agno 2.0.5__py3-none-any.whl → 2.0.7__py3-none-any.whl

agno/tools/memori.py

@@ -1,7 +1,6 @@
 import json
 from typing import Any, Dict, List, Optional
 
-from agno.agent import Agent
 from agno.tools.toolkit import Toolkit
 from agno.utils.log import log_debug, log_error, log_info, log_warning
 
@@ -122,7 +121,6 @@ class MemoriTools(Toolkit):
 
     def search_memory(
         self,
-        agent: Agent,
         query: str,
         limit: Optional[int] = None,
     ) -> str:
@@ -180,7 +178,7 @@ class MemoriTools(Toolkit):
             log_error(f"Error searching memory: {e}")
             return json.dumps({"success": False, "error": f"Memory search error: {str(e)}"})
 
-    def record_conversation(self, agent: Agent, content: str) -> str:
+    def record_conversation(self, content: str) -> str:
         """
         Add important information or facts to memory.
 
@@ -222,7 +220,6 @@ class MemoriTools(Toolkit):
 
     def get_memory_stats(
         self,
-        agent: Agent,
     ) -> str:
         """
         Get statistics about the memory system.
@@ -340,52 +337,3 @@ class MemoriTools(Toolkit):
         except Exception as e:
             log_error(f"Failed to disable memory system: {e}")
             return False
-
-
-def create_memori_search_tool(memori_toolkit: MemoriTools):
-    """
-    Create a standalone memory search function for use with Agno agents.
-
-    This is a convenience function that creates a memory search tool similar
-    to the pattern shown in the Memori example code.
-
-    Args:
-        memori_toolkit: An initialized MemoriTools instance
-
-    Returns:
-        Callable: A memory search function that can be used as an agent tool
-
-    Example:
-        ```python
-        memori_tools = MemoriTools(database_connect="sqlite:///memory.db")
-        search_tool = create_memori_search_tool(memori_tools)
-
-        agent = Agent(
-            model=OpenAIChat(),
-            tools=[search_tool],
-            description="Agent with memory search capability"
-        )
-        ```
-    """
-
-    def search_memory(query: str) -> str:
-        """
-        Search the agent's memory for past conversations and information.
-
-        Args:
-            query: What to search for in memory
-
-        Returns:
-            str: Search results or error message
-        """
-        try:
-            if not query.strip():
-                return "Please provide a search query"
-
-            result = memori_toolkit._memory_tool.execute(query=query.strip())
-            return str(result) if result else "No relevant memories found"
-
-        except Exception as e:
-            return f"Memory search error: {str(e)}"
-
-    return search_memory

agno/utils/events.py

@@ -1,4 +1,4 @@
-from typing import Any, List, Optional
+from typing import Any, Dict, List, Optional
 
 from agno.media import Audio, Image
 from agno.models.message import Citations
@@ -76,6 +76,7 @@ def create_team_run_completed_event(from_run_response: TeamRunOutput) -> TeamRun
         content_type=from_run_response.content_type,  # type: ignore
         reasoning_content=from_run_response.reasoning_content,  # type: ignore
         citations=from_run_response.citations,  # type: ignore
+        model_provider_data=from_run_response.model_provider_data,  # type: ignore
         images=from_run_response.images,  # type: ignore
         videos=from_run_response.videos,  # type: ignore
         audio=from_run_response.audio,  # type: ignore
@@ -100,6 +101,7 @@ def create_run_completed_event(from_run_response: RunOutput) -> RunCompletedEven
         content_type=from_run_response.content_type,  # type: ignore
         reasoning_content=from_run_response.reasoning_content,  # type: ignore
         citations=from_run_response.citations,  # type: ignore
+        model_provider_data=from_run_response.model_provider_data,  # type: ignore
         images=from_run_response.images,  # type: ignore
         videos=from_run_response.videos,  # type: ignore
         audio=from_run_response.audio,  # type: ignore
@@ -343,6 +345,7 @@ def create_run_output_content_event(
     content_type: Optional[str] = None,
     reasoning_content: Optional[str] = None,
     redacted_reasoning_content: Optional[str] = None,
+    model_provider_data: Optional[Dict[str, Any]] = None,
     citations: Optional[Citations] = None,
     response_audio: Optional[Audio] = None,
     image: Optional[Image] = None,
@@ -364,6 +367,7 @@ def create_run_output_content_event(
         additional_input=from_run_response.additional_input,
         reasoning_steps=from_run_response.reasoning_steps,
         reasoning_messages=from_run_response.reasoning_messages,
+        model_provider_data=model_provider_data,
     )
 
 
@@ -374,6 +378,7 @@ def create_team_run_output_content_event(
     reasoning_content: Optional[str] = None,
     redacted_reasoning_content: Optional[str] = None,
     citations: Optional[Citations] = None,
+    model_provider_data: Optional[Dict[str, Any]] = None,
     response_audio: Optional[Audio] = None,
     image: Optional[Image] = None,
 ) -> TeamRunContentEvent:
@@ -388,6 +393,7 @@ def create_team_run_output_content_event(
         content_type=content_type or "str",
         reasoning_content=thinking_combined,
         citations=citations,
+        model_provider_data=model_provider_data,
         response_audio=response_audio,
         image=image,
         references=from_run_response.references,  # type: ignore

agno/utils/gemini.py

@@ -146,13 +146,24 @@ def convert_schema(schema_dict: Dict[str, Any], root_schema: Optional[Dict[str,
                 # For Gemini, we need to represent Dict[str, T] as an object with at least one property
                 # to avoid the "properties should be non-empty" error.
                 # We'll create a generic property that represents the dictionary structure
-                value_type = additional_props.get("type", "string").upper()
+
+                # Handle both single types and union types (arrays) from Zod schemas
+                type_value = additional_props.get("type", "string")
+                if isinstance(type_value, list):
+                    value_type = type_value[0].upper() if type_value else "STRING"
+                    union_types = ", ".join(type_value)
+                    type_description_suffix = f" (supports union types: {union_types})"
+                else:
+                    # Single type
+                    value_type = type_value.upper()
+                    type_description_suffix = ""
+
                 # Create a placeholder property to satisfy Gemini's requirements
                 # This is a workaround since Gemini doesn't support additionalProperties directly
                 placeholder_properties = {
                     "example_key": Schema(
                         type=value_type,
-                        description=f"Example key-value pair. This object can contain any number of keys with {value_type.lower()} values.",
+                        description=f"Example key-value pair. This object can contain any number of keys with {value_type.lower()} values{type_description_suffix}.",
                     )
                 }
                 if value_type == "ARRAY":
@@ -162,7 +173,7 @@ def convert_schema(schema_dict: Dict[str, Any], root_schema: Optional[Dict[str,
                     type=Type.OBJECT,
                     properties=placeholder_properties,
                     description=description
-                    or f"Dictionary with {value_type.lower()} values. Can contain any number of key-value pairs.",
+                    or f"Dictionary with {value_type.lower()} values{type_description_suffix}. Can contain any number of key-value pairs.",
                     default=default,
                 )
             else:
@@ -174,7 +185,10 @@ def convert_schema(schema_dict: Dict[str, Any], root_schema: Optional[Dict[str,
             return Schema(type=Type.OBJECT, description=description, default=default)
 
     elif schema_type == "array" and "items" in schema_dict:
-        items = convert_schema(schema_dict["items"], root_schema)
+        if not schema_dict["items"]:  # Handle empty {}
+            items = Schema(type=Type.STRING)
+        else:
+            items = convert_schema(schema_dict["items"], root_schema)
         min_items = schema_dict.get("minItems")
         max_items = schema_dict.get("maxItems")
         return Schema(
@@ -233,6 +247,12 @@ def convert_schema(schema_dict: Dict[str, Any], root_schema: Optional[Dict[str,
                 default=default,
             )
     else:
+        if isinstance(schema_type, list):
+            non_null_types = [t for t in schema_type if t != "null"]
+            if non_null_types:
+                schema_type = non_null_types[0]
+            else:
+                schema_type = ""
         # Only convert to uppercase if schema_type is not empty
         if schema_type:
             schema_type = schema_type.upper()

agno/vectordb/chroma/chromadb.py

@@ -1,4 +1,5 @@
 import asyncio
+import json
 from hashlib import md5
 from typing import Any, Dict, List, Mapping, Optional, Union, cast
 
@@ -60,6 +61,44 @@ class ChromaDb(VectorDb):
         # Chroma client kwargs
         self.kwargs = kwargs
 
+    def _flatten_metadata(self, metadata: Dict[str, Any]) -> Dict[str, Union[str, int, float, bool]]:
+        """
+        Flatten nested metadata to ChromaDB-compatible format.
+
+        Args:
+            metadata: Dictionary that may contain nested structures
+
+        Returns:
+            Flattened dictionary with only primitive values
+        """
+        flattened: Dict[str, Any] = {}
+
+        def _flatten_recursive(obj: Any, prefix: str = "") -> None:
+            if isinstance(obj, dict):
+                if len(obj) == 0:
+                    # Handle empty dictionaries by converting to JSON string
+                    flattened[prefix] = json.dumps(obj)
+                else:
+                    for key, value in obj.items():
+                        new_key = f"{prefix}.{key}" if prefix else key
+                        _flatten_recursive(value, new_key)
+            elif isinstance(obj, (list, tuple)):
+                # Convert lists/tuples to JSON strings
+                flattened[prefix] = json.dumps(obj)
+            elif isinstance(obj, (str, int, float, bool)) or obj is None:
+                if obj is not None:  # ChromaDB doesn't accept None values
+                    flattened[prefix] = obj
+            else:
+                # Convert other complex types to JSON strings
+                try:
+                    flattened[prefix] = json.dumps(obj)
+                except (TypeError, ValueError):
+                    # If it can't be serialized, convert to string
+                    flattened[prefix] = str(obj)
+
+        _flatten_recursive(metadata)
+        return flattened
+
     @property
     def client(self) -> ClientAPI:
         if self._client is None:
@@ -147,11 +186,14 @@ class ChromaDb(VectorDb):
 
             metadata["content_hash"] = content_hash
 
+            # Flatten metadata for ChromaDB compatibility
+            flattened_metadata = self._flatten_metadata(metadata)
+
             docs_embeddings.append(document.embedding)
             docs.append(cleaned_content)
             ids.append(doc_id)
-            docs_metadata.append(metadata)
-            log_debug(f"Prepared document: {document.id} | {document.name} | {metadata}")
+            docs_metadata.append(flattened_metadata)
+            log_debug(f"Prepared document: {document.id} | {document.name} | {flattened_metadata}")
 
         if self._collection is None:
             logger.warning("Collection does not exist")
@@ -196,11 +238,14 @@ class ChromaDb(VectorDb):
 
             metadata["content_hash"] = content_hash
 
+            # Flatten metadata for ChromaDB compatibility
+            flattened_metadata = self._flatten_metadata(metadata)
+
             docs_embeddings.append(document.embedding)
             docs.append(cleaned_content)
             ids.append(doc_id)
-            docs_metadata.append(metadata)
-            log_debug(f"Prepared document: {document.id} | {document.name} | {metadata}")
+            docs_metadata.append(flattened_metadata)
+            log_debug(f"Prepared document: {document.id} | {document.name} | {flattened_metadata}")
 
         if self._collection is None:
             logger.warning("Collection does not exist")
@@ -262,11 +307,14 @@ class ChromaDb(VectorDb):
 
             metadata["content_hash"] = content_hash
 
+            # Flatten metadata for ChromaDB compatibility
+            flattened_metadata = self._flatten_metadata(metadata)
+
             docs_embeddings.append(document.embedding)
             docs.append(cleaned_content)
             ids.append(doc_id)
-            docs_metadata.append(metadata)
-            log_debug(f"Upserted document: {document.id} | {document.name} | {metadata}")
+            docs_metadata.append(flattened_metadata)
+            log_debug(f"Upserted document: {document.id} | {document.name} | {flattened_metadata}")
 
         if self._collection is None:
             logger.warning("Collection does not exist")
@@ -313,11 +361,14 @@ class ChromaDb(VectorDb):
 
             metadata["content_hash"] = content_hash
 
+            # Flatten metadata for ChromaDB compatibility
+            flattened_metadata = self._flatten_metadata(metadata)
+
             docs_embeddings.append(document.embedding)
             docs.append(cleaned_content)
             ids.append(doc_id)
-            docs_metadata.append(metadata)
-            log_debug(f"Upserted document: {document.id} | {document.name} | {metadata}")
+            docs_metadata.append(flattened_metadata)
+            log_debug(f"Upserted document: {document.id} | {document.name} | {flattened_metadata}")
 
         if self._collection is None:
             logger.warning("Collection does not exist")
@@ -747,6 +798,9 @@ class ChromaDb(VectorDb):
                     logger.debug(f"No documents found with content_id: {content_id}")
                     return
 
+                # Flatten the new metadata first
+                flattened_new_metadata = self._flatten_metadata(metadata)
+
                 # Merge metadata for each document
                 updated_metadatas = []
                 for i, current_meta in enumerate(current_metadatas or []):
@@ -754,26 +808,13 @@ class ChromaDb(VectorDb):
                         meta_dict: Dict[str, Any] = {}
                     else:
                         meta_dict = dict(current_meta)  # Convert Mapping to dict
-                    updated_meta: Dict[str, Any] = meta_dict.copy()
-                    updated_meta.update(metadata)
-
-                    if "filters" not in updated_meta:
-                        updated_meta["filters"] = {}
-                    if isinstance(updated_meta["filters"], dict):
-                        updated_meta["filters"].update(metadata)
-                    else:
-                        updated_meta["filters"] = metadata
-                    updated_metadatas.append(updated_meta)
 
-                # Update the documents
-                # Filter out None values from metadata as ChromaDB doesn't accept them
-                cleaned_metadatas = []
-                for meta in updated_metadatas:
-                    cleaned_meta = {k: v for k, v in meta.items() if v is not None}
-                    cleaned_metadatas.append(cleaned_meta)
+                    # Update with flattened metadata
+                    meta_dict.update(flattened_new_metadata)
+                    updated_metadatas.append(meta_dict)
 
                 # Convert to the expected type for ChromaDB
-                chroma_metadatas = cast(List[Mapping[str, Union[str, int, float, bool]]], cleaned_metadatas)
+                chroma_metadatas = cast(List[Mapping[str, Union[str, int, float, bool]]], updated_metadatas)
                 collection.update(ids=ids, metadatas=chroma_metadatas)  # type: ignore
                 logger.debug(f"Updated metadata for {len(ids)} documents with content_id: {content_id}")
 

agno/vectordb/lancedb/lance_db.py

@@ -950,17 +950,28 @@ class LanceDb(VectorDb):
                 logger.error("Table not initialized")
                 return
 
-            # Search for documents with the given content_id
-            query_filter = f"payload->>'content_id' = '{content_id}'"
-            results = self.table.search().where(query_filter).to_pandas()
+            # Get all documents and filter in Python (LanceDB doesn't support JSON operators)
+            total_count = self.table.count_rows()
+            results = self.table.search().select(["id", "payload"]).limit(total_count).to_pandas()
 
             if results.empty:
+                logger.debug("No documents found")
+                return
+
+            # Find matching documents with the given content_id
+            matching_rows = []
+            for _, row in results.iterrows():
+                payload = json.loads(row["payload"])
+                if payload.get("content_id") == content_id:
+                    matching_rows.append(row)
+
+            if not matching_rows:
                 logger.debug(f"No documents found with content_id: {content_id}")
                 return
 
             # Update each matching document
             updated_count = 0
-            for _, row in results.iterrows():
+            for row in matching_rows:
                 row_id = row["id"]
                 current_payload = json.loads(row["payload"])
 

agno/vectordb/milvus/milvus.py

@@ -423,6 +423,9 @@ class Milvus(VectorDb):
         else:
             for document in documents:
                 document.embed(embedder=self.embedder)
+                if not document.embedding:
+                    log_debug(f"Skipping document without embedding: {document.name} ({document.meta_data})")
+                    continue
                 cleaned_content = document.content.replace("\x00", "\ufffd")
                 doc_id = md5(cleaned_content.encode()).hexdigest()
 
@@ -465,6 +468,9 @@ class Milvus(VectorDb):
 
             async def process_document(document):
                 document.embed(embedder=self.embedder)
+                if not document.embedding:
+                    log_debug(f"Skipping document without embedding: {document.name} ({document.meta_data})")
+                    return None
                 cleaned_content = document.content.replace("\x00", "\ufffd")
                 doc_id = md5(cleaned_content.encode()).hexdigest()
 

agno/workflow/workflow.py

@@ -1708,6 +1708,7 @@ class Workflow:
         # Create workflow run response with PENDING status
         workflow_run_response = WorkflowRunOutput(
             run_id=run_id,
+            input=input,
             session_id=session_id,
             workflow_id=self.id,
             workflow_name=self.name,
@@ -1798,6 +1799,7 @@ class Workflow:
         # Create workflow run response with PENDING status
         workflow_run_response = WorkflowRunOutput(
             run_id=run_id,
+            input=input,
             session_id=session_id,
             workflow_id=self.id,
             workflow_name=self.name,
@@ -1971,6 +1973,7 @@ class Workflow:
         # Create workflow run response that will be updated by reference
         workflow_run_response = WorkflowRunOutput(
             run_id=run_id,
+            input=input,
             session_id=session_id,
             workflow_id=self.id,
             workflow_name=self.name,
@@ -2139,6 +2142,7 @@ class Workflow:
         # Create workflow run response that will be updated by reference
         workflow_run_response = WorkflowRunOutput(
             run_id=run_id,
+            input=input,
             session_id=session_id,
             workflow_id=self.id,
             workflow_name=self.name,
@@ -2367,6 +2371,34 @@ class Workflow:
         """Convert workflow to dictionary representation"""
 
         def serialize_step(step):
+            # Handle callable functions (not wrapped in Step objects)
+            if callable(step) and hasattr(step, "__name__"):
+                step_dict = {
+                    "name": step.__name__,
+                    "description": "User-defined callable step",
+                    "type": StepType.STEP.value,
+                }
+                return step_dict
+
+            # Handle Agent and Team objects directly
+            if isinstance(step, Agent):
+                step_dict = {
+                    "name": step.name or "unnamed_agent",
+                    "description": step.description or "Agent step",
+                    "type": StepType.STEP.value,
+                    "agent": step,
+                }
+                return step_dict
+
+            if isinstance(step, Team):
+                step_dict = {
+                    "name": step.name or "unnamed_team",
+                    "description": step.description or "Team step",
+                    "type": StepType.STEP.value,
+                    "team": step,
+                }
+                return step_dict
+
             step_dict = {
                 "name": step.name if hasattr(step, "name") else f"unnamed_{type(step).__name__.lower()}",
                 "description": step.description if hasattr(step, "description") else "User-defined callable step",

{agno-2.0.5.dist-info → agno-2.0.7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agno
-Version: 2.0.5
+Version: 2.0.7
 Summary: Agno: a lightweight library for building Multi-Agent Systems
 Author-email: Ashpreet Bedi <ashpreet@agno.com>
 Project-URL: homepage, https://agno.com
@@ -161,6 +161,8 @@ Requires-Dist: opencv-python; extra == "opencv"
 Provides-Extra: psycopg
 Requires-Dist: psycopg-binary; extra == "psycopg"
 Requires-Dist: psycopg; extra == "psycopg"
+Provides-Extra: reportlab
+Requires-Dist: reportlab; extra == "reportlab"
 Provides-Extra: todoist
 Requires-Dist: todoist-api-python; extra == "todoist"
 Provides-Extra: valyu
@@ -306,6 +308,7 @@ Requires-Dist: agno[mem0]; extra == "tools"
 Requires-Dist: agno[memori]; extra == "tools"
 Requires-Dist: agno[google_bigquery]; extra == "tools"
 Requires-Dist: agno[psycopg]; extra == "tools"
+Requires-Dist: agno[reportlab]; extra == "tools"
 Requires-Dist: agno[trafilatura]; extra == "tools"
 Requires-Dist: agno[neo4j]; extra == "tools"
 Provides-Extra: storage