agno 2.4.2__tar.gz → 2.4.4__tar.gz

{agno-2.4.2 → agno-2.4.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agno
-Version: 2.4.2
+Version: 2.4.4
 Summary: Agno: a lightweight library for building Multi-Agent Systems
 Author-email: Ashpreet Bedi <ashpreet@agno.com>
 Project-URL: homepage, https://agno.com
@@ -260,8 +260,9 @@ Provides-Extra: text
 Requires-Dist: aiofiles; extra == "text"
 Provides-Extra: csv
 Requires-Dist: aiofiles; extra == "csv"
-Requires-Dist: openpyxl; extra == "csv"
-Requires-Dist: xlrd; extra == "csv"
+Provides-Extra: excel
+Requires-Dist: openpyxl; extra == "excel"
+Requires-Dist: xlrd; extra == "excel"
 Provides-Extra: markdown
 Requires-Dist: unstructured; extra == "markdown"
 Requires-Dist: markdown; extra == "markdown"
@@ -376,6 +377,7 @@ Requires-Dist: agno[docx]; extra == "knowledge"
 Requires-Dist: agno[pptx]; extra == "knowledge"
 Requires-Dist: agno[text]; extra == "knowledge"
 Requires-Dist: agno[csv]; extra == "knowledge"
+Requires-Dist: agno[excel]; extra == "knowledge"
 Requires-Dist: agno[markdown]; extra == "knowledge"
 Requires-Dist: agno[chonkie]; extra == "knowledge"
 Provides-Extra: embedders
@@ -450,7 +452,7 @@ Dynamic: license-file
   <span>&nbsp;•&nbsp;</span>
   <a href="https://community.agno.com/">Community</a>
   <span>&nbsp;•&nbsp;</span>
-  <a href="https://discord.gg/4MtYHHrgA8">Discord</a>
+  <a href="https://www.agno.com/discord">Discord</a>
 </div>
 
 ## What is Agno?
@@ -529,7 +531,7 @@ https://github.com/user-attachments/assets/feb23db8-15cc-4e88-be7c-01a21a03ebf6
 
 ## Getting Started
 
-1. Follow the [getting started guide](https://github.com/agno-agi/agno/tree/main/cookbook/00_getting_started)
+1. Follow the [quickstart guide](https://github.com/agno-agi/agno/tree/main/cookbook/00_quickstart)
 2. Browse the [cookbook](https://github.com/agno-agi/agno/tree/main/cookbook) for real-world examples
 3. Read the [docs](https://docs.agno.com) to go deeper
 

{agno-2.4.2 → agno-2.4.4}/README.md

@@ -19,7 +19,7 @@
   <span>&nbsp;•&nbsp;</span>
   <a href="https://community.agno.com/">Community</a>
   <span>&nbsp;•&nbsp;</span>
-  <a href="https://discord.gg/4MtYHHrgA8">Discord</a>
+  <a href="https://www.agno.com/discord">Discord</a>
 </div>
 
 ## What is Agno?
@@ -98,7 +98,7 @@ https://github.com/user-attachments/assets/feb23db8-15cc-4e88-be7c-01a21a03ebf6
 
 ## Getting Started
 
-1. Follow the [getting started guide](https://github.com/agno-agi/agno/tree/main/cookbook/00_getting_started)
+1. Follow the [quickstart guide](https://github.com/agno-agi/agno/tree/main/cookbook/00_quickstart)
 2. Browse the [cookbook](https://github.com/agno-agi/agno/tree/main/cookbook) for real-world examples
 3. Read the [docs](https://docs.agno.com) to go deeper
 

{agno-2.4.2 → agno-2.4.4}/agno/agent/agent.py

@@ -5181,6 +5181,8 @@ class Agent:
         functions: Optional[Dict[str, Function]] = None,
         stream_events: bool = False,
     ) -> Iterator[RunOutputEvent]:
+        from agno.run.agent import CustomEvent
+
         self.model = cast(Model, self.model)
         # Execute the tool
         function_call = self.model.get_function_call_to_run_from_tool_execution(tool, functions)
@@ -5222,6 +5224,10 @@ class Agent:
                                 events_to_skip=self.events_to_skip,  # type: ignore
                                 store_events=self.store_events,
                             )
+            # Yield CustomEvent instances from sync tool generators
+            elif isinstance(call_result, CustomEvent):
+                if stream_events:
+                    yield call_result  # type: ignore
 
         if len(function_call_results) > 0:
             run_messages.messages.extend(function_call_results)
@@ -5246,6 +5252,8 @@ class Agent:
         functions: Optional[Dict[str, Function]] = None,
         stream_events: bool = False,
     ) -> AsyncIterator[RunOutputEvent]:
+        from agno.run.agent import CustomEvent
+
         self.model = cast(Model, self.model)
 
         # Execute the tool
@@ -5288,6 +5296,11 @@ class Agent:
                                 events_to_skip=self.events_to_skip,  # type: ignore
                                 store_events=self.store_events,
                             )
+            # Yield CustomEvent instances from async tool generators
+            elif isinstance(call_result, CustomEvent):
+                if stream_events:
+                    yield call_result  # type: ignore
+
         if len(function_call_results) > 0:
             run_messages.messages.extend(function_call_results)
 

{agno-2.4.2 → agno-2.4.4}/agno/db/firestore/firestore.py

@@ -1,7 +1,7 @@
 import json
 import time
 from datetime import date, datetime, timedelta, timezone
-from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union
+from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union, cast
 from uuid import uuid4
 
 if TYPE_CHECKING:
@@ -25,7 +25,7 @@ from agno.db.schemas.culture import CulturalKnowledge
 from agno.db.schemas.evals import EvalFilterType, EvalRunRecord, EvalType
 from agno.db.schemas.knowledge import KnowledgeRow
 from agno.db.schemas.memory import UserMemory
-from agno.db.utils import deserialize_session_json_fields, serialize_session_json_fields
+from agno.db.utils import deserialize_session_json_fields
 from agno.session import AgentSession, Session, TeamSession, WorkflowSession
 from agno.utils.log import log_debug, log_error, log_info
 from agno.utils.string import generate_id
@@ -496,6 +496,29 @@ class FirestoreDb(BaseDb):
             if doc_ref is None:
                 return None
 
+            # Check if session_data is stored as JSON string (legacy) or native map.
+            # Legacy sessions stored session_data as a JSON string, but Firestore's dot notation
+            # (e.g., "session_data.session_name") only works with native maps. Using dot notation
+            # on a JSON string overwrites the entire field, causing data loss.
+            # For legacy sessions, we use read-modify-write which also migrates them to native maps.
+            current_doc = doc_ref.get()
+            if not current_doc.exists:
+                return None
+
+            current_data = current_doc.to_dict()
+            session_data = current_data.get("session_data") if current_data else None
+
+            if session_data is None or isinstance(session_data, str):
+                existing_session = self.get_session(session_id, session_type, deserialize=True)
+                if existing_session is None:
+                    return None
+                existing_session = cast(Session, existing_session)
+                if existing_session.session_data is None:
+                    existing_session.session_data = {}
+                existing_session.session_data["session_name"] = session_name
+                return self.upsert_session(existing_session, deserialize=deserialize)
+
+            # Native map format - use efficient dot notation update
             doc_ref.update({"session_data.session_name": session_name, "updated_at": int(time.time())})
 
             updated_doc = doc_ref.get()
@@ -539,50 +562,50 @@ class FirestoreDb(BaseDb):
         """
         try:
             collection_ref = self._get_collection(table_type="sessions", create_collection_if_not_found=True)
-            serialized_session_dict = serialize_session_json_fields(session.to_dict())
+            session_dict = session.to_dict()
 
             if isinstance(session, AgentSession):
                 record = {
-                    "session_id": serialized_session_dict.get("session_id"),
+                    "session_id": session_dict.get("session_id"),
                     "session_type": SessionType.AGENT.value,
-                    "agent_id": serialized_session_dict.get("agent_id"),
-                    "user_id": serialized_session_dict.get("user_id"),
-                    "runs": serialized_session_dict.get("runs"),
-                    "agent_data": serialized_session_dict.get("agent_data"),
-                    "session_data": serialized_session_dict.get("session_data"),
-                    "summary": serialized_session_dict.get("summary"),
-                    "metadata": serialized_session_dict.get("metadata"),
-                    "created_at": serialized_session_dict.get("created_at"),
+                    "agent_id": session_dict.get("agent_id"),
+                    "user_id": session_dict.get("user_id"),
+                    "runs": session_dict.get("runs"),
+                    "agent_data": session_dict.get("agent_data"),
+                    "session_data": session_dict.get("session_data"),
+                    "summary": session_dict.get("summary"),
+                    "metadata": session_dict.get("metadata"),
+                    "created_at": session_dict.get("created_at"),
                     "updated_at": int(time.time()),
                 }
 
             elif isinstance(session, TeamSession):
                 record = {
-                    "session_id": serialized_session_dict.get("session_id"),
+                    "session_id": session_dict.get("session_id"),
                     "session_type": SessionType.TEAM.value,
-                    "team_id": serialized_session_dict.get("team_id"),
-                    "user_id": serialized_session_dict.get("user_id"),
-                    "runs": serialized_session_dict.get("runs"),
-                    "team_data": serialized_session_dict.get("team_data"),
-                    "session_data": serialized_session_dict.get("session_data"),
-                    "summary": serialized_session_dict.get("summary"),
-                    "metadata": serialized_session_dict.get("metadata"),
-                    "created_at": serialized_session_dict.get("created_at"),
+                    "team_id": session_dict.get("team_id"),
+                    "user_id": session_dict.get("user_id"),
+                    "runs": session_dict.get("runs"),
+                    "team_data": session_dict.get("team_data"),
+                    "session_data": session_dict.get("session_data"),
+                    "summary": session_dict.get("summary"),
+                    "metadata": session_dict.get("metadata"),
+                    "created_at": session_dict.get("created_at"),
                     "updated_at": int(time.time()),
                 }
 
             elif isinstance(session, WorkflowSession):
                 record = {
-                    "session_id": serialized_session_dict.get("session_id"),
+                    "session_id": session_dict.get("session_id"),
                     "session_type": SessionType.WORKFLOW.value,
-                    "workflow_id": serialized_session_dict.get("workflow_id"),
-                    "user_id": serialized_session_dict.get("user_id"),
-                    "runs": serialized_session_dict.get("runs"),
-                    "workflow_data": serialized_session_dict.get("workflow_data"),
-                    "session_data": serialized_session_dict.get("session_data"),
-                    "summary": serialized_session_dict.get("summary"),
-                    "metadata": serialized_session_dict.get("metadata"),
-                    "created_at": serialized_session_dict.get("created_at"),
+                    "workflow_id": session_dict.get("workflow_id"),
+                    "user_id": session_dict.get("user_id"),
+                    "runs": session_dict.get("runs"),
+                    "workflow_data": session_dict.get("workflow_data"),
+                    "session_data": session_dict.get("session_data"),
+                    "summary": session_dict.get("summary"),
+                    "metadata": session_dict.get("metadata"),
+                    "created_at": session_dict.get("created_at"),
                     "updated_at": int(time.time()),
                 }
 

{agno-2.4.2 → agno-2.4.4}/agno/db/surrealdb/models.py

@@ -292,25 +292,25 @@ def get_schema(table_type: TableType, table_name: str) -> str:
     elif table_type == "knowledge":
         return dedent(f"""
             {define_table}
-            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime VALUE time::now();
+            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime DEFAULT time::now();
             DEFINE FIELD OVERWRITE updated_at ON {table_name} TYPE datetime VALUE time::now();
             """)
     elif table_type == "culture":
         return dedent(f"""
             {define_table}
-            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime VALUE time::now();
+            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime DEFAULT time::now();
             DEFINE FIELD OVERWRITE updated_at ON {table_name} TYPE datetime VALUE time::now();
             """)
     elif table_type == "sessions":
         return dedent(f"""
             {define_table}
-            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime VALUE time::now();
+            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime DEFAULT time::now();
             DEFINE FIELD OVERWRITE updated_at ON {table_name} TYPE datetime VALUE time::now();
             """)
     elif table_type == "traces":
         return dedent(f"""
             {define_table}
-            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime VALUE time::now();
+            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime DEFAULT time::now();
             DEFINE INDEX idx_trace_id ON {table_name} FIELDS trace_id UNIQUE;
             DEFINE INDEX idx_run_id ON {table_name} FIELDS run_id;
             DEFINE INDEX idx_session_id ON {table_name} FIELDS session_id;
@@ -324,7 +324,7 @@ def get_schema(table_type: TableType, table_name: str) -> str:
     elif table_type == "spans":
         return dedent(f"""
             {define_table}
-            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime VALUE time::now();
+            DEFINE FIELD OVERWRITE created_at ON {table_name} TYPE datetime DEFAULT time::now();
             DEFINE INDEX idx_span_id ON {table_name} FIELDS span_id UNIQUE;
             DEFINE INDEX idx_trace_id ON {table_name} FIELDS trace_id;
             DEFINE INDEX idx_parent_span_id ON {table_name} FIELDS parent_span_id;

{agno-2.4.2 → agno-2.4.4}/agno/db/surrealdb/surrealdb.py

@@ -287,6 +287,7 @@ class SurrealDb(BaseDb):
         where = WhereClause()
         if user_id is not None:
             where = where.and_("user_id", user_id)
+
         where_clause, where_vars = where.build()
         query = dedent(f"""
             SELECT *
@@ -295,7 +296,18 @@ class SurrealDb(BaseDb):
         """)
         vars = {"record": record, **where_vars}
         raw = self._query_one(query, vars, dict)
-        if raw is None or not deserialize:
+        if raw is None:
+            return None
+
+        # Verify session type matches
+        if session_type == SessionType.AGENT and raw.get("agent") is None:
+            return None
+        elif session_type == SessionType.TEAM and raw.get("team") is None:
+            return None
+        elif session_type == SessionType.WORKFLOW and raw.get("workflow") is None:
+            return None
+
+        if not deserialize:
             return raw
 
         return deserialize_session(session_type, raw)

{agno-2.4.2 → agno-2.4.4}/agno/knowledge/chunking/markdown.py

@@ -37,6 +37,103 @@ class MarkdownChunking(ChunkingStrategy):
             if not (1 <= split_on_headings <= 6):
                 raise ValueError("split_on_headings must be between 1 and 6 when using integer value")
 
+    def _split_large_section(self, section: str) -> List[str]:
+        """
+        Split a large section into smaller chunks while preserving the heading context.
+        Each sub-chunk will include the original heading for context.
+
+        Args:
+            section: The section content to split (may start with a heading)
+
+        Returns:
+            List of chunks, each respecting chunk_size
+        """
+        if len(section) <= self.chunk_size:
+            return [section]
+
+        # Extract heading and content from the section
+        lines = section.split("\n")
+        if lines and re.match(r"^#{1,6}\s+", lines[0]):
+            heading = lines[0]
+            content_lines = lines[1:]
+        else:
+            heading = ""
+            content_lines = lines
+
+        content = "\n".join(content_lines).strip()
+
+        # If just heading and small content, return as-is
+        if not content or len(section) <= self.chunk_size:
+            return [section]
+
+        # Split content by paragraphs
+        paragraphs = re.split(r"\n\n+", content)
+
+        chunks: List[str] = []
+        current_chunk_content: List[str] = []
+        # Account for heading size in each chunk
+        heading_size = len(heading) + 2 if heading else 0  # +2 for "\n\n"
+
+        for para in paragraphs:
+            para = para.strip()
+            if not para:
+                continue
+
+            current_size = sum(len(p) for p in current_chunk_content) + len(current_chunk_content) * 2  # \n\n
+            para_size = len(para)
+
+            # Check if adding this paragraph would exceed chunk_size
+            if current_chunk_content and (heading_size + current_size + para_size + 2) > self.chunk_size:
+                # Save current chunk
+                chunk_text = (
+                    heading + "\n\n" + "\n\n".join(current_chunk_content)
+                    if heading
+                    else "\n\n".join(current_chunk_content)
+                )
+                chunks.append(chunk_text.strip())
+                current_chunk_content = []
+
+            # If single paragraph exceeds chunk_size, split it further
+            if para_size + heading_size > self.chunk_size:
+                # Save any accumulated content first
+                if current_chunk_content:
+                    chunk_text = (
+                        heading + "\n\n" + "\n\n".join(current_chunk_content)
+                        if heading
+                        else "\n\n".join(current_chunk_content)
+                    )
+                    chunks.append(chunk_text.strip())
+                    current_chunk_content = []
+
+                # Split the large paragraph by sentences or fixed size
+                available_size = self.chunk_size - heading_size
+                words = para.split()
+                current_words: List[str] = []
+                current_word_len = 0
+
+                for word in words:
+                    if current_word_len + len(word) + 1 > available_size and current_words:
+                        chunk_text = heading + "\n\n" + " ".join(current_words) if heading else " ".join(current_words)
+                        chunks.append(chunk_text.strip())
+                        current_words = []
+                        current_word_len = 0
+                    current_words.append(word)
+                    current_word_len += len(word) + 1
+
+                if current_words:
+                    current_chunk_content.append(" ".join(current_words))
+            else:
+                current_chunk_content.append(para)
+
+        # add the remaining content
+        if current_chunk_content:
+            chunk_text = (
+                heading + "\n\n" + "\n\n".join(current_chunk_content) if heading else "\n\n".join(current_chunk_content)
+            )
+            chunks.append(chunk_text.strip())
+
+        return chunks if chunks else [section]
+
     def _split_by_headings(self, content: str) -> List[str]:
         """
         Split markdown content by headings, keeping each heading with its content.
@@ -163,18 +260,22 @@ class MarkdownChunking(ChunkingStrategy):
             section_size = len(section)
 
             # When split_on_headings is True or an int, each section becomes its own chunk
+            # But if section exceeds chunk_size, split it further
             if self.split_on_headings:
-                meta_data = chunk_meta_data.copy()
-                meta_data["chunk"] = chunk_number
-                chunk_id = None
-                if document.id:
-                    chunk_id = f"{document.id}_{chunk_number}"
-                elif document.name:
-                    chunk_id = f"{document.name}_{chunk_number}"
-                meta_data["chunk_size"] = section_size
-
-                chunks.append(Document(id=chunk_id, name=document.name, meta_data=meta_data, content=section))
-                chunk_number += 1
+                # Split large sections to respect chunk_size
+                sub_chunks = self._split_large_section(section)
+                for sub_chunk in sub_chunks:
+                    meta_data = chunk_meta_data.copy()
+                    meta_data["chunk"] = chunk_number
+                    chunk_id = None
+                    if document.id:
+                        chunk_id = f"{document.id}_{chunk_number}"
+                    elif document.name:
+                        chunk_id = f"{document.name}_{chunk_number}"
+                    meta_data["chunk_size"] = len(sub_chunk)
+
+                    chunks.append(Document(id=chunk_id, name=document.name, meta_data=meta_data, content=sub_chunk))
+                    chunk_number += 1
             elif current_size + section_size <= self.chunk_size:
                 current_chunk.append(section)
                 current_size += section_size

{agno-2.4.2 → agno-2.4.4}/agno/knowledge/embedder/openai.py

@@ -71,7 +71,8 @@ class OpenAIEmbedder(Embedder):
         }
         if self.user is not None:
             _request_params["user"] = self.user
-        if self.id.startswith("text-embedding-3"):
+        # Pass dimensions for text-embedding-3 models or when using custom base_url (third-party APIs)
+        if self.id.startswith("text-embedding-3") or self.base_url is not None:
             _request_params["dimensions"] = self.dimensions
         if self.request_params:
             _request_params.update(self.request_params)
@@ -106,7 +107,8 @@ class OpenAIEmbedder(Embedder):
         }
         if self.user is not None:
             req["user"] = self.user
-        if self.id.startswith("text-embedding-3"):
+        # Pass dimensions for text-embedding-3 models or when using custom base_url (third-party APIs)
+        if self.id.startswith("text-embedding-3") or self.base_url is not None:
             req["dimensions"] = self.dimensions
         if self.request_params:
             req.update(self.request_params)
@@ -126,7 +128,8 @@ class OpenAIEmbedder(Embedder):
         }
         if self.user is not None:
             req["user"] = self.user
-        if self.id.startswith("text-embedding-3"):
+        # Pass dimensions for text-embedding-3 models or when using custom base_url (third-party APIs)
+        if self.id.startswith("text-embedding-3") or self.base_url is not None:
             req["dimensions"] = self.dimensions
         if self.request_params:
             req.update(self.request_params)
@@ -166,7 +169,8 @@ class OpenAIEmbedder(Embedder):
             }
             if self.user is not None:
                 req["user"] = self.user
-            if self.id.startswith("text-embedding-3"):
+            # Pass dimensions for text-embedding-3 models or when using custom base_url (third-party APIs)
+            if self.id.startswith("text-embedding-3") or self.base_url is not None:
                 req["dimensions"] = self.dimensions
             if self.request_params:
                 req.update(self.request_params)

{agno-2.4.2 → agno-2.4.4}/agno/knowledge/knowledge.py

@@ -983,6 +983,11 @@ class Knowledge:
         """CSV reader - lazy loaded via factory."""
         return self._get_reader("csv")
 
+    @property
+    def excel_reader(self) -> Optional[Reader]:
+        """Excel reader - lazy loaded via factory."""
+        return self._get_reader("excel")
+
     @property
     def docx_reader(self) -> Optional[Reader]:
         """Docx reader - lazy loaded via factory."""
@@ -1123,6 +1128,8 @@ class Knowledge:
             return self.json_reader, ""
         elif file_extension == ".markdown":
             return self.markdown_reader, ""
+        elif file_extension in [".xlsx", ".xls"]:
+            return self.excel_reader, ""
         else:
             return self.text_reader, ""
 
@@ -1153,6 +1160,8 @@ class Knowledge:
             return self.json_reader
         elif uri_lower.endswith(".markdown"):
             return self.markdown_reader
+        elif uri_lower.endswith(".xlsx") or uri_lower.endswith(".xls"):
+            return self.excel_reader
         else:
             return self.text_reader
 
@@ -1284,6 +1293,10 @@ class Knowledge:
             if self._should_include_file(str(path), include, exclude):
                 log_debug(f"Adding file {path} due to include/exclude filters")
 
+                # Set name from path if not provided
+                if not content.name:
+                    content.name = path.name
+
                 await self._ainsert_contents_db(content)
                 if self._should_skip(content.content_hash, skip_if_exists):  # type: ignore[arg-type]
                     content.status = ContentStatus.COMPLETED
@@ -1365,6 +1378,10 @@ class Knowledge:
             if self._should_include_file(str(path), include, exclude):
                 log_debug(f"Adding file {path} due to include/exclude filters")
 
+                # Set name from path if not provided
+                if not content.name:
+                    content.name = path.name
+
                 self._insert_contents_db(content)
                 if self._should_skip(content.content_hash, skip_if_exists):  # type: ignore[arg-type]
                     content.status = ContentStatus.COMPLETED
@@ -1450,6 +1467,14 @@ class Knowledge:
         if not content.url:
             raise ValueError("No url provided")
 
+        # Set name from URL if not provided
+        if not content.name and content.url:
+            from urllib.parse import urlparse
+
+            parsed = urlparse(content.url)
+            url_path = Path(parsed.path)
+            content.name = url_path.name if url_path.name else content.url
+
         # 1. Add content to contents database
         await self._ainsert_contents_db(content)
         if self._should_skip(content.content_hash, skip_if_exists):  # type: ignore[arg-type]
@@ -1547,6 +1572,14 @@ class Knowledge:
         if not content.url:
             raise ValueError("No url provided")
 
+        # Set name from URL if not provided
+        if not content.name and content.url:
+            from urllib.parse import urlparse
+
+            parsed = urlparse(content.url)
+            url_path = Path(parsed.path)
+            content.name = url_path.name if url_path.name else content.url
+
         # 1. Add content to contents database
         self._insert_contents_db(content)
         if self._should_skip(content.content_hash, skip_if_exists):  # type: ignore[arg-type]
@@ -1633,6 +1666,8 @@ class Knowledge:
 
         if content.name:
             name = content.name
+        elif content.file_data and content.file_data.filename:
+            name = content.file_data.filename
         elif content.file_data and content.file_data.content:
             if isinstance(content.file_data.content, bytes):
                 name = content.file_data.content[:10].decode("utf-8", errors="ignore")
@@ -1696,9 +1731,17 @@ class Knowledge:
                     log_debug(f"Using reader: {content.reader.__class__.__name__} to read content")
                     reader = content.reader
                 else:
-                    reader = self._select_reader(content.file_data.type)
-                name = content.name if content.name else f"content_{content.file_data.type}"
-                read_documents = await reader.async_read(content_io, name=name)
+                    # Prefer filename extension over MIME type for reader selection
+                    # (browsers often send wrong MIME types for Excel files)
+                    reader_hint = content.file_data.type
+                    if content.file_data.filename:
+                        ext = Path(content.file_data.filename).suffix.lower()
+                        if ext:
+                            reader_hint = ext
+                    reader = self._select_reader(reader_hint)
+                # Use file_data.filename for reader (preserves extension for format detection)
+                reader_name = content.file_data.filename or content.name or f"content_{content.file_data.type}"
+                read_documents = await reader.async_read(content_io, name=reader_name)
                 if not content.id:
                     content.id = generate_id(content.content_hash or "")
                 self._prepare_documents_for_insert(read_documents, content.id, metadata=content.metadata)
@@ -1730,6 +1773,8 @@ class Knowledge:
 
         if content.name:
             name = content.name
+        elif content.file_data and content.file_data.filename:
+            name = content.file_data.filename
         elif content.file_data and content.file_data.content:
             if isinstance(content.file_data.content, bytes):
                 name = content.file_data.content[:10].decode("utf-8", errors="ignore")
@@ -1793,9 +1838,17 @@ class Knowledge:
                     log_debug(f"Using reader: {content.reader.__class__.__name__} to read content")
                     reader = content.reader
                 else:
-                    reader = self._select_reader(content.file_data.type)
-                name = content.name if content.name else f"content_{content.file_data.type}"
-                read_documents = reader.read(content_io, name=name)
+                    # Prefer filename extension over MIME type for reader selection
+                    # (browsers often send wrong MIME types for Excel files)
+                    reader_hint = content.file_data.type
+                    if content.file_data.filename:
+                        ext = Path(content.file_data.filename).suffix.lower()
+                        if ext:
+                            reader_hint = ext
+                    reader = self._select_reader(reader_hint)
+                # Use file_data.filename for reader (preserves extension for format detection)
+                reader_name = content.file_data.filename or content.name or f"content_{content.file_data.type}"
+                read_documents = reader.read(content_io, name=reader_name)
                 if not content.id:
                     content.id = generate_id(content.content_hash or "")
                 self._prepare_documents_for_insert(read_documents, content.id, metadata=content.metadata)