary=None, separate_tables=True, delimiter=None):
         if binary:
             encoding = find_codec(binary)
             txt = binary.decode(encoding, errors="ignore")
@@ -602,7 +606,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
         _SerializedRelationships.load_from_xml = load_from_xml_v2
         sections, tables = Docx()(filename, binary)
 
-        tables=vision_figure_parser_docx_wrapper(sections=sections,tbls=tables,callback=callback,**kwargs)
+        tables = vision_figure_parser_docx_wrapper(sections=sections, tbls=tables, callback=callback, **kwargs)
 
         res = tokenize_table(tables, doc, is_english)
         callback(0.8, "Finish parsing.")
@@ -653,18 +657,47 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
 
         if name in ["tcadp", "docling", "mineru"]:
             parser_config["chunk_token_num"] = 0
-        
+
         res = tokenize_table(tables, doc, is_english)
         callback(0.8, "Finish parsing.")
 
     elif re.search(r"\.(csv|xlsx?)$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
-        excel_parser = ExcelParser()
-        if parser_config.get("html4excel"):
-            sections = [(_, "") for _ in excel_parser.html(binary, 12) if _]
+
+        # Check if tcadp_parser is selected for spreadsheet files
+        layout_recognizer = parser_config.get("layout_recognize", "DeepDOC")
+        if layout_recognizer == "TCADP Parser":
+            table_result_type = parser_config.get("table_result_type", "1")
+            markdown_image_response_type = parser_config.get("markdown_image_response_type", "1")
+            tcadp_parser = TCADPParser(
+                table_result_type=table_result_type,
+                markdown_image_response_type=markdown_image_response_type
+            )
+            if not tcadp_parser.check_installation():
+                callback(-1, "TCADP parser not available. Please check Tencent Cloud API configuration.")
+                return res
+
+            # Determine file type based on extension
+            file_type = "XLSX" if re.search(r"\.xlsx?$", filename, re.IGNORECASE) else "CSV"
+
+            sections, tables = tcadp_parser.parse_pdf(
+                filepath=filename,
+                binary=binary,
+                callback=callback,
+                output_dir=os.environ.get("TCADP_OUTPUT_DIR", ""),
+                file_type=file_type
+            )
+            parser_config["chunk_token_num"] = 0
+            res = tokenize_table(tables, doc, is_english)
+            callback(0.8, "Finish parsing.")
         else:
-            sections = [(_, "") for _ in excel_parser(binary) if _]
-        parser_config["chunk_token_num"] = 12800
+            # Default DeepDOC parser
+            excel_parser = ExcelParser()
+            if parser_config.get("html4excel"):
+                sections = [(_, "") for _ in excel_parser.html(binary, 12) if _]
+            else:
+                sections = [(_, "") for _ in excel_parser(binary) if _]
+            parser_config["chunk_token_num"] = 12800
 
     elif re.search(r"\.(txt|py|js|java|c|cpp|h|php|go|ts|sh|cs|kt|sql)$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
@@ -676,7 +709,7 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
     elif re.search(r"\.(md|markdown)$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")
         markdown_parser = Markdown(int(parser_config.get("chunk_token_num", 128)))
-        sections, tables = markdown_parser(filename, binary, separate_tables=False,delimiter=parser_config.get("delimiter", "\n!?;。；！？"))
+        sections, tables = markdown_parser(filename, binary, separate_tables=False, delimiter=parser_config.get("delimiter", "\n!?;。；！？"))
 
         try:
             vision_model = LLMBundle(kwargs["tenant_id"], LLMType.IMAGE2TEXT)

rag/app/paper.py

@@ -21,8 +21,10 @@ import re
 from deepdoc.parser.figure_parser import vision_figure_parser_pdf_wrapper
 from common.constants import ParserType
 from rag.nlp import rag_tokenizer, tokenize, tokenize_table, add_positions, bullets_category, title_frequency, tokenize_chunks
-from deepdoc.parser import PdfParser, PlainParser
+from deepdoc.parser import PdfParser
 import numpy as np
+from rag.app.naive import by_plaintext, PARSERS
+
 
 class Pdf(PdfParser):
     def __init__(self):
@@ -147,19 +149,40 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
         "parser_config", {
             "chunk_token_num": 512, "delimiter": "\n!?。；！？", "layout_recognize": "DeepDOC"})
     if re.search(r"\.pdf$", filename, re.IGNORECASE):
-        if parser_config.get("layout_recognize", "DeepDOC") == "Plain Text":
-            pdf_parser = PlainParser()
+        layout_recognizer = parser_config.get("layout_recognize", "DeepDOC")
+        
+        if isinstance(layout_recognizer, bool):
+            layout_recognizer = "DeepDOC" if layout_recognizer else "Plain Text"
+
+        name = layout_recognizer.strip().lower()
+        pdf_parser = PARSERS.get(name, by_plaintext)
+        callback(0.1, "Start to parse.")
+
+        if name == "deepdoc":
+            pdf_parser = Pdf()
+            paper = pdf_parser(filename if not binary else binary,
+                               from_page=from_page, to_page=to_page, callback=callback)
+        else:
+            sections, tables, pdf_parser = pdf_parser(
+                filename=filename,
+                binary=binary,
+                from_page=from_page,
+                to_page=to_page,
+                lang=lang,
+                callback=callback,
+                pdf_cls=Pdf,
+                parse_method="paper",
+                **kwargs
+            )
+
             paper = {
                 "title": filename,
                 "authors": " ",
                 "abstract": "",
-                "sections": pdf_parser(filename if not binary else binary, from_page=from_page, to_page=to_page)[0],
-                "tables": []
+                "sections": sections,
+                "tables": tables
             }
-        else:
-            pdf_parser = Pdf()
-            paper = pdf_parser(filename if not binary else binary,
-                               from_page=from_page, to_page=to_page, callback=callback)
+
         tbls=paper["tables"]
         tbls=vision_figure_parser_pdf_wrapper(tbls=tbls,callback=callback,**kwargs)
         paper["tables"] = tbls

rag/flow/parser/parser.py

@@ -16,6 +16,7 @@ import io
 import json
 import os
 import random
+import re
 from functools import partial
 
 import trio
@@ -83,6 +84,7 @@ class ParserParam(ProcessParamBase):
                 "output_format": "json",
             },
             "spreadsheet": {
+                "parse_method": "deepdoc",  # deepdoc/tcadp_parser
                 "output_format": "html",
                 "suffix": [
                     "xls",
@@ -102,8 +104,10 @@ class ParserParam(ProcessParamBase):
                 "output_format": "json",
             },
             "slides": {
+                "parse_method": "deepdoc",  # deepdoc/tcadp_parser
                 "suffix": [
                     "pptx",
+                    "ppt"
                 ],
                 "output_format": "json",
             },
@@ -245,7 +249,12 @@ class Parser(ProcessBase):
                 bboxes.append(box)
         elif conf.get("parse_method").lower() == "tcadp parser":
             # ADP is a document parsing tool using Tencent Cloud API
-            tcadp_parser = TCADPParser()
+            table_result_type = conf.get("table_result_type", "1")
+            markdown_image_response_type = conf.get("markdown_image_response_type", "1")
+            tcadp_parser = TCADPParser(
+                table_result_type=table_result_type,
+                markdown_image_response_type=markdown_image_response_type
+            )
             sections, _ = tcadp_parser.parse_pdf(
                 filepath=name,
                 binary=blob,
@@ -301,14 +310,86 @@ class Parser(ProcessBase):
         self.callback(random.randint(1, 5) / 100.0, "Start to work on a Spreadsheet.")
         conf = self._param.setups["spreadsheet"]
         self.set_output("output_format", conf["output_format"])
-        spreadsheet_parser = ExcelParser()
-        if conf.get("output_format") == "html":
-            htmls = spreadsheet_parser.html(blob, 1000000000)
-            self.set_output("html", htmls[0])
-        elif conf.get("output_format") == "json":
-            self.set_output("json", [{"text": txt} for txt in spreadsheet_parser(blob) if txt])
-        elif conf.get("output_format") == "markdown":
-            self.set_output("markdown", spreadsheet_parser.markdown(blob))
+
+        parse_method = conf.get("parse_method", "deepdoc")
+
+        # Handle TCADP parser
+        if parse_method.lower() == "tcadp parser":
+            table_result_type = conf.get("table_result_type", "1")
+            markdown_image_response_type = conf.get("markdown_image_response_type", "1")
+            tcadp_parser = TCADPParser(
+                table_result_type=table_result_type,
+                markdown_image_response_type=markdown_image_response_type
+            )
+            if not tcadp_parser.check_installation():
+                raise RuntimeError("TCADP parser not available. Please check Tencent Cloud API configuration.")
+
+            # Determine file type based on extension
+            if re.search(r"\.xlsx?$", name, re.IGNORECASE):
+                file_type = "XLSX"
+            else:
+                file_type = "CSV"
+
+            self.callback(0.2, f"Using TCADP parser for {file_type} file.")
+            sections, tables = tcadp_parser.parse_pdf(
+                filepath=name,
+                binary=blob,
+                callback=self.callback,
+                file_type=file_type,
+                file_start_page=1,
+                file_end_page=1000
+            )
+
+            # Process TCADP parser output based on configured output_format
+            output_format = conf.get("output_format", "html")
+
+            if output_format == "html":
+                # For HTML output, combine sections and tables into HTML
+                html_content = ""
+                for section, position_tag in sections:
+                    if section:
+                        html_content += section + "\n"
+                for table in tables:
+                    if table:
+                        html_content += table + "\n"
+
+                self.set_output("html", html_content)
+
+            elif output_format == "json":
+                # For JSON output, create a list of text items
+                result = []
+                # Add sections as text
+                for section, position_tag in sections:
+                    if section:
+                        result.append({"text": section})
+                # Add tables as text
+                for table in tables:
+                    if table:
+                        result.append({"text": table})
+
+                self.set_output("json", result)
+
+            elif output_format == "markdown":
+                # For markdown output, combine into markdown
+                md_content = ""
+                for section, position_tag in sections:
+                    if section:
+                        md_content += section + "\n\n"
+                for table in tables:
+                    if table:
+                        md_content += table + "\n\n"
+
+                self.set_output("markdown", md_content)
+        else:
+            # Default DeepDOC parser
+            spreadsheet_parser = ExcelParser()
+            if conf.get("output_format") == "html":
+                htmls = spreadsheet_parser.html(blob, 1000000000)
+                self.set_output("html", htmls[0])
+            elif conf.get("output_format") == "json":
+                self.set_output("json", [{"text": txt} for txt in spreadsheet_parser(blob) if txt])
+            elif conf.get("output_format") == "markdown":
+                self.set_output("markdown", spreadsheet_parser.markdown(blob))
 
     def _word(self, name, blob):
         self.callback(random.randint(1, 5) / 100.0, "Start to work on a Word Processor Document")
@@ -326,22 +407,69 @@ class Parser(ProcessBase):
             self.set_output("markdown", markdown_text)
 
     def _slides(self, name, blob):
-        from deepdoc.parser.ppt_parser import RAGFlowPptParser as ppt_parser
-
         self.callback(random.randint(1, 5) / 100.0, "Start to work on a PowerPoint Document")
 
         conf = self._param.setups["slides"]
         self.set_output("output_format", conf["output_format"])
 
-        ppt_parser = ppt_parser()
-        txts = ppt_parser(blob, 0, 100000, None)
+        parse_method = conf.get("parse_method", "deepdoc")
 
-        sections = [{"text": section} for section in txts if section.strip()]
+        # Handle TCADP parser
+        if parse_method.lower() == "tcadp parser":
+            table_result_type = conf.get("table_result_type", "1")
+            markdown_image_response_type = conf.get("markdown_image_response_type", "1")
+            tcadp_parser = TCADPParser(
+                table_result_type=table_result_type,
+                markdown_image_response_type=markdown_image_response_type
+            )
+            if not tcadp_parser.check_installation():
+                raise RuntimeError("TCADP parser not available. Please check Tencent Cloud API configuration.")
 
-        # json
-        assert conf.get("output_format") == "json", "have to be json for ppt"
-        if conf.get("output_format") == "json":
-            self.set_output("json", sections)
+            # Determine file type based on extension
+            if re.search(r"\.pptx?$", name, re.IGNORECASE):
+                file_type = "PPTX"
+            else:
+                file_type = "PPT"
+
+            self.callback(0.2, f"Using TCADP parser for {file_type} file.")
+
+            sections, tables = tcadp_parser.parse_pdf(
+                filepath=name,
+                binary=blob,
+                callback=self.callback,
+                file_type=file_type,
+                file_start_page=1,
+                file_end_page=1000
+            )
+
+            # Process TCADP parser output - PPT only supports json format
+            output_format = conf.get("output_format", "json")
+            if output_format == "json":
+                # For JSON output, create a list of text items
+                result = []
+                # Add sections as text
+                for section, position_tag in sections:
+                    if section:
+                        result.append({"text": section})
+                # Add tables as text
+                for table in tables:
+                    if table:
+                        result.append({"text": table})
+
+                self.set_output("json", result)
+        else:
+            # Default DeepDOC parser (supports .pptx format)
+            from deepdoc.parser.ppt_parser import RAGFlowPptParser as ppt_parser
+
+            ppt_parser = ppt_parser()
+            txts = ppt_parser(blob, 0, 100000, None)
+
+            sections = [{"text": section} for section in txts if section.strip()]
+
+            # json
+            assert conf.get("output_format") == "json", "have to be json for ppt"
+            if conf.get("output_format") == "json":
+                self.set_output("json", sections)
 
     def _markdown(self, name, blob):
         from functools import reduce
@@ -579,6 +707,7 @@ class Parser(ProcessBase):
             "video": self._video,
             "email": self._email,
         }
+
         try:
             from_upstream = ParserFromUpstream.model_validate(kwargs)
         except Exception as e:

rag/llm/rerank_model.py

@@ -234,7 +234,11 @@ class CoHereRerank(Base):
     def __init__(self, key, model_name, base_url=None):
         from cohere import Client
 
-        self.client = Client(api_key=key, base_url=base_url)
+        # Only pass base_url if it's a non-empty string, otherwise use default Cohere API endpoint
+        client_kwargs = {"api_key": key}
+        if base_url and base_url.strip():
+            client_kwargs["base_url"] = base_url
+        self.client = Client(**client_kwargs)
         self.model_name = model_name.split("___")[0]
 
     def similarity(self, query: str, texts: list):

rag/nlp/query.py

@@ -83,6 +83,7 @@ class FulltextQueryer:
         return txt
 
     def question(self, txt, tbl="qa", min_match: float = 0.6):
+        original_query = txt
         txt = FulltextQueryer.add_space_between_eng_zh(txt)
         txt = re.sub(
             r"[ :|\r\n\t,，。？?/`!！&^%%()\[\]{}<>]+",
@@ -127,7 +128,7 @@ class FulltextQueryer:
                 q.append(txt)
             query = " ".join(q)
             return MatchTextExpr(
-                self.query_fields, query, 100
+                self.query_fields, query, 100, {"original_query": original_query}
             ), keywords
 
         def need_fine_grained_tokenize(tk):
@@ -212,7 +213,7 @@ class FulltextQueryer:
             if not query:
                 query = otxt
             return MatchTextExpr(
-                self.query_fields, query, 100, {"minimum_should_match": min_match}
+                self.query_fields, query, 100, {"minimum_should_match": min_match, "original_query": original_query}
             ), keywords
         return None, keywords
 
@@ -259,6 +260,7 @@ class FulltextQueryer:
             content_tks = [c.strip() for c in content_tks.strip() if c.strip()]
         tks_w = self.tw.weights(content_tks, preprocess=False)
 
+        origin_keywords = keywords.copy()
         keywords = [f'"{k.strip()}"' for k in keywords]
         for tk, w in sorted(tks_w, key=lambda x: x[1] * -1)[:keywords_topn]:
             tk_syns = self.syn.lookup(tk)
@@ -274,4 +276,4 @@ class FulltextQueryer:
                 keywords.append(f"{tk}^{w}")
 
         return MatchTextExpr(self.query_fields, " ".join(keywords), 100,
-                             {"minimum_should_match": min(3, len(keywords) // 10)})
+                             {"minimum_should_match": min(3, len(keywords) / 10), "original_query": " ".join(origin_keywords)})

rag/nlp/search.py

@@ -355,75 +355,102 @@ class Dealer:
                                            rag_tokenizer.tokenize(ans).split(),
                                            rag_tokenizer.tokenize(inst).split())
 
-    def retrieval(self, question, embd_mdl, tenant_ids, kb_ids, page, page_size, similarity_threshold=0.2,
-                  vector_similarity_weight=0.3, top=1024, doc_ids=None, aggs=True,
-                  rerank_mdl=None, highlight=False,
-                  rank_feature: dict | None = {PAGERANK_FLD: 10}):
+    def retrieval(
+        self,
+        question,
+        embd_mdl,
+        tenant_ids,
+        kb_ids,
+        page,
+        page_size,
+        similarity_threshold=0.2,
+        vector_similarity_weight=0.3,
+        top=1024,
+        doc_ids=None,
+        aggs=True,
+        rerank_mdl=None,
+        highlight=False,
+        rank_feature: dict | None = {PAGERANK_FLD: 10},
+    ):
         ranks = {"total": 0, "chunks": [], "doc_aggs": {}}
         if not question:
             return ranks
 
         # Ensure RERANK_LIMIT is multiple of page_size
-        RERANK_LIMIT = math.ceil(64/page_size) * page_size if page_size>1 else 1
-        req = {"kb_ids": kb_ids, "doc_ids": doc_ids, "page": math.ceil(page_size*page/RERANK_LIMIT), "size": RERANK_LIMIT,
-               "question": question, "vector": True, "topk": top,
-               "similarity": similarity_threshold,
-               "available_int": 1}
-
+        RERANK_LIMIT = math.ceil(64 / page_size) * page_size if page_size > 1 else 1
+        req = {
+            "kb_ids": kb_ids,
+            "doc_ids": doc_ids,
+            "page": math.ceil(page_size * page / RERANK_LIMIT),
+            "size": RERANK_LIMIT,
+            "question": question,
+            "vector": True,
+            "topk": top,
+            "similarity": similarity_threshold,
+            "available_int": 1,
+        }
 
         if isinstance(tenant_ids, str):
             tenant_ids = tenant_ids.split(",")
 
-        sres = self.search(req, [index_name(tid) for tid in tenant_ids],
-                           kb_ids, embd_mdl, highlight, rank_feature=rank_feature)
+        sres = self.search(req, [index_name(tid) for tid in tenant_ids], kb_ids, embd_mdl, highlight, rank_feature=rank_feature)
 
         if rerank_mdl and sres.total > 0:
-            sim, tsim, vsim = self.rerank_by_model(rerank_mdl,
-                                                   sres, question, 1 - vector_similarity_weight,
-                                                   vector_similarity_weight,
-                                                   rank_feature=rank_feature)
+            sim, tsim, vsim = self.rerank_by_model(
+                rerank_mdl,
+                sres,
+                question,
+                1 - vector_similarity_weight,
+                vector_similarity_weight,
+                rank_feature=rank_feature,
+            )
         else:
-            lower_case_doc_engine = os.getenv('DOC_ENGINE', 'elasticsearch')
-            if lower_case_doc_engine in ["elasticsearch","opensearch"]:
+            lower_case_doc_engine = os.getenv("DOC_ENGINE", "elasticsearch")
+            if lower_case_doc_engine in ["elasticsearch", "opensearch"]:
                 # ElasticSearch doesn't normalize each way score before fusion.
                 sim, tsim, vsim = self.rerank(
-                    sres, question, 1 - vector_similarity_weight, vector_similarity_weight,
-                    rank_feature=rank_feature)
+                    sres,
+                    question,
+                    1 - vector_similarity_weight,
+                    vector_similarity_weight,
+                    rank_feature=rank_feature,
+                )
             else:
                 # Don't need rerank here since Infinity normalizes each way score before fusion.
                 sim = [sres.field[id].get("_score", 0.0) for id in sres.ids]
-                sim = [s if s is not None else 0. for s in sim]
+                sim = [s if s is not None else 0.0 for s in sim]
                 tsim = sim
                 vsim = sim
-        # Already paginated in search function
-        max_pages = RERANK_LIMIT // page_size
-        page_index = (page % max_pages) - 1
-        begin = max(page_index * page_size, 0)
-        sim = sim[begin : begin + page_size]
+
         sim_np = np.array(sim, dtype=np.float64)
-        idx = np.argsort(sim_np * -1)
+        if sim_np.size == 0:
+            return ranks
+
+        sorted_idx = np.argsort(sim_np * -1)
+
+        valid_idx = [int(i) for i in sorted_idx if sim_np[i] >= similarity_threshold]
+        filtered_count = len(valid_idx)
+        ranks["total"] = int(filtered_count)
+
+        if filtered_count == 0:
+            return ranks
+
+        max_pages = max(RERANK_LIMIT // max(page_size, 1), 1)
+        page_index = (page - 1) % max_pages
+        begin = page_index * page_size
+        end = begin + page_size
+        page_idx = valid_idx[begin:end]
+
         dim = len(sres.query_vector)
         vector_column = f"q_{dim}_vec"
         zero_vector = [0.0] * dim
-        filtered_count = (sim_np >= similarity_threshold).sum()
-        ranks["total"] = int(filtered_count) # Convert from np.int64 to Python int otherwise JSON serializable error
-        for i in idx:
-            if np.float64(sim[i]) < similarity_threshold:
-                break
 
+        for i in page_idx:
             id = sres.ids[i]
             chunk = sres.field[id]
             dnm = chunk.get("docnm_kwd", "")
             did = chunk.get("doc_id", "")
 
-            if len(ranks["chunks"]) >= page_size:
-                if aggs:
-                    if dnm not in ranks["doc_aggs"]:
-                        ranks["doc_aggs"][dnm] = {"doc_id": did, "count": 0}
-                    ranks["doc_aggs"][dnm]["count"] += 1
-                    continue
-                break
-
             position_int = chunk.get("position_int", [])
             d = {
                 "chunk_id": id,
@@ -434,12 +461,12 @@ class Dealer:
                 "kb_id": chunk["kb_id"],
                 "important_kwd": chunk.get("important_kwd", []),
                 "image_id": chunk.get("img_id", ""),
-                "similarity": sim[i],
-                "vector_similarity": vsim[i],
-                "term_similarity": tsim[i],
+                "similarity": float(sim_np[i]),
+                "vector_similarity": float(vsim[i]),
+                "term_similarity": float(tsim[i]),
                 "vector": chunk.get(vector_column, zero_vector),
                 "positions": position_int,
-                "doc_type_kwd": chunk.get("doc_type_kwd", "")
+                "doc_type_kwd": chunk.get("doc_type_kwd", ""),
             }
             if highlight and sres.highlight:
                 if id in sres.highlight:
@@ -447,15 +474,30 @@ class Dealer:
                 else:
                     d["highlight"] = d["content_with_weight"]
             ranks["chunks"].append(d)
-            if dnm not in ranks["doc_aggs"]:
-                ranks["doc_aggs"][dnm] = {"doc_id": did, "count": 0}
-            ranks["doc_aggs"][dnm]["count"] += 1
-        ranks["doc_aggs"] = [{"doc_name": k,
-                              "doc_id": v["doc_id"],
-                              "count": v["count"]} for k,
-                                                       v in sorted(ranks["doc_aggs"].items(),
-                                                                   key=lambda x: x[1]["count"] * -1)]
-        ranks["chunks"] = ranks["chunks"][:page_size]
+
+        if aggs:
+            for i in valid_idx:
+                id = sres.ids[i]
+                chunk = sres.field[id]
+                dnm = chunk.get("docnm_kwd", "")
+                did = chunk.get("doc_id", "")
+                if dnm not in ranks["doc_aggs"]:
+                    ranks["doc_aggs"][dnm] = {"doc_id": did, "count": 0}
+                ranks["doc_aggs"][dnm]["count"] += 1
+
+            ranks["doc_aggs"] = [
+                {
+                    "doc_name": k,
+                    "doc_id": v["doc_id"],
+                    "count": v["count"],
+                }
+                for k, v in sorted(
+                    ranks["doc_aggs"].items(),
+                    key=lambda x: x[1]["count"] * -1,
+                )
+            ]
+        else:
+            ranks["doc_aggs"] = []
 
         return ranks
 
@@ -564,7 +606,7 @@ class Dealer:
         ids = relevant_chunks_with_toc(query, toc, chat_mdl, topn*2)
         if not ids:
             return chunks
-        
+
         vector_size = 1024
         id2idx = {ck["chunk_id"]: i for i, ck in enumerate(chunks)}
         for cid, sim in ids:

rag/prompts/generator.py

@@ -429,7 +429,7 @@ def rank_memories(chat_mdl, goal:str, sub_goal:str, tool_call_summaries: list[st
     return re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
 
 
-def gen_meta_filter(chat_mdl, meta_data:dict, query: str) -> list:
+def gen_meta_filter(chat_mdl, meta_data:dict, query: str) -> dict:
     sys_prompt = PROMPT_JINJA_ENV.from_string(META_FILTER).render(
         current_date=datetime.datetime.today().strftime('%Y-%m-%d'),
         metadata_keys=json.dumps(meta_data),
@@ -440,11 +440,13 @@ def gen_meta_filter(chat_mdl, meta_data:dict, query: str) -> list:
     ans = re.sub(r"(^.*</think>|```json\n|```\n*$)", "", ans, flags=re.DOTALL)
     try:
         ans = json_repair.loads(ans)
-        assert isinstance(ans, list), ans
+        assert isinstance(ans, dict), ans
+        assert "conditions" in ans and isinstance(ans["conditions"], list), ans
         return ans
     except Exception:
         logging.exception(f"Loading json failure: {ans}")
-    return []
+
+    return {"conditions": []}
 
 
 def gen_json(system_prompt:str, user_prompt:str, chat_mdl, gen_conf = None):

rag/prompts/meta_filter.md

@@ -9,11 +9,13 @@ You are a metadata filtering condition generator. Analyze the user's question an
      }
 
 2. **Output Requirements**:
-   - Always output a JSON array of filter objects
-   - Each object must have:
+   - Always output a JSON dictionary with only 2 keys: 'conditions'(filter objects) and 'logic' between the conditions ('and' or 'or').
+   - Each filter object in conditions must have:
         "key": (metadata attribute name),
         "value": (string value to compare),
         "op": (operator from allowed list)
+   - Logic between all the conditions: 'and'(Intersection of results for each condition) / 'or' (union of results for all conditions)
+
 
 3. **Operator Guide**:
    - Use these operators only: ["contains", "not contains", "start with", "end with", "empty", "not empty", "=", "≠", ">", "<", "≥", "≤"]
@@ -32,22 +34,97 @@ You are a metadata filtering condition generator. Analyze the user's question an
         - Attribute doesn't exist in metadata
         - Value has no match in metadata
 
-5. **Example**:
+5. **Example A**:
    - User query: "上市日期七月份的有哪些商品，不要蓝色的"
    - Metadata: { "color": {...}, "listing_date": {...} }
    - Output: 
-        [
+   {
+        "logic": "and",
+        "conditions": [
           {"key": "listing_date", "value": "2025-07-01", "op": "≥"},
           {"key": "listing_date", "value": "2025-08-01", "op": "<"},
           {"key": "color", "value": "blue", "op": "≠"}
         ]
+   }
 
-6. **Final Output**:
-   - ONLY output valid JSON array
+6. **Example B**:
+   - User query: "Both blue and red are acceptable."
+   - Metadata: { "color": {...}, "listing_date": {...} }
+   - Output: 
+   {
+        "logic": "or",
+        "conditions": [
+          {"key": "color", "value": "blue", "op": "="},
+          {"key": "color", "value": "red", "op": "="}
+        ]
+   }
+
+7. **Final Output**:
+   - ONLY output valid JSON dictionary
    - NO additional text/explanations
+   - Json schema is as following:
+```json
+{
+  "type": "object",
+  "properties": {
+    "logic": {
+      "type": "string",
+      "description": "Logic relationship between all the conditions, the default is 'and'.",
+      "enum": [
+        "and",
+        "or"
+      ]
+    },
+    "conditions": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "properties": {
+          "key": {
+            "type": "string",
+            "description": "Metadata attribute name."
+          },
+          "value": {
+            "type": "string",
+            "description": "Value to compare."
+          },
+          "op": {
+            "type": "string",
+            "description": "Operator from allowed list.",
+            "enum": [
+              "contains",
+              "not contains",
+              "start with",
+              "end with",
+              "empty",
+              "not empty",
+              "=",
+              "≠",
+              ">",
+              "<",
+              "≥",
+              "≤"
+            ]
+          }
+        },
+        "required": [
+          "key",
+          "value",
+          "op"
+        ],
+        "additionalProperties": false
+      }
+    }
+  },
+  "required": [
+    "conditions"
+  ],
+  "additionalProperties": false
+}
+```
 
 **Current Task**:
-- Today's date: {{current_date}}
-- Available metadata keys: {{metadata_keys}}
-- User query: "{{user_question}}"
+- Today's date: {{ current_date }}
+- Available metadata keys: {{ metadata_keys }}
+- User query: "{{ user_question }}"
 

sdk/python/ragflow_sdk/modules/document.py

@@ -69,7 +69,7 @@ class Document(Base):
             response = res.json()
             actual_keys = set(response.keys())
             if actual_keys == error_keys:
-                raise Exception(res.get("message"))
+                raise Exception(response.get("message"))
             else:
                 return res.content
         except json.JSONDecodeError:

sdk/python/ragflow_sdk/modules/session.py

@@ -80,6 +80,7 @@ class Session(Base):
         
 
     def _structure_answer(self, json_data):
+        answer = ""
         if self.__session_type == "agent":
            answer = json_data["data"]["content"]
         elif self.__session_type == "chat":

web/package-lock.json

@@ -66,6 +66,7 @@
         "input-otp": "^1.4.1",
         "js-base64": "^3.7.5",
         "jsencrypt": "^3.3.2",
+        "jsoneditor": "^10.4.2",
         "lexical": "^0.23.1",
         "lodash": "^4.17.21",
         "lucide-react": "^0.546.0",
@@ -8998,6 +8999,12 @@
         "@sinonjs/commons": "^3.0.0"
       }
     },
+    "node_modules/@sphinxxxx/color-conversion": {
+      "version": "2.2.2",
+      "resolved": "https://registry.npmmirror.com/@sphinxxxx/color-conversion/-/color-conversion-2.2.2.tgz",
+      "integrity": "sha512-XExJS3cLqgrmNBIP3bBw6+1oQ1ksGjFh0+oClDKFYpCCqx/hlqwWO5KO/S63fzUo67SxI9dMrF0y5T/Ey7h8Zw==",
+      "license": "ISC"
+    },
     "node_modules/@storybook/addon-docs": {
       "version": "9.1.4",
       "resolved": "https://registry.npmmirror.com/@storybook/addon-docs/-/addon-docs-9.1.4.tgz",
@@ -12962,6 +12969,12 @@
         "node": ">= 0.6"
       }
     },
+    "node_modules/ace-builds": {
+      "version": "1.43.4",
+      "resolved": "https://registry.npmmirror.com/ace-builds/-/ace-builds-1.43.4.tgz",
+      "integrity": "sha512-8hAxVfo2ImICd69BWlZwZlxe9rxDGDjuUhh+WeWgGDvfBCE+r3lkynkQvIovDz4jcMi8O7bsEaFygaDT+h9sBA==",
+      "license": "BSD-3-Clause"
+    },
     "node_modules/acorn": {
       "version": "8.15.0",
       "resolved": "https://registry.npmmirror.com/acorn/-/acorn-8.15.0.tgz",
@@ -21894,6 +21907,12 @@
         "@pkgjs/parseargs": "^0.11.0"
       }
     },
+    "node_modules/javascript-natural-sort": {
+      "version": "0.7.1",
+      "resolved": "https://registry.npmmirror.com/javascript-natural-sort/-/javascript-natural-sort-0.7.1.tgz",
+      "integrity": "sha512-nO6jcEfZWQXDhOiBtG2KvKyEptz7RVbpGP4vTD2hLBdmNQSsCiicO2Ioinv6UI4y9ukqnBpy+XZ9H6uLNgJTlw==",
+      "license": "MIT"
+    },
     "node_modules/javascript-stringify": {
       "version": "2.1.0",
       "resolved": "https://registry.npmmirror.com/javascript-stringify/-/javascript-stringify-2.1.0.tgz",
@@ -24253,6 +24272,15 @@
         "jiti": "bin/jiti.js"
       }
     },
+    "node_modules/jmespath": {
+      "version": "0.16.0",
+      "resolved": "https://registry.npmmirror.com/jmespath/-/jmespath-0.16.0.tgz",
+      "integrity": "sha512-9FzQjJ7MATs1tSpnco1K6ayiYE3figslrXA72G2HQ/n76RzvYlofyi5QM+iX4YRs/pu3yzxlVQSST23+dMDknw==",
+      "license": "Apache-2.0",
+      "engines": {
+        "node": ">= 0.6.0"
+      }
+    },
     "node_modules/js-base64": {
       "version": "3.7.5",
       "resolved": "https://registry.npmmirror.com/js-base64/-/js-base64-3.7.5.tgz",
@@ -24357,6 +24385,12 @@
       "integrity": "sha512-NM8/P9n3XjXhIZn1lLhkFaACTOURQXjWhV4BA/RnOv8xvgqtqpAX9IO4mRQxSx1Rlo4tqzeqb0sOlruaOy3dug==",
       "license": "MIT"
     },
+    "node_modules/json-source-map": {
+      "version": "0.6.1",
+      "resolved": "https://registry.npmmirror.com/json-source-map/-/json-source-map-0.6.1.tgz",
+      "integrity": "sha512-1QoztHPsMQqhDq0hlXY5ZqcEdUzxQEIxgFkKl4WUp2pgShObl+9ovi4kRh2TfvAfxAoHOJ9vIMEqk3k4iex7tg==",
+      "license": "MIT"
+    },
     "node_modules/json-stable-stringify-without-jsonify": {
       "version": "1.0.1",
       "resolved": "https://registry.npmmirror.com/json-stable-stringify-without-jsonify/-/json-stable-stringify-without-jsonify-1.0.1.tgz",
@@ -24393,6 +24427,44 @@
         "node": ">=6"
       }
     },
+    "node_modules/jsoneditor": {
+      "version": "10.4.2",
+      "resolved": "https://registry.npmmirror.com/jsoneditor/-/jsoneditor-10.4.2.tgz",
+      "integrity": "sha512-SQPCXlanU4PqdVsYuj2X7yfbLiiJYjklbksGfMKPsuwLhAIPxDlG43jYfXieGXvxpuq1fkw08YoRbkKXKabcLA==",
+      "license": "Apache-2.0",
+      "dependencies": {
+        "ace-builds": "^1.36.2",
+        "ajv": "^6.12.6",
+        "javascript-natural-sort": "^0.7.1",
+        "jmespath": "^0.16.0",
+        "json-source-map": "^0.6.1",
+        "jsonrepair": "^3.8.1",
+        "picomodal": "^3.0.0",
+        "vanilla-picker": "^2.12.3"
+      }
+    },
+    "node_modules/jsoneditor/node_modules/ajv": {
+      "version": "6.12.6",
+      "resolved": "https://registry.npmmirror.com/ajv/-/ajv-6.12.6.tgz",
+      "integrity": "sha512-j3fVLgvTo527anyYyJOGTYJbG+vnnQYvE0m5mmkc1TK+nxAppkCLMIL0aZ4dblVCNoGShhm+kzE4ZUykBoMg4g==",
+      "license": "MIT",
+      "dependencies": {
+        "fast-deep-equal": "^3.1.1",
+        "fast-json-stable-stringify": "^2.0.0",
+        "json-schema-traverse": "^0.4.1",
+        "uri-js": "^4.2.2"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/epoberezkin"
+      }
+    },
+    "node_modules/jsoneditor/node_modules/json-schema-traverse": {
+      "version": "0.4.1",
+      "resolved": "https://registry.npmmirror.com/json-schema-traverse/-/json-schema-traverse-0.4.1.tgz",
+      "integrity": "sha512-xbbCH5dCYU5T8LcEhhuh7HJ88HXuW3qsI3Y0zOZFKfZEHcpWiHU/Jxzk629Brsab/mMiHQti9wMP+845RPe3Vg==",
+      "license": "MIT"
+    },
     "node_modules/jsonfile": {
       "version": "6.1.0",
       "resolved": "https://registry.npmmirror.com/jsonfile/-/jsonfile-6.1.0.tgz",
@@ -24404,6 +24476,15 @@
         "graceful-fs": "^4.1.6"
       }
     },
+    "node_modules/jsonrepair": {
+      "version": "3.13.1",
+      "resolved": "https://registry.npmmirror.com/jsonrepair/-/jsonrepair-3.13.1.tgz",
+      "integrity": "sha512-WJeiE0jGfxYmtLwBTEk8+y/mYcaleyLXWaqp5bJu0/ZTSeG0KQq/wWQ8pmnkKenEdN6pdnn6QtcoSUkbqDHWNw==",
+      "license": "ISC",
+      "bin": {
+        "jsonrepair": "bin/cli.js"
+      }
+    },
     "node_modules/jsx-ast-utils": {
       "version": "3.3.5",
       "resolved": "https://registry.npmmirror.com/jsx-ast-utils/-/jsx-ast-utils-3.3.5.tgz",
@@ -27499,6 +27580,12 @@
         "node": ">=8.6"
       }
     },
+    "node_modules/picomodal": {
+      "version": "3.0.0",
+      "resolved": "https://registry.npmmirror.com/picomodal/-/picomodal-3.0.0.tgz",
+      "integrity": "sha512-FoR3TDfuLlqUvcEeK5ifpKSVVns6B4BQvc8SDF6THVMuadya6LLtji0QgUDSStw0ZR2J7I6UGi5V2V23rnPWTw==",
+      "license": "MIT"
+    },
     "node_modules/pidtree": {
       "version": "0.6.0",
       "resolved": "https://registry.npmmirror.com/pidtree/-/pidtree-0.6.0.tgz",
@@ -36235,6 +36322,15 @@
       "dev": true,
       "peer": true
     },
+    "node_modules/vanilla-picker": {
+      "version": "2.12.3",
+      "resolved": "https://registry.npmmirror.com/vanilla-picker/-/vanilla-picker-2.12.3.tgz",
+      "integrity": "sha512-qVkT1E7yMbUsB2mmJNFmaXMWE2hF8ffqzMMwe9zdAikd8u2VfnsVY2HQcOUi2F38bgbxzlJBEdS1UUhOXdF9GQ==",
+      "license": "ISC",
+      "dependencies": {
+        "@sphinxxxx/color-conversion": "^2.2.2"
+      }
+    },
     "node_modules/vary": {
       "version": "1.1.2",
       "resolved": "https://registry.npmmirror.com/vary/-/vary-1.1.2.tgz",

web/package.json

@@ -79,6 +79,7 @@
     "input-otp": "^1.4.1",
     "js-base64": "^3.7.5",
     "jsencrypt": "^3.3.2",
+    "jsoneditor": "^10.4.2",
     "lexical": "^0.23.1",
     "lodash": "^4.17.21",
     "lucide-react": "^0.546.0",

web/src/components/json-edit/css/cloud9_night.less

@@ -0,0 +1,132 @@
+.ace-tomorrow-night .ace_gutter {
+  background: var(--bg-card);
+  color: rgb(var(--text-primary));
+}
+.ace-tomorrow-night .ace_print-margin {
+  width: 1px;
+  background: #25282c;
+}
+
+.ace-tomorrow-night {
+  background: var(--bg-card);
+  color: rgb(var(--text-primary));
+  .ace_editor {
+    background: var(--bg-card);
+  }
+}
+
+.ace-tomorrow-night .ace_cursor {
+  color: #aeafad;
+}
+
+.ace-tomorrow-night .ace_marker-layer .ace_selection {
+  background: #373b41;
+}
+
+.ace-tomorrow-night.ace_multiselect .ace_selection.ace_start {
+  box-shadow: 0 0 3px 0px #1d1f21;
+}
+
+.ace-tomorrow-night .ace_marker-layer .ace_step {
+  background: rgb(102, 82, 0);
+}
+
+.ace-tomorrow-night .ace_marker-layer .ace_bracket {
+  margin: -1px 0 0 -1px;
+  border: 1px solid #4b4e55;
+}
+
+.ace-tomorrow-night .ace_marker-layer .ace_active-line {
+  background: var(--bg-card);
+}
+
+.ace-tomorrow-night .ace_gutter-active-line {
+  background-color: var(--bg-card);
+}
+
+.ace-tomorrow-night .ace_marker-layer .ace_selected-word {
+  border: 1px solid #373b41;
+}
+
+.ace-tomorrow-night .ace_invisible {
+  color: #4b4e55;
+}
+
+.ace-tomorrow-night .ace_keyword,
+.ace-tomorrow-night .ace_meta,
+.ace-tomorrow-night .ace_storage,
+.ace-tomorrow-night .ace_storage.ace_type,
+.ace-tomorrow-night .ace_support.ace_type {
+  color: #b294bb;
+}
+
+.ace-tomorrow-night .ace_keyword.ace_operator {
+  color: #8abeb7;
+}
+
+.ace-tomorrow-night .ace_constant.ace_character,
+.ace-tomorrow-night .ace_constant.ace_language,
+.ace-tomorrow-night .ace_constant.ace_numeric,
+.ace-tomorrow-night .ace_keyword.ace_other.ace_unit,
+.ace-tomorrow-night .ace_support.ace_constant,
+.ace-tomorrow-night .ace_variable.ace_parameter {
+  color: #de935f;
+}
+
+.ace-tomorrow-night .ace_constant.ace_other {
+  color: #ced1cf;
+}
+
+.ace-tomorrow-night .ace_invalid {
+  color: #ced2cf;
+  background-color: #df5f5f;
+}
+
+.ace-tomorrow-night .ace_invalid.ace_deprecated {
+  color: #ced2cf;
+  background-color: #b798bf;
+}
+
+.ace-tomorrow-night .ace_fold {
+  background-color: #81a2be;
+  border-color: #c5c8c6;
+}
+
+.ace-tomorrow-night .ace_entity.ace_name.ace_function,
+.ace-tomorrow-night .ace_support.ace_function,
+.ace-tomorrow-night .ace_variable {
+  color: #81a2be;
+}
+
+.ace-tomorrow-night .ace_support.ace_class,
+.ace-tomorrow-night .ace_support.ace_type {
+  color: #f0c674;
+}
+
+.ace-tomorrow-night .ace_heading,
+.ace-tomorrow-night .ace_markup.ace_heading,
+.ace-tomorrow-night .ace_string {
+  color: #b5bd68;
+}
+
+.ace-tomorrow-night .ace_entity.ace_name.ace_tag,
+.ace-tomorrow-night .ace_entity.ace_other.ace_attribute-name,
+.ace-tomorrow-night .ace_meta.ace_tag,
+.ace-tomorrow-night .ace_string.ace_regexp,
+.ace-tomorrow-night .ace_variable {
+  color: #cc6666;
+}
+
+.ace-tomorrow-night .ace_comment {
+  color: #969896;
+}
+
+.ace-tomorrow-night .ace_indent-guide {
+  background: url(data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAACCAYAAACZgbYnAAAAEklEQVQImWNgYGBgYHB3d/8PAAOIAdULw8qMAAAAAElFTkSuQmCC)
+    right repeat-y;
+}
+
+.ace-tomorrow-night .ace_indent-guide-active {
+  background: url(data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAACCAYAAACZgbYnAAAAEklEQVQIW2PQ1dX9zzBz5sz/ABCcBFFentLlAAAAAElFTkSuQmCC)
+    right repeat-y;
+}

web/src/components/json-edit/css/index.less

@@ -0,0 +1,83 @@
+.jsoneditor {
+  border: none;
+  color: rgb(var(--text-primary));
+  overflow: auto;
+  scrollbar-width: none;
+  background-color: var(--bg-base);
+  .jsoneditor-menu {
+    background-color: var(--bg-base);
+    // border-color: var(--border-button);
+    border-bottom: thin solid var(--border-button);
+  }
+  .jsoneditor-navigation-bar {
+    border-bottom: 1px solid var(--border-button);
+    background-color: var(--bg-input);
+  }
+  .jsoneditor-tree {
+    background: var(--bg-base);
+  }
+  .jsoneditor-highlight {
+    background-color: var(--bg-card);
+  }
+}
+.jsoneditor-popover,
+.jsoneditor-schema-error,
+div.jsoneditor td,
+div.jsoneditor textarea,
+div.jsoneditor th,
+div.jsoneditor-field,
+div.jsoneditor-value,
+pre.jsoneditor-preview {
+  font-family: consolas, menlo, monaco, 'Ubuntu Mono', source-code-pro,
+    monospace;
+  font-size: 14px;
+  color: rgb(var(--text-primary));
+}
+
+div.jsoneditor-field.jsoneditor-highlight,
+div.jsoneditor-field[contenteditable='true']:focus,
+div.jsoneditor-field[contenteditable='true']:hover,
+div.jsoneditor-value.jsoneditor-highlight,
+div.jsoneditor-value[contenteditable='true']:focus,
+div.jsoneditor-value[contenteditable='true']:hover {
+  background-color: var(--bg-input);
+  border: 1px solid var(--border-button);
+  border-radius: 2px;
+}
+
+.jsoneditor-selected,
+.jsoneditor-contextmenu .jsoneditor-menu li ul {
+  background: var(--bg-base);
+}
+
+.jsoneditor-contextmenu .jsoneditor-menu button {
+  color: rgb(var(--text-secondary));
+}
+.jsoneditor-menu a.jsoneditor-poweredBy {
+  display: none;
+}
+.ace-jsoneditor .ace_scroller {
+  background-color: var(--bg-base);
+}
+.jsoneditor-statusbar {
+  border-top: 1px solid var(--border-button);
+  background-color: var(--bg-base);
+  color: rgb(var(--text-primary));
+}
+.jsoneditor-menu > .jsoneditor-modes > button,
+.jsoneditor-menu > button {
+  // color: rgb(var(--text-secondary));
+  background-color: var(--text-disabled);
+}
+
+.jsoneditor-menu > .jsoneditor-modes > button:active,
+.jsoneditor-menu > .jsoneditor-modes > button:focus,
+.jsoneditor-menu > button:active,
+.jsoneditor-menu > button:focus {
+  background-color: rgb(var(--text-secondary));
+}
+.jsoneditor-menu > .jsoneditor-modes > button:hover,
+.jsoneditor-menu > button:hover {
+  background-color: rgb(var(--text-secondary));
+  border: 1px solid var(--border-button);
+}

web/src/components/json-edit/index.tsx

@@ -0,0 +1,142 @@
+import React, { useEffect, useRef } from 'react';
+import { useTranslation } from 'react-i18next';
+import './css/cloud9_night.less';
+import './css/index.less';
+import { JsonEditorOptions, JsonEditorProps } from './interface';
+const defaultConfig: JsonEditorOptions = {
+  mode: 'code',
+  modes: ['tree', 'code'],
+  history: false,
+  search: false,
+  mainMenuBar: false,
+  navigationBar: false,
+  enableSort: false,
+  enableTransform: false,
+  indentation: 2,
+};
+const JsonEditor: React.FC<JsonEditorProps> = ({
+  value,
+  onChange,
+  height = '400px',
+  className = '',
+  options = {},
+}) => {
+  const containerRef = useRef<HTMLDivElement>(null);
+  const editorRef = useRef<any>(null);
+  const { i18n } = useTranslation();
+  const currentLanguageRef = useRef<string>(i18n.language);
+
+  useEffect(() => {
+    if (typeof window !== 'undefined') {
+      const JSONEditor = require('jsoneditor');
+      import('jsoneditor/dist/jsoneditor.min.css');
+
+      if (containerRef.current) {
+        // Default configuration options
+        const defaultOptions: JsonEditorOptions = {
+          ...defaultConfig,
+          language: i18n.language === 'zh' ? 'zh-CN' : 'en',
+          onChange: () => {
+            if (editorRef.current && onChange) {
+              try {
+                const updatedJson = editorRef.current.get();
+                onChange(updatedJson);
+              } catch (err) {
+                // Do not trigger onChange when parsing error occurs
+                console.error(err);
+              }
+            }
+          },
+          ...options, // Merge user provided options with defaults
+        };
+
+        editorRef.current = new JSONEditor(
+          containerRef.current,
+          defaultOptions,
+        );
+
+        if (value) {
+          editorRef.current.set(value);
+        }
+      }
+    }
+
+    return () => {
+      if (editorRef.current) {
+        if (typeof editorRef.current.destroy === 'function') {
+          editorRef.current.destroy();
+        }
+        editorRef.current = null;
+      }
+    };
+  }, []);
+
+  useEffect(() => {
+    // Update language when i18n language changes
+    // Since JSONEditor doesn't have a setOptions method, we need to recreate the editor
+    if (editorRef.current && currentLanguageRef.current !== i18n.language) {
+      currentLanguageRef.current = i18n.language;
+
+      // Save current data
+      let currentData;
+      try {
+        currentData = editorRef.current.get();
+      } catch (e) {
+        // If there's an error getting data, use the passed value or empty object
+        currentData = value || {};
+      }
+
+      // Destroy the current editor
+      if (typeof editorRef.current.destroy === 'function') {
+        editorRef.current.destroy();
+      }
+
+      // Recreate the editor with new language
+      const JSONEditor = require('jsoneditor');
+
+      const newOptions: JsonEditorOptions = {
+        ...defaultConfig,
+        language: i18n.language === 'zh' ? 'zh-CN' : 'en',
+        onChange: () => {
+          if (editorRef.current && onChange) {
+            try {
+              const updatedJson = editorRef.current.get();
+              onChange(updatedJson);
+            } catch (err) {
+              // Do not trigger onChange when parsing error occurs
+            }
+          }
+        },
+        ...options, // Merge user provided options with defaults
+      };
+
+      editorRef.current = new JSONEditor(containerRef.current, newOptions);
+      editorRef.current.set(currentData);
+    }
+  }, [i18n.language, value, onChange, options]);
+
+  useEffect(() => {
+    if (editorRef.current && value !== undefined) {
+      try {
+        // Only update the editor when the value actually changes
+        const currentJson = editorRef.current.get();
+        if (JSON.stringify(currentJson) !== JSON.stringify(value)) {
+          editorRef.current.set(value);
+        }
+      } catch (err) {
+        // Skip update if there is a syntax error in the current editor
+        editorRef.current.set(value);
+      }
+    }
+  }, [value]);
+
+  return (
+    <div
+      ref={containerRef}
+      style={{ height }}
+      className={`ace-tomorrow-night w-full border border-border-button rounded-lg overflow-hidden bg-bg-input ${className} `}
+    />
+  );
+};
+
+export default JsonEditor;

web/src/components/json-edit/interface.ts

@@ -0,0 +1,339 @@
+// JSONEditor configuration options interface see: https://github.com/josdejong/jsoneditor/blob/master/docs/api.md
+export interface JsonEditorOptions {
+  /**
+   * Editor mode. Available values: 'tree' (default), 'view', 'form', 'text', and 'code'.
+   */
+  mode?: 'tree' | 'view' | 'form' | 'text' | 'code';
+
+  /**
+   * Array of available modes
+   */
+  modes?: Array<'tree' | 'view' | 'form' | 'text' | 'code'>;
+
+  /**
+   * Field name for the root node. Only applicable for modes 'tree', 'view', and 'form'
+   */
+  name?: string;
+
+  /**
+   * Theme for the editor
+   */
+  theme?: string;
+
+  /**
+   * Enable history (undo/redo). True by default. Only applicable for modes 'tree', 'view', and 'form'
+   */
+  history?: boolean;
+
+  /**
+   * Enable search box. True by default. Only applicable for modes 'tree', 'view', and 'form'
+   */
+  search?: boolean;
+
+  /**
+   * Main menu bar visibility
+   */
+  mainMenuBar?: boolean;
+
+  /**
+   * Navigation bar visibility
+   */
+  navigationBar?: boolean;
+
+  /**
+   * Status bar visibility
+   */
+  statusBar?: boolean;
+
+  /**
+   * If true, object keys are sorted before display. false by default.
+   */
+  sortObjectKeys?: boolean;
+
+  /**
+   * Enable transform functionality
+   */
+  enableTransform?: boolean;
+
+  /**
+   * Enable sort functionality
+   */
+  enableSort?: boolean;
+
+  /**
+   * Limit dragging functionality
+   */
+  limitDragging?: boolean;
+
+  /**
+   * A JSON schema object
+   */
+  schema?: any;
+
+  /**
+   * Schemas that are referenced using the `$ref` property from the JSON schema
+   */
+  schemaRefs?: Record<string, any>;
+
+  /**
+   * Array of template objects
+   */
+  templates?: Array<{
+    text: string;
+    title?: string;
+    className?: string;
+    field?: string;
+    value: any;
+  }>;
+
+  /**
+   * Ace editor instance
+   */
+  ace?: any;
+
+  /**
+   * An instance of Ajv JSON schema validator
+   */
+  ajv?: any;
+
+  /**
+   * Switch to enable/disable autocomplete
+   */
+  autocomplete?: {
+    confirmKey?: string | string[];
+    caseSensitive?: boolean;
+    getOptions?: (
+      text: string,
+      path: Array<string | number>,
+      input: string,
+      editor: any,
+    ) => string[] | Promise<string[]> | null;
+  };
+
+  /**
+   * Number of indentation spaces. 4 by default. Only applicable for modes 'text' and 'code'
+   */
+  indentation?: number;
+
+  /**
+   * Available languages
+   */
+  languages?: string[];
+
+  /**
+   * Language of the editor
+   */
+  language?: string;
+
+  /**
+   * Callback method, triggered on change of contents. Does not pass the contents itself.
+   * See also onChangeJSON and onChangeText.
+   */
+  onChange?: () => void;
+
+  /**
+   * Callback method, triggered in modes on change of contents, passing the changed contents as JSON.
+   * Only applicable for modes 'tree', 'view', and 'form'.
+   */
+  onChangeJSON?: (json: any) => void;
+
+  /**
+   * Callback method, triggered in modes on change of contents, passing the changed contents as stringified JSON.
+   */
+  onChangeText?: (text: string) => void;
+
+  /**
+   * Callback method, triggered when an error occurs
+   */
+  onError?: (error: Error) => void;
+
+  /**
+   * Callback method, triggered when node is expanded
+   */
+  onExpand?: (node: any) => void;
+
+  /**
+   * Callback method, triggered when node is collapsed
+   */
+  onCollapse?: (node: any) => void;
+
+  /**
+   * Callback method, determines if a node is editable
+   */
+  onEditable?: (node: any) => boolean | { field: boolean; value: boolean };
+
+  /**
+   * Callback method, triggered when an event occurs in a JSON field or value.
+   * Only applicable for modes 'form', 'tree' and 'view'
+   */
+  onEvent?: (node: any, event: Event) => void;
+
+  /**
+   * Callback method, triggered when the editor comes into focus, passing an object {type, target}.
+   * Applicable for all modes
+   */
+  onFocus?: (node: any) => void;
+
+  /**
+   * Callback method, triggered when the editor goes out of focus, passing an object {type, target}.
+   * Applicable for all modes
+   */
+  onBlur?: (node: any) => void;
+
+  /**
+   * Callback method, triggered when creating menu items
+   */
+  onCreateMenu?: (menuItems: any[], node: any) => any[];
+
+  /**
+   * Callback method, triggered on node selection change. Only applicable for modes 'tree', 'view', and 'form'
+   */
+  onSelectionChange?: (selection: any) => void;
+
+  /**
+   * Callback method, triggered on text selection change. Only applicable for modes 'text' and 'code'
+   */
+  onTextSelectionChange?: (selection: any) => void;
+
+  /**
+   * Callback method, triggered when a Node DOM is rendered. Function returns a css class name to be set on a node.
+   * Only applicable for modes 'form', 'tree' and 'view'
+   */
+  onClassName?: (node: any) => string | undefined;
+
+  /**
+   * Callback method, triggered when validating nodes
+   */
+  onValidate?: (
+    json: any,
+  ) =>
+    | Array<{ path: Array<string | number>; message: string }>
+    | Promise<Array<{ path: Array<string | number>; message: string }>>;
+
+  /**
+   * Callback method, triggered when node name is determined
+   */
+  onNodeName?: (parentNode: any, childNode: any, name: string) => string;
+
+  /**
+   * Callback method, triggered when mode changes
+   */
+  onModeChange?: (newMode: string, oldMode: string) => void;
+
+  /**
+   * Color picker options
+   */
+  colorPicker?: boolean;
+
+  /**
+   * Callback method for color picker
+   */
+  onColorPicker?: (
+    callback: (color: string) => void,
+    parent: HTMLElement,
+  ) => void;
+
+  /**
+   * If true, shows timestamp tag
+   */
+  timestampTag?: boolean;
+
+  /**
+   * Format for timestamps
+   */
+  timestampFormat?: string;
+
+  /**
+   * If true, unicode characters are escaped. false by default.
+   */
+  escapeUnicode?: boolean;
+
+  /**
+   * Number of children allowed for a node in 'tree', 'view', or 'form' mode before
+   * the "show more/show all" buttons appear. 100 by default.
+   */
+  maxVisibleChilds?: number;
+
+  /**
+   * Callback method for validation errors
+   */
+  onValidationError?: (
+    errors: Array<{ path: Array<string | number>; message: string }>,
+  ) => void;
+
+  /**
+   * Callback method for validation warnings
+   */
+  onValidationWarning?: (
+    warnings: Array<{ path: Array<string | number>; message: string }>,
+  ) => void;
+
+  /**
+   * The anchor element to apply an overlay and display the modals in a centered location. Defaults to document.body
+   */
+  modalAnchor?: HTMLElement | null;
+
+  /**
+   * Anchor element for popups
+   */
+  popupAnchor?: HTMLElement | null;
+
+  /**
+   * Function to create queries
+   */
+  createQuery?: () => void;
+
+  /**
+   * Function to execute queries
+   */
+  executeQuery?: () => void;
+
+  /**
+   * Query description
+   */
+  queryDescription?: string;
+
+  /**
+   * Allow schema suggestions
+   */
+  allowSchemaSuggestions?: boolean;
+
+  /**
+   * Show error table
+   */
+  showErrorTable?: boolean;
+
+  /**
+   * Validate current JSON object against the configured JSON schema
+   * Must be implemented by tree mode and text mode
+   */
+  validate?: () => Promise<any[]>;
+
+  /**
+   * Refresh the rendered contents
+   * Can be implemented by tree mode and text mode
+   */
+  refresh?: () => void;
+
+  /**
+   * Callback method triggered when schema changes
+   */
+  _onSchemaChange?: (schema: any, schemaRefs: any) => void;
+}
+
+export interface JsonEditorProps {
+  // JSON data to be displayed in the editor
+  value?: any;
+
+  // Callback function triggered when the JSON data changes
+  onChange?: (value: any) => void;
+
+  // Height of the editor
+  height?: string;
+
+  // Additional CSS class names
+  className?: string;
+
+  // Configuration options for the JSONEditor
+  options?: JsonEditorOptions;
+}

web/src/locales/en.ts

@@ -25,6 +25,7 @@ export default {
       portugueseBr: 'Portuguese (Brazil)',
       chinese: 'Simplified Chinese',
       traditionalChinese: 'Traditional Chinese',
+      russian: 'Russian',
       language: 'Language',
       languageMessage: 'Please input your language!',
       languagePlaceholder: 'select your language',
@@ -1752,6 +1753,8 @@ The variable aggregation node (originally the variable assignment node) is a cru
 The Indexer will store the content in the corresponding data structures for the selected methods.`,
       // file: 'File',
       parserMethod: 'PDF parser',
+      tableResultType: 'Table Result Type',
+      markdownImageResponseType: 'Markdown Image Response Type',
       // systemPrompt: 'System Prompt',
       systemPromptPlaceholder:
         'Enter system prompt for image analysis, if empty the system default value will be used',
@@ -1934,6 +1937,7 @@ Important structured information may include: names, dates, locations, events, k
       japanese: 'Japanese',
       korean: 'Korean',
       vietnamese: 'Vietnamese',
+      russian: 'Russian',
     },
     pagination: {
       total: 'Total {{total}}',

web/src/locales/zh.ts

@@ -1629,6 +1629,8 @@ General：实体和关系提取提示来自 GitHub - microsoft/graphrag：基于
 Tokenizer 会根据所选方式将内容存储为对应的数据结构。`,
       filenameEmbdWeight: '文件名嵌入权重',
       parserMethod: '解析方法',
+      tableResultType: '表格返回形式',
+      markdownImageResponseType: '图片返回形式',
       systemPromptPlaceholder:
         '请输入用于图像分析的系统提示词，若为空则使用系统缺省值',
       exportJson: '导出 JSON',

web/src/pages/agent/constant/pipeline.tsx

@@ -169,6 +169,7 @@ export const initialParserValues = {
     {
       fileFormat: FileType.Spreadsheet,
       output_format: SpreadsheetOutputFormat.Html,
+      parse_method: ParseDocumentType.DeepDOC,
     },
     {
       fileFormat: FileType.Image,
@@ -192,6 +193,7 @@ export const initialParserValues = {
     {
       fileFormat: FileType.PowerPoint,
       output_format: PptOutputFormat.Json,
+      parse_method: ParseDocumentType.DeepDOC,
     },
   ],
 };
@@ -243,7 +245,7 @@ export const FileTypeSuffixMap = {
   [FileType.Email]: ['eml', 'msg'],
   [FileType.TextMarkdown]: ['md', 'markdown', 'mdx', 'txt'],
   [FileType.Docx]: ['doc', 'docx'],
-  [FileType.PowerPoint]: ['pptx'],
+  [FileType.PowerPoint]: ['pptx', 'ppt'],
   [FileType.Video]: ['mp4', 'avi', 'mkv'],
   [FileType.Audio]: [
     'da',

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

@@ -22,7 +22,8 @@ import { Switch } from '@/components/ui/switch';
 import { LlmModelType } from '@/constants/knowledge';
 import { useFindLlmByUuid } from '@/hooks/use-llm-request';
 import { zodResolver } from '@hookform/resolvers/zod';
-import { memo, useCallback, useEffect, useMemo } from 'react';
+import { get } from 'lodash';
+import { memo, useEffect, useMemo } from 'react';
 import { useForm, useWatch } from 'react-hook-form';
 import { useTranslation } from 'react-i18next';
 import { z } from 'zod';
@@ -45,7 +46,10 @@ import { AgentTools, Agents } from './agent-tools';
 import { StructuredOutputDialog } from './structured-output-dialog';
 import { StructuredOutputPanel } from './structured-output-panel';
 import { useBuildPromptExtraPromptOptions } from './use-build-prompt-options';
-import { useShowStructuredOutputDialog } from './use-show-structured-output-dialog';
+import {
+  useHandleShowStructuredOutput,
+  useShowStructuredOutputDialog,
+} from './use-show-structured-output-dialog';
 import { useValues } from './use-values';
 import { useWatchFormChange } from './use-watch-change';
 
@@ -121,22 +125,19 @@ function AgentForm({ node }: INextOperatorForm) {
   });
 
   const {
-    initialStructuredOutput,
     showStructuredOutputDialog,
     structuredOutputDialogVisible,
     hideStructuredOutputDialog,
     handleStructuredOutputDialogOk,
   } = useShowStructuredOutputDialog(node?.id);
 
-  const updateNodeForm = useGraphStore((state) => state.updateNodeForm);
+  const structuredOutput = get(
+    node,
+    `data.form.outputs.${AgentStructuredOutputField}`,
+  );
 
-  const handleShowStructuredOutput = useCallback(
-    (val: boolean) => {
-      if (node?.id && val) {
-        updateNodeForm(node?.id, {}, ['outputs', AgentStructuredOutputField]);
-      }
-    },
-    [node?.id, updateNodeForm],
+  const { handleShowStructuredOutput } = useHandleShowStructuredOutput(
+    node?.id,
   );
 
   useEffect(() => {
@@ -327,7 +328,7 @@ function AgentForm({ node }: INextOperatorForm) {
               </div>
 
               <StructuredOutputPanel
-                value={initialStructuredOutput}
+                value={structuredOutput}
               ></StructuredOutputPanel>
             </section>
           )}
@@ -337,7 +338,7 @@ function AgentForm({ node }: INextOperatorForm) {
         <StructuredOutputDialog
           hideModal={hideStructuredOutputDialog}
           onOk={handleStructuredOutputDialogOk}
-          initialValues={initialStructuredOutput}
+          initialValues={structuredOutput}
         ></StructuredOutputDialog>
       )}
     </>

web/src/pages/agent/form/agent-form/use-show-structured-output-dialog.ts

@@ -1,6 +1,8 @@
 import { JSONSchema } from '@/components/jsonjoy-builder';
+import { AgentStructuredOutputField } from '@/constants/agent';
 import { useSetModalState } from '@/hooks/common-hooks';
 import { useCallback } from 'react';
+import { initialAgentValues } from '../../constant';
 import useGraphStore from '../../store';
 
 export function useShowStructuredOutputDialog(nodeId?: string) {
@@ -9,15 +11,13 @@ export function useShowStructuredOutputDialog(nodeId?: string) {
     showModal: showStructuredOutputDialog,
     hideModal: hideStructuredOutputDialog,
   } = useSetModalState();
-  const { updateNodeForm, getNode } = useGraphStore((state) => state);
-
-  const initialStructuredOutput = getNode(nodeId)?.data.form.outputs.structured;
+  const { updateNodeForm } = useGraphStore((state) => state);
 
   const handleStructuredOutputDialogOk = useCallback(
     (values: JSONSchema) => {
       // Sync data to canvas
       if (nodeId) {
-        updateNodeForm(nodeId, values, ['outputs', 'structured']);
+        updateNodeForm(nodeId, values, ['outputs', AgentStructuredOutputField]);
       }
       hideStructuredOutputDialog();
     },
@@ -25,10 +25,30 @@ export function useShowStructuredOutputDialog(nodeId?: string) {
   );
 
   return {
-    initialStructuredOutput,
     structuredOutputDialogVisible,
     showStructuredOutputDialog,
     hideStructuredOutputDialog,
     handleStructuredOutputDialogOk,
   };
 }
+
+export function useHandleShowStructuredOutput(nodeId?: string) {
+  const updateNodeForm = useGraphStore((state) => state.updateNodeForm);
+
+  const handleShowStructuredOutput = useCallback(
+    (val: boolean) => {
+      if (nodeId) {
+        if (val) {
+          updateNodeForm(nodeId, {}, ['outputs', AgentStructuredOutputField]);
+        } else {
+          updateNodeForm(nodeId, initialAgentValues.outputs, ['outputs']);
+        }
+      }
+    },
+    [nodeId, updateNodeForm],
+  );
+
+  return {
+    handleShowStructuredOutput,
+  };
+}

web/src/pages/agent/form/agent-form/use-values.ts

@@ -6,8 +6,10 @@ import { initialAgentValues } from '../../constant';
 
 // You need to exclude the mcp and tools fields that are not in the form,
 // otherwise the form data update will reset the tools or mcp data to an array
+// Exclude data that is not in the form to avoid writing this data to the canvas when using useWatch.
+// Outputs, tools, and MCP data are directly synchronized to the canvas without going through the form.
 function omitToolsAndMcp(values: Record<string, any>) {
-  return omit(values, ['mcp', 'tools']);
+  return omit(values, ['mcp', 'tools', 'outputs']);
 }
 
 export function useValues(node?: RAGFlowNodeType) {

web/src/pages/agent/form/agent-form/use-watch-change.ts

@@ -1,7 +1,6 @@
-import { omit } from 'lodash';
 import { useEffect } from 'react';
 import { UseFormReturn, useWatch } from 'react-hook-form';
-import { AgentStructuredOutputField, PromptRole } from '../../constant';
+import { PromptRole } from '../../constant';
 import useGraphStore from '../../store';
 
 export function useWatchFormChange(id?: string, form?: UseFormReturn<any>) {
@@ -17,14 +16,6 @@ export function useWatchFormChange(id?: string, form?: UseFormReturn<any>) {
         prompts: [{ role: PromptRole.User, content: values.prompts }],
       };
 
-      if (!values.showStructuredOutput) {
-        nextValues = {
-          ...nextValues,
-          outputs: omit(values.outputs, [AgentStructuredOutputField]),
-        };
-      } else {
-        nextValues = omit(nextValues, 'outputs');
-      }
       updateNodeForm(id, nextValues);
     }
   }, [form?.formState.isDirty, id, updateNodeForm, values]);

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

@@ -34,6 +34,8 @@ import { OutputFormatFormField } from './common-form-fields';
 import { EmailFormFields } from './email-form-fields';
 import { ImageFormFields } from './image-form-fields';
 import { PdfFormFields } from './pdf-form-fields';
+import { PptFormFields } from './ppt-form-fields';
+import { SpreadsheetFormFields } from './spreadsheet-form-fields';
 import { buildFieldNameWithPrefix } from './utils';
 import { AudioFormFields, VideoFormFields } from './video-form-fields';
 
@@ -41,6 +43,8 @@ const outputList = buildOutputList(initialParserValues.outputs);
 
 const FileFormatWidgetMap = {
   [FileType.PDF]: PdfFormFields,
+  [FileType.Spreadsheet]: SpreadsheetFormFields,
+  [FileType.PowerPoint]: PptFormFields,
   [FileType.Video]: VideoFormFields,
   [FileType.Audio]: AudioFormFields,
   [FileType.Email]: EmailFormFields,
@@ -65,6 +69,8 @@ export const FormSchema = z.object({
       fields: z.array(z.string()).optional(),
       llm_id: z.string().optional(),
       system_prompt: z.string().optional(),
+      table_result_type: z.string().optional(),
+      markdown_image_response_type: z.string().optional(),
     }),
   ),
 });
@@ -184,6 +190,8 @@ const ParserForm = ({ node }: INextOperatorForm) => {
       lang: '',
       fields: [],
       llm_id: '',
+      table_result_type: '',
+      markdown_image_response_type: '',
     });
   }, [append]);
 

web/src/pages/agent/form/parser-form/pdf-form-fields.tsx

@@ -1,13 +1,30 @@
 import { ParseDocumentType } from '@/components/layout-recognize-form-field';
+import {
+  SelectWithSearch,
+  SelectWithSearchFlagOptionType,
+} from '@/components/originui/select-with-search';
+import { RAGFlowFormItem } from '@/components/ragflow-form';
 import { isEmpty } from 'lodash';
 import { useEffect, useMemo } from 'react';
 import { useFormContext, useWatch } from 'react-hook-form';
+import { useTranslation } from 'react-i18next';
 import { LanguageFormField, ParserMethodFormField } from './common-form-fields';
 import { CommonProps } from './interface';
 import { useSetInitialLanguage } from './use-set-initial-language';
 import { buildFieldNameWithPrefix } from './utils';
 
+const tableResultTypeOptions: SelectWithSearchFlagOptionType[] = [
+  { label: 'Markdown', value: '0' },
+  { label: 'HTML', value: '1' },
+];
+
+const markdownImageResponseTypeOptions: SelectWithSearchFlagOptionType[] = [
+  { label: 'URL', value: '0' },
+  { label: 'Text', value: '1' },
+];
+
 export function PdfFormFields({ prefix }: CommonProps) {
+  const { t } = useTranslation();
   const form = useFormContext();
 
   const parseMethodName = buildFieldNameWithPrefix('parse_method', prefix);
@@ -25,6 +42,12 @@ export function PdfFormFields({ prefix }: CommonProps) {
     );
   }, [parseMethod]);
 
+  const tcadpOptionsShown = useMemo(() => {
+    return (
+      !isEmpty(parseMethod) && parseMethod === ParseDocumentType.TCADPParser
+    );
+  }, [parseMethod]);
+
   useSetInitialLanguage({ prefix, languageShown });
 
   useEffect(() => {
@@ -36,10 +59,68 @@ export function PdfFormFields({ prefix }: CommonProps) {
     }
   }, [form, parseMethodName]);
 
+  // Set default values for TCADP options when TCADP is selected
+  useEffect(() => {
+    if (tcadpOptionsShown) {
+      const tableResultTypeName = buildFieldNameWithPrefix(
+        'table_result_type',
+        prefix,
+      );
+      const markdownImageResponseTypeName = buildFieldNameWithPrefix(
+        'markdown_image_response_type',
+        prefix,
+      );
+
+      if (isEmpty(form.getValues(tableResultTypeName))) {
+        form.setValue(tableResultTypeName, '1', {
+          shouldValidate: true,
+          shouldDirty: true,
+        });
+      }
+      if (isEmpty(form.getValues(markdownImageResponseTypeName))) {
+        form.setValue(markdownImageResponseTypeName, '1', {
+          shouldValidate: true,
+          shouldDirty: true,
+        });
+      }
+    }
+  }, [tcadpOptionsShown, form, prefix]);
+
   return (
     <>
       <ParserMethodFormField prefix={prefix}></ParserMethodFormField>
       {languageShown && <LanguageFormField prefix={prefix}></LanguageFormField>}
+      {tcadpOptionsShown && (
+        <>
+          <RAGFlowFormItem
+            name={buildFieldNameWithPrefix('table_result_type', prefix)}
+            label={t('flow.tableResultType') || '表格返回形式'}
+          >
+            {(field) => (
+              <SelectWithSearch
+                value={field.value}
+                onChange={field.onChange}
+                options={tableResultTypeOptions}
+              ></SelectWithSearch>
+            )}
+          </RAGFlowFormItem>
+          <RAGFlowFormItem
+            name={buildFieldNameWithPrefix(
+              'markdown_image_response_type',
+              prefix,
+            )}
+            label={t('flow.markdownImageResponseType') || '图片返回形式'}
+          >
+            {(field) => (
+              <SelectWithSearch
+                value={field.value}
+                onChange={field.onChange}
+                options={markdownImageResponseTypeOptions}
+              ></SelectWithSearch>
+            )}
+          </RAGFlowFormItem>
+        </>
+      )}
     </>
   );
 }

web/src/pages/agent/form/parser-form/ppt-form-fields.tsx

@@ -0,0 +1,125 @@
+import { ParseDocumentType } from '@/components/layout-recognize-form-field';
+import {
+  SelectWithSearch,
+  SelectWithSearchFlagOptionType,
+} from '@/components/originui/select-with-search';
+import { RAGFlowFormItem } from '@/components/ragflow-form';
+import { isEmpty } from 'lodash';
+import { useEffect, useMemo } from 'react';
+import { useFormContext, useWatch } from 'react-hook-form';
+import { useTranslation } from 'react-i18next';
+import { ParserMethodFormField } from './common-form-fields';
+import { CommonProps } from './interface';
+import { buildFieldNameWithPrefix } from './utils';
+
+const tableResultTypeOptions: SelectWithSearchFlagOptionType[] = [
+  { label: 'Markdown', value: '0' },
+  { label: 'HTML', value: '1' },
+];
+
+const markdownImageResponseTypeOptions: SelectWithSearchFlagOptionType[] = [
+  { label: 'URL', value: '0' },
+  { label: 'Text', value: '1' },
+];
+
+export function PptFormFields({ prefix }: CommonProps) {
+  const { t } = useTranslation();
+  const form = useFormContext();
+
+  const parseMethodName = buildFieldNameWithPrefix('parse_method', prefix);
+
+  const parseMethod = useWatch({
+    name: parseMethodName,
+  });
+
+  // PPT only supports DeepDOC and TCADPParser
+  const optionsWithoutLLM = [
+    { label: ParseDocumentType.DeepDOC, value: ParseDocumentType.DeepDOC },
+    {
+      label: ParseDocumentType.TCADPParser,
+      value: ParseDocumentType.TCADPParser,
+    },
+  ];
+
+  const tcadpOptionsShown = useMemo(() => {
+    return (
+      !isEmpty(parseMethod) && parseMethod === ParseDocumentType.TCADPParser
+    );
+  }, [parseMethod]);
+
+  useEffect(() => {
+    if (isEmpty(form.getValues(parseMethodName))) {
+      form.setValue(parseMethodName, ParseDocumentType.DeepDOC, {
+        shouldValidate: true,
+        shouldDirty: true,
+      });
+    }
+  }, [form, parseMethodName]);
+
+  // Set default values for TCADP options when TCADP is selected
+  useEffect(() => {
+    if (tcadpOptionsShown) {
+      const tableResultTypeName = buildFieldNameWithPrefix(
+        'table_result_type',
+        prefix,
+      );
+      const markdownImageResponseTypeName = buildFieldNameWithPrefix(
+        'markdown_image_response_type',
+        prefix,
+      );
+
+      if (isEmpty(form.getValues(tableResultTypeName))) {
+        form.setValue(tableResultTypeName, '1', {
+          shouldValidate: true,
+          shouldDirty: true,
+        });
+      }
+      if (isEmpty(form.getValues(markdownImageResponseTypeName))) {
+        form.setValue(markdownImageResponseTypeName, '1', {
+          shouldValidate: true,
+          shouldDirty: true,
+        });
+      }
+    }
+  }, [tcadpOptionsShown, form, prefix]);
+
+  return (
+    <>
+      <ParserMethodFormField
+        prefix={prefix}
+        optionsWithoutLLM={optionsWithoutLLM}
+      ></ParserMethodFormField>
+      {tcadpOptionsShown && (
+        <>
+          <RAGFlowFormItem
+            name={buildFieldNameWithPrefix('table_result_type', prefix)}
+            label={t('flow.tableResultType') || '表格返回形式'}
+          >
+            {(field) => (
+              <SelectWithSearch
+                value={field.value}
+                onChange={field.onChange}
+                options={tableResultTypeOptions}
+              ></SelectWithSearch>
+            )}
+          </RAGFlowFormItem>
+          <RAGFlowFormItem
+            name={buildFieldNameWithPrefix(
+              'markdown_image_response_type',
+              prefix,
+            )}
+            label={t('flow.markdownImageResponseType') || '图片返回形式'}
+          >
+            {(field) => (
+              <SelectWithSearch
+                value={field.value}
+                onChange={field.onChange}
+                options={markdownImageResponseTypeOptions}
+              ></SelectWithSearch>
+            )}
+          </RAGFlowFormItem>
+        </>
+      )}
+    </>
+  );
+}

web/src/pages/agent/form/parser-form/spreadsheet-form-fields.tsx

@@ -0,0 +1,125 @@
+import { ParseDocumentType } from '@/components/layout-recognize-form-field';
+import {
+  SelectWithSearch,
+  SelectWithSearchFlagOptionType,
+} from '@/components/originui/select-with-search';
+import { RAGFlowFormItem } from '@/components/ragflow-form';
+import { isEmpty } from 'lodash';
+import { useEffect, useMemo } from 'react';
+import { useFormContext, useWatch } from 'react-hook-form';
+import { useTranslation } from 'react-i18next';
+import { ParserMethodFormField } from './common-form-fields';
+import { CommonProps } from './interface';
+import { buildFieldNameWithPrefix } from './utils';
+
+const tableResultTypeOptions: SelectWithSearchFlagOptionType[] = [
+  { label: 'Markdown', value: '0' },
+  { label: 'HTML', value: '1' },
+];
+
+const markdownImageResponseTypeOptions: SelectWithSearchFlagOptionType[] = [
+  { label: 'URL', value: '0' },
+  { label: 'Text', value: '1' },
+];
+
+export function SpreadsheetFormFields({ prefix }: CommonProps) {
+  const { t } = useTranslation();
+  const form = useFormContext();
+
+  const parseMethodName = buildFieldNameWithPrefix('parse_method', prefix);
+
+  const parseMethod = useWatch({
+    name: parseMethodName,
+  });
+
+  // Spreadsheet only supports DeepDOC and TCADPParser
+  const optionsWithoutLLM = [
+    { label: ParseDocumentType.DeepDOC, value: ParseDocumentType.DeepDOC },
+    {
+      label: ParseDocumentType.TCADPParser,
+      value: ParseDocumentType.TCADPParser,
+    },
+  ];
+
+  const tcadpOptionsShown = useMemo(() => {
+    return (
+      !isEmpty(parseMethod) && parseMethod === ParseDocumentType.TCADPParser
+    );
+  }, [parseMethod]);
+
+  useEffect(() => {
+    if (isEmpty(form.getValues(parseMethodName))) {
+      form.setValue(parseMethodName, ParseDocumentType.DeepDOC, {
+        shouldValidate: true,
+        shouldDirty: true,
+      });
+    }
+  }, [form, parseMethodName]);
+
+  // Set default values for TCADP options when TCADP is selected
+  useEffect(() => {
+    if (tcadpOptionsShown) {
+      const tableResultTypeName = buildFieldNameWithPrefix(
+        'table_result_type',
+        prefix,
+      );
+      const markdownImageResponseTypeName = buildFieldNameWithPrefix(
+        'markdown_image_response_type',
+        prefix,
+      );
+
+      if (isEmpty(form.getValues(tableResultTypeName))) {
+        form.setValue(tableResultTypeName, '1', {
+          shouldValidate: true,
+          shouldDirty: true,
+        });
+      }
+      if (isEmpty(form.getValues(markdownImageResponseTypeName))) {
+        form.setValue(markdownImageResponseTypeName, '1', {
+          shouldValidate: true,
+          shouldDirty: true,
+        });
+      }
+    }
+  }, [tcadpOptionsShown, form, prefix]);
+
+  return (
+    <>
+      <ParserMethodFormField
+        prefix={prefix}
+        optionsWithoutLLM={optionsWithoutLLM}
+      ></ParserMethodFormField>
+      {tcadpOptionsShown && (
+        <>
+          <RAGFlowFormItem
+            name={buildFieldNameWithPrefix('table_result_type', prefix)}
+            label={t('flow.tableResultType') || '表格返回形式'}
+          >
+            {(field) => (
+              <SelectWithSearch
+                value={field.value}
+                onChange={field.onChange}
+                options={tableResultTypeOptions}
+              ></SelectWithSearch>
+            )}
+          </RAGFlowFormItem>
+          <RAGFlowFormItem
+            name={buildFieldNameWithPrefix(
+              'markdown_image_response_type',
+              prefix,
+            )}
+            label={t('flow.markdownImageResponseType') || '图片返回形式'}
+          >
+            {(field) => (
+              <SelectWithSearch
+                value={field.value}
+                onChange={field.onChange}
+                options={markdownImageResponseTypeOptions}
+              ></SelectWithSearch>
+            )}
+          </RAGFlowFormItem>
+        </>
+      )}
+    </>
+  );
+}

web/src/pages/agent/gobal-variable-sheet/hooks/use-object-fields.tsx

@@ -1,7 +1,7 @@
+import JsonEditor from '@/components/json-edit';
 import { BlockButton, Button } from '@/components/ui/button';
 import { Input } from '@/components/ui/input';
 import { Segmented } from '@/components/ui/segmented';
-import { Editor } from '@monaco-editor/react';
 import { t } from 'i18next';
 import { Trash2, X } from 'lucide-react';
 import { useCallback } from 'react';
@@ -31,32 +31,80 @@ export const useObjectFields = () => {
     },
     [],
   );
+  const validateKeys = (
+    obj: any,
+    path: (string | number)[] = [],
+  ): Array<{ path: (string | number)[]; message: string }> => {
+    const errors: Array<{ path: (string | number)[]; message: string }> = [];
 
+    if (obj !== null && typeof obj === 'object' && !Array.isArray(obj)) {
+      for (const key in obj) {
+        if (obj.hasOwnProperty(key)) {
+          if (!/^[a-zA-Z_]+$/.test(key)) {
+            errors.push({
+              path: [...path, key],
+              message: `Key "${key}" is invalid. Keys can only contain letters and underscores.`,
+            });
+          }
+          const nestedErrors = validateKeys(obj[key], [...path, key]);
+          errors.push(...nestedErrors);
+        }
+      }
+    } else if (Array.isArray(obj)) {
+      obj.forEach((item, index) => {
+        const nestedErrors = validateKeys(item, [...path, index]);
+        errors.push(...nestedErrors);
+      });
+    }
+
+    return errors;
+  };
   const objectRender = useCallback((field: FieldValues) => {
-    const fieldValue =
-      typeof field.value === 'object'
-        ? JSON.stringify(field.value, null, 2)
-        : JSON.stringify({}, null, 2);
-    console.log('object-render-field', field, fieldValue);
+    // const fieldValue =
+    //   typeof field.value === 'object'
+    //     ? JSON.stringify(field.value, null, 2)
+    //     : JSON.stringify({}, null, 2);
+    // console.log('object-render-field', field, fieldValue);
     return (
-      <Editor
-        height={200}
-        defaultLanguage="json"
-        theme="vs-dark"
-        value={fieldValue}
+      // <Editor
+      //   height={200}
+      //   defaultLanguage="json"
+      //   theme="vs-dark"
+      //   value={fieldValue}
+      //   onChange={field.onChange}
+      // />
+      <JsonEditor
+        value={field.value}
         onChange={field.onChange}
+        height="400px"
+        options={{
+          mode: 'code',
+          navigationBar: false,
+          mainMenuBar: true,
+          history: true,
+          onValidate: (json) => {
+            return validateKeys(json);
+          },
+        }}
       />
     );
   }, []);
 
   const objectValidate = useCallback((value: any) => {
     try {
-      if (!JSON.parse(value)) {
-        throw new Error(t('knowledgeDetails.formatTypeError'));
+      if (validateKeys(value, [])?.length > 0) {
+        throw new Error(t('flow.formatTypeError'));
+      }
+      if (!z.object({}).safeParse(value).success) {
+        throw new Error(t('flow.formatTypeError'));
+      }
+      if (value && typeof value === 'string' && !JSON.parse(value)) {
+        throw new Error(t('flow.formatTypeError'));
       }
       return true;
     } catch (e) {
-      throw new Error(t('knowledgeDetails.formatTypeError'));
+      console.log('object-render-error', e, value);
+      throw new Error(t('flow.formatTypeError'));
     }
   }, []);
 
@@ -219,6 +267,10 @@ export const useObjectFields = () => {
   };
   const handleCustomSchema = (value: TypesWithArray) => {
     switch (value) {
+      case TypesWithArray.Object:
+        return z.object({});
+      case TypesWithArray.ArrayObject:
+        return z.array(z.object({}));
       case TypesWithArray.ArrayString:
         return z.array(z.string());
       case TypesWithArray.ArrayNumber:

web/src/pages/agent/utils.ts

@@ -214,6 +214,36 @@ function transformParserParams(params: ParserFormSchemaType) {
             parse_method: cur.parse_method,
             lang: cur.lang,
           };
+          // Only include TCADP parameters if TCADP Parser is selected
+          if (cur.parse_method?.toLowerCase() === 'tcadp parser') {
+            filteredSetup.table_result_type = cur.table_result_type;
+            filteredSetup.markdown_image_response_type =
+              cur.markdown_image_response_type;
+          }
+          break;
+        case FileType.Spreadsheet:
+          filteredSetup = {
+            ...filteredSetup,
+            parse_method: cur.parse_method,
+          };
+          // Only include TCADP parameters if TCADP Parser is selected
+          if (cur.parse_method?.toLowerCase() === 'tcadp parser') {
+            filteredSetup.table_result_type = cur.table_result_type;
+            filteredSetup.markdown_image_response_type =
+              cur.markdown_image_response_type;
+          }
+          break;
+        case FileType.PowerPoint:
+          filteredSetup = {
+            ...filteredSetup,
+            parse_method: cur.parse_method,
+          };
+          // Only include TCADP parameters if TCADP Parser is selected
+          if (cur.parse_method?.toLowerCase() === 'tcadp parser') {
+            filteredSetup.table_result_type = cur.table_result_type;
+            filteredSetup.markdown_image_response_type =
+              cur.markdown_image_response_type;
+          }
           break;
         case FileType.Image:
           filteredSetup = {

web/src/pages/data-flow/form/parser-form/ppt-form-fields.tsx

@@ -0,0 +1,40 @@
+import { ParseDocumentType } from '@/components/layout-recognize-form-field';
+import { isEmpty } from 'lodash';
+import { useEffect } from 'react';
+import { useFormContext } from 'react-hook-form';
+import { ParserMethodFormField } from './common-form-fields';
+import { CommonProps } from './interface';
+import { buildFieldNameWithPrefix } from './utils';
+
+export function PptFormFields({ prefix }: CommonProps) {
+  const form = useFormContext();
+
+  const parseMethodName = buildFieldNameWithPrefix('parse_method', prefix);
+
+  // PPT only supports DeepDOC and TCADPParser
+  const optionsWithoutLLM = [
+    { label: ParseDocumentType.DeepDOC, value: ParseDocumentType.DeepDOC },
+    {
+      label: ParseDocumentType.TCADPParser,
+      value: ParseDocumentType.TCADPParser,
+    },
+  ];
+
+  useEffect(() => {
+    if (isEmpty(form.getValues(parseMethodName))) {
+      form.setValue(parseMethodName, ParseDocumentType.DeepDOC, {
+        shouldValidate: true,
+        shouldDirty: true,
+      });
+    }
+  }, [form, parseMethodName]);
+
+  return (
+    <>
+      <ParserMethodFormField
+        prefix={prefix}
+        optionsWithoutLLM={optionsWithoutLLM}
+      ></ParserMethodFormField>
+    </>
+  );
+}

web/src/pages/data-flow/form/parser-form/spreadsheet-form-fields.tsx

@@ -0,0 +1,40 @@
+import { ParseDocumentType } from '@/components/layout-recognize-form-field';
+import { isEmpty } from 'lodash';
+import { useEffect } from 'react';
+import { useFormContext } from 'react-hook-form';
+import { ParserMethodFormField } from './common-form-fields';
+import { CommonProps } from './interface';
+import { buildFieldNameWithPrefix } from './utils';
+
+export function SpreadsheetFormFields({ prefix }: CommonProps) {
+  const form = useFormContext();
+
+  const parseMethodName = buildFieldNameWithPrefix('parse_method', prefix);
+
+  // Spreadsheet only supports DeepDOC and TCADPParser
+  const optionsWithoutLLM = [
+    { label: ParseDocumentType.DeepDOC, value: ParseDocumentType.DeepDOC },
+    {
+      label: ParseDocumentType.TCADPParser,
+      value: ParseDocumentType.TCADPParser,
+    },
+  ];
+
+  useEffect(() => {
+    if (isEmpty(form.getValues(parseMethodName))) {
+      form.setValue(parseMethodName, ParseDocumentType.DeepDOC, {
+        shouldValidate: true,
+        shouldDirty: true,
+      });
+    }
+  }, [form, parseMethodName]);
+
+  return (
+    <>
+      <ParserMethodFormField
+        prefix={prefix}
+        optionsWithoutLLM={optionsWithoutLLM}
+      ></ParserMethodFormField>
+    </>
+  );
+}