agno 2.0.6__tar.gz → 2.0.7__tar.gz

{agno-2.0.6 → agno-2.0.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agno
-Version: 2.0.6
+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

{agno-2.0.6 → agno-2.0.7}/agno/agent/agent.py

@@ -762,6 +762,7 @@ class Agent:
             tool_call_limit=self.tool_call_limit,
             response_format=response_format,
             run_response=run_response,
+            send_media_to_model=self.send_media_to_model,
         )
 
         # Check for cancellation after model call
@@ -1359,6 +1360,7 @@ class Agent:
             tool_choice=self.tool_choice,
             tool_call_limit=self.tool_call_limit,
             response_format=response_format,
+            send_media_to_model=self.send_media_to_model,
         )
 
         # Check for cancellation after model call
@@ -3100,6 +3102,8 @@ class Agent:
         # Update the run_response citations with the model response citations
         if model_response.citations is not None:
             run_response.citations = model_response.citations
+        if model_response.provider_data is not None:
+            run_response.model_provider_data = model_response.provider_data
 
         # Update the run_response tools with the model response tool_executions
         if model_response.tool_executions is not None:
@@ -3174,6 +3178,7 @@ class Agent:
             tool_call_limit=self.tool_call_limit,
             stream_model_response=stream_model_response,
             run_response=run_response,
+            send_media_to_model=self.send_media_to_model,
         ):
             yield from self._handle_model_response_chunk(
                 session=session,
@@ -3250,6 +3255,7 @@ class Agent:
             tool_call_limit=self.tool_call_limit,
             stream_model_response=stream_model_response,
             run_response=run_response,
+            send_media_to_model=self.send_media_to_model,
         )  # type: ignore
 
         async for model_response_event in model_response_stream:  # type: ignore
@@ -3352,6 +3358,10 @@ class Agent:
                         model_response.reasoning_content += model_response_event.redacted_reasoning_content
                     run_response.reasoning_content = model_response.reasoning_content
 
+                if model_response_event.provider_data is not None:
+                    # We get citations in one chunk
+                    run_response.model_provider_data = model_response.provider_data
+
                 if model_response_event.citations is not None:
                     # We get citations in one chunk
                     run_response.citations = model_response_event.citations
@@ -3372,6 +3382,7 @@ class Agent:
                     or model_response_event.reasoning_content is not None
                     or model_response_event.redacted_reasoning_content is not None
                     or model_response_event.citations is not None
+                    or model_response_event.provider_data is not None
                 ):
                     yield self._handle_event(
                         create_run_output_content_event(
@@ -3380,6 +3391,7 @@ class Agent:
                             reasoning_content=model_response_event.reasoning_content,
                             redacted_reasoning_content=model_response_event.redacted_reasoning_content,
                             citations=model_response_event.citations,
+                            model_provider_data=model_response_event.provider_data,
                         ),
                         run_response,
                         workflow_context=workflow_context,
@@ -3813,7 +3825,9 @@ class Agent:
             self._rebuild_tools = True
         if self.search_session_history:
             agent_tools.append(
-                self._get_previous_sessions_messages_function(num_history_sessions=self.num_history_sessions, user_id=user_id)
+                self._get_previous_sessions_messages_function(
+                    num_history_sessions=self.num_history_sessions, user_id=user_id
+                )
             )
             self._rebuild_tools = True
 

{agno-2.0.6 → agno-2.0.7}/agno/knowledge/chunking/semantic.py

@@ -1,4 +1,5 @@
-from typing import List, Optional
+import inspect
+from typing import Any, Dict, List, Optional
 
 from agno.knowledge.chunking.strategy import ChunkingStrategy
 from agno.knowledge.document.base import Document
@@ -26,11 +27,37 @@ class SemanticChunking(ChunkingStrategy):
                     "Please install it using `pip install chonkie` to use SemanticChunking."
                 )
 
-            self.chunker = SemanticChunker(
-                embedding_model=self.embedder.id,  # type: ignore
-                chunk_size=self.chunk_size,
-                threshold=self.similarity_threshold,
-            )
+            # Build arguments dynamically based on chonkie's supported signature
+            params: Dict[str, Any] = {
+                "chunk_size": self.chunk_size,
+                "threshold": self.similarity_threshold,
+            }
+
+            try:
+                sig = inspect.signature(SemanticChunker)
+                param_names = set(sig.parameters.keys())
+
+                # Prefer passing a callable to avoid Chonkie initializing its own client
+                if "embedding_fn" in param_names:
+                    params["embedding_fn"] = self.embedder.get_embedding  # type: ignore[attr-defined]
+                    # If chonkie allows specifying dimensions, provide them
+                    if "embedding_dimensions" in param_names and getattr(self.embedder, "dimensions", None):
+                        params["embedding_dimensions"] = self.embedder.dimensions  # type: ignore[attr-defined]
+                elif "embedder" in param_names:
+                    # Some versions may accept an embedder object directly
+                    params["embedder"] = self.embedder
+                else:
+                    # Fallback to model id
+                    params["embedding_model"] = getattr(self.embedder, "id", None) or "text-embedding-3-small"
+
+                self.chunker = SemanticChunker(**params)
+            except Exception:
+                # As a final fallback, use the original behavior
+                self.chunker = SemanticChunker(
+                    embedding_model=getattr(self.embedder, "id", None) or "text-embedding-3-small",
+                    chunk_size=self.chunk_size,
+                    threshold=self.similarity_threshold,
+                )
 
     def chunk(self, document: Document) -> List[Document]:
         """Split document into semantic chunks using chonkie"""

{agno-2.0.6 → agno-2.0.7}/agno/media.py

@@ -371,6 +371,8 @@ class File(BaseModel):
             "application/pdf",
             "application/json",
             "application/x-javascript",
+            "application/json",
+            "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
             "text/javascript",
             "application/x-python",
             "text/x-python",

{agno-2.0.6 → agno-2.0.7}/agno/models/base.py

@@ -196,6 +196,7 @@ class Model(ABC):
         tool_choice: Optional[Union[str, Dict[str, Any]]] = None,
         tool_call_limit: Optional[int] = None,
         run_response: Optional[RunOutput] = None,
+        send_media_to_model: bool = True,
     ) -> ModelResponse:
         """
         Generate a response from the model.
@@ -301,7 +302,11 @@ class Model(ABC):
 
                 if any(msg.images or msg.videos or msg.audio or msg.files for msg in function_call_results):
                     # Handle function call media
-                    self._handle_function_call_media(messages=messages, function_call_results=function_call_results)
+                    self._handle_function_call_media(
+                        messages=messages,
+                        function_call_results=function_call_results,
+                        send_media_to_model=send_media_to_model,
+                    )
 
                 for function_call_result in function_call_results:
                     function_call_result.log(metrics=True)
@@ -339,6 +344,7 @@ class Model(ABC):
         functions: Optional[Dict[str, Function]] = None,
         tool_choice: Optional[Union[str, Dict[str, Any]]] = None,
         tool_call_limit: Optional[int] = None,
+        send_media_to_model: bool = True,
     ) -> ModelResponse:
         """
         Generate an asynchronous response from the model.
@@ -441,7 +447,11 @@ class Model(ABC):
 
                 if any(msg.images or msg.videos or msg.audio or msg.files for msg in function_call_results):
                     # Handle function call media
-                    self._handle_function_call_media(messages=messages, function_call_results=function_call_results)
+                    self._handle_function_call_media(
+                        messages=messages,
+                        function_call_results=function_call_results,
+                        send_media_to_model=send_media_to_model,
+                    )
 
                 for function_call_result in function_call_results:
                     function_call_result.log(metrics=True)
@@ -689,6 +699,7 @@ class Model(ABC):
         tool_call_limit: Optional[int] = None,
         stream_model_response: bool = True,
         run_response: Optional[RunOutput] = None,
+        send_media_to_model: bool = True,
     ) -> Iterator[Union[ModelResponse, RunOutputEvent, TeamRunOutputEvent]]:
         """
         Generate a streaming response from the model.
@@ -778,7 +789,11 @@ class Model(ABC):
 
                 # Handle function call media
                 if any(msg.images or msg.videos or msg.audio for msg in function_call_results):
-                    self._handle_function_call_media(messages=messages, function_call_results=function_call_results)
+                    self._handle_function_call_media(
+                        messages=messages,
+                        function_call_results=function_call_results,
+                        send_media_to_model=send_media_to_model,
+                    )
 
                 for function_call_result in function_call_results:
                     function_call_result.log(metrics=True)
@@ -848,6 +863,7 @@ class Model(ABC):
         tool_call_limit: Optional[int] = None,
         stream_model_response: bool = True,
         run_response: Optional[RunOutput] = None,
+        send_media_to_model: bool = True,
     ) -> AsyncIterator[Union[ModelResponse, RunOutputEvent, TeamRunOutputEvent]]:
         """
         Generate an asynchronous streaming response from the model.
@@ -937,7 +953,11 @@ class Model(ABC):
 
                 # Handle function call media
                 if any(msg.images or msg.videos or msg.audio for msg in function_call_results):
-                    self._handle_function_call_media(messages=messages, function_call_results=function_call_results)
+                    self._handle_function_call_media(
+                        messages=messages,
+                        function_call_results=function_call_results,
+                        send_media_to_model=send_media_to_model,
+                    )
 
                 for function_call_result in function_call_results:
                     function_call_result.log(metrics=True)
@@ -1041,7 +1061,13 @@ class Model(ABC):
         if model_response_delta.extra is not None:
             if stream_data.extra is None:
                 stream_data.extra = {}
-            stream_data.extra.update(model_response_delta.extra)
+            for key in model_response_delta.extra:
+                if isinstance(model_response_delta.extra[key], list):
+                    if not stream_data.extra.get(key):
+                        stream_data.extra[key] = []
+                    stream_data.extra[key].extend(model_response_delta.extra[key])
+                else:
+                    stream_data.extra[key] = model_response_delta.extra[key]
 
         if should_yield:
             yield model_response_delta
@@ -1708,7 +1734,9 @@ class Model(ABC):
         if len(function_call_results) > 0:
             messages.extend(function_call_results)
 
-    def _handle_function_call_media(self, messages: List[Message], function_call_results: List[Message]) -> None:
+    def _handle_function_call_media(
+        self, messages: List[Message], function_call_results: List[Message], send_media_to_model: bool = True
+    ) -> None:
         """
         Handle media artifacts from function calls by adding follow-up user messages for generated media if needed.
         """
@@ -1739,9 +1767,10 @@ class Model(ABC):
                 all_files.extend(result_message.files)
                 result_message.files = None
 
-        # If we have media artifacts, add a follow-up "user" message instead of a "tool"
-        # message with the media artifacts which throws error for some models
-        if all_images or all_videos or all_audio or all_files:
+        # Only add media message if we should send media to model
+        if send_media_to_model and (all_images or all_videos or all_audio or all_files):
+            # If we have media artifacts, add a follow-up "user" message instead of a "tool"
+            # message with the media artifacts which throws error for some models
             media_message = Message(
                 role="user",
                 content="Take note of the following content",

agno-2.0.7/agno/models/llama_cpp/__init__.py

@@ -0,0 +1,5 @@
+from agno.models.llama_cpp.llama_cpp import LlamaCpp
+
+__all__ = [
+    "LlamaCpp",
+]

agno-2.0.7/agno/models/llama_cpp/llama_cpp.py

@@ -0,0 +1,22 @@
+from dataclasses import dataclass
+
+from agno.models.openai.like import OpenAILike
+
+
+@dataclass
+class LlamaCpp(OpenAILike):
+    """
+    A class for interacting with LLMs using Llama CPP.
+
+    Attributes:
+        id (str): The id of the Llama CPP model. Default is "ggml-org/gpt-oss-20b-GGUF".
+        name (str): The name of this chat model instance. Default is "LlamaCpp".
+        provider (str): The provider of the model. Default is "LlamaCpp".
+        base_url (str): The base url to which the requests are sent.
+    """
+
+    id: str = "ggml-org/gpt-oss-20b-GGUF"
+    name: str = "LlamaCpp"
+    provider: str = "LlamaCpp"
+
+    base_url: str = "http://127.0.0.1:8080/v1"

{agno-2.0.6 → agno-2.0.7}/agno/models/nexus/__init__.py

@@ -1,3 +1,3 @@
 from agno.models.nexus.nexus import Nexus
 
-__all__ = ["Nexus"]
+__all__ = ["Nexus"]

{agno-2.0.6 → agno-2.0.7}/agno/models/nexus/nexus.py

@@ -6,13 +6,12 @@ from agno.models.openai.like import OpenAILike
 @dataclass
 class Nexus(OpenAILike):
     """
-    A class for interacting with Nvidia models.
+    A class for interacting with LLMs using Nexus.
 
     Attributes:
-        id (str): The id of the Nexus model to use. Default is "nvidia/llama-3.1-nemotron-70b-instruct".
+        id (str): The id of the Nexus model to use. Default is "openai/gpt-4".
         name (str): The name of this chat model instance. Default is "Nexus"
         provider (str): The provider of the model. Default is "Nexus".
-        api_key (str): The api key to authorize request to Nexus.
         base_url (str): The base url to which the requests are sent.
     """
 
@@ -21,5 +20,3 @@ class Nexus(OpenAILike):
     provider: str = "Nexus"
 
     base_url: str = "http://localhost:8000/llm/v1/"
-
-    supports_native_structured_outputs: bool = False

{agno-2.0.6 → agno-2.0.7}/agno/os/app.py

@@ -9,7 +9,6 @@ from fastapi.responses import JSONResponse
 from fastapi.routing import APIRoute
 from rich import box
 from rich.panel import Panel
-from starlette.middleware.cors import CORSMiddleware
 from starlette.requests import Request
 
 from agno.agent.agent import Agent
@@ -37,6 +36,7 @@ from agno.os.routers.memory import get_memory_router
 from agno.os.routers.metrics import get_metrics_router
 from agno.os.routers.session import get_session_router
 from agno.os.settings import AgnoAPISettings
+from agno.os.utils import update_cors_middleware
 from agno.team.team import Team
 from agno.utils.log import logger
 from agno.utils.string import generate_id, generate_id_from_name
@@ -286,14 +286,8 @@ class AgentOS:
 
             self.fastapi_app.middleware("http")(general_exception_handler)
 
-            self.fastapi_app.add_middleware(
-                CORSMiddleware,
-                allow_origins=self.settings.cors_origin_list,  # type: ignore
-                allow_credentials=True,
-                allow_methods=["*"],
-                allow_headers=["*"],
-                expose_headers=["*"],
-            )
+        # Update CORS middleware
+        update_cors_middleware(self.fastapi_app, self.settings.cors_origin_list)  # type: ignore
 
         return self.fastapi_app
 
@@ -368,7 +362,7 @@ class AgentOS:
                 for route in self.fastapi_app.routes:
                     for conflict in conflicts:
                         if isinstance(route, APIRoute):
-                            if route.path == conflict["path"] and list(route.methods) == list(conflict["methods"]):
+                            if route.path == conflict["path"] and list(route.methods) == list(conflict["methods"]):  # type: ignore
                                 self.fastapi_app.routes.pop(self.fastapi_app.routes.index(route))
 
                 self.fastapi_app.include_router(router)

{agno-2.0.6 → agno-2.0.7}/agno/os/router.py

@@ -731,10 +731,9 @@ def get_base_router(
                 ]:
                     # Process document files
                     try:
-                        file_content = await file.read()
-                        input_files.append(
-                            FileMedia(content=file_content, filename=file.filename, mime_type=file.content_type)
-                        )
+                        input_file = process_document(file)
+                        if input_file is not None:
+                            input_files.append(input_file)
                     except Exception as e:
                         log_error(f"Error processing file {file.filename}: {e}")
                         continue

{agno-2.0.6 → agno-2.0.7}/agno/os/routers/evals/evals.py

@@ -380,7 +380,7 @@ def parse_eval_types_filter(
     eval_types: Optional[str] = Query(
         default=None,
         description="Comma-separated eval types (accuracy,performance,reliability)",
-        example="accuracy,performance",
+        examples=["accuracy,performance"],
     ),
 ) -> Optional[List[EvalType]]:
     """Parse comma-separated eval types into EvalType enums for filtering evaluation runs."""

{agno-2.0.6 → agno-2.0.7}/agno/os/routers/memory/memory.py

@@ -396,7 +396,7 @@ def parse_topics(
     topics: Optional[List[str]] = Query(
         default=None,
         description="Comma-separated list of topics to filter by",
-        example=["preferences,technical,communication_style"],
+        examples=["preferences,technical,communication_style"],
     ),
 ) -> Optional[List[str]]:
     """Parse comma-separated topics into a list for filtering memories by topic."""

{agno-2.0.6 → agno-2.0.7}/agno/os/schema.py

@@ -461,11 +461,8 @@ class TeamResponse(BaseModel):
             "stream_member_events": False,
         }
 
-        if team.model is None:
-            raise ValueError("Team model is required")
-
         team.determine_tools_for_model(
-            model=team.model,
+            model=team.model,  # type: ignore
             session=TeamSession(session_id=str(uuid4()), session_data={}),
             run_response=TeamRunOutput(run_id=str(uuid4())),
             async_mode=True,
@@ -763,6 +760,7 @@ class TeamSessionDetailSchema(BaseModel):
     session_state: Optional[dict]
     metrics: Optional[dict]
     team_data: Optional[dict]
+    chat_history: Optional[List[dict]]
     created_at: Optional[datetime]
     updated_at: Optional[datetime]
     total_tokens: Optional[int]
@@ -784,6 +782,7 @@ class TeamSessionDetailSchema(BaseModel):
             if session.session_data
             else None,
             metrics=session.session_data.get("session_metrics", {}) if session.session_data else None,
+            chat_history=[message.to_dict() for message in session.get_chat_history()],
             created_at=datetime.fromtimestamp(session.created_at, tz=timezone.utc) if session.created_at else None,
             updated_at=datetime.fromtimestamp(session.updated_at, tz=timezone.utc) if session.updated_at else None,
         )

{agno-2.0.6 → agno-2.0.7}/agno/os/utils.py

@@ -1,6 +1,7 @@
 from typing import Any, Callable, Dict, List, Optional, Union
 
-from fastapi import HTTPException, UploadFile
+from fastapi import FastAPI, HTTPException, UploadFile
+from starlette.middleware.cors import CORSMiddleware
 
 from agno.agent.agent import Agent
 from agno.db.base import BaseDb
@@ -109,27 +110,21 @@ def process_image(file: UploadFile) -> Image:
     content = file.file.read()
     if not content:
         raise HTTPException(status_code=400, detail="Empty file")
-    return Image(content=content)
+    return Image(content=content, format=extract_format(file), mime_type=file.content_type)
 
 
 def process_audio(file: UploadFile) -> Audio:
     content = file.file.read()
     if not content:
         raise HTTPException(status_code=400, detail="Empty file")
-    format = None
-    if file.filename and "." in file.filename:
-        format = file.filename.split(".")[-1].lower()
-    elif file.content_type:
-        format = file.content_type.split("/")[-1]
-
-    return Audio(content=content, format=format)
+    return Audio(content=content, format=extract_format(file), mime_type=file.content_type)
 
 
 def process_video(file: UploadFile) -> Video:
     content = file.file.read()
     if not content:
         raise HTTPException(status_code=400, detail="Empty file")
-    return Video(content=content, format=file.content_type)
+    return Video(content=content, format=extract_format(file), mime_type=file.content_type)
 
 
 def process_document(file: UploadFile) -> Optional[FileMedia]:
@@ -137,13 +132,23 @@ def process_document(file: UploadFile) -> Optional[FileMedia]:
         content = file.file.read()
         if not content:
             raise HTTPException(status_code=400, detail="Empty file")
-
-        return FileMedia(content=content, filename=file.filename, mime_type=file.content_type)
+        return FileMedia(
+            content=content, filename=file.filename, format=extract_format(file), mime_type=file.content_type
+        )
     except Exception as e:
         logger.error(f"Error processing document {file.filename}: {e}")
         return None
 
 
+def extract_format(file: UploadFile):
+    format = None
+    if file.filename and "." in file.filename:
+        format = file.filename.split(".")[-1].lower()
+    elif file.content_type:
+        format = file.content_type.split("/")[-1]
+    return format
+
+
 def format_tools(agent_tools: List[Union[Dict[str, Any], Toolkit, Function, Callable]]):
     formatted_tools = []
     if agent_tools is not None:
@@ -260,3 +265,33 @@ def _generate_schema_from_params(params: Dict[str, Any]) -> Dict[str, Any]:
         schema["required"] = required
 
     return schema
+
+
+def update_cors_middleware(app: FastAPI, new_origins: list):
+    existing_origins: List[str] = []
+
+    # TODO: Allow more options where CORS is properly merged and user can disable this behaviour
+
+    # Extract existing origins from current CORS middleware
+    for middleware in app.user_middleware:
+        if middleware.cls == CORSMiddleware:
+            if hasattr(middleware, "kwargs"):
+                existing_origins = middleware.kwargs.get("allow_origins", [])
+            break
+    # Merge origins
+    merged_origins = list(set(new_origins + existing_origins))
+    final_origins = [origin for origin in merged_origins if origin != "*"]
+
+    # Remove existing CORS
+    app.user_middleware = [m for m in app.user_middleware if m.cls != CORSMiddleware]
+    app.middleware_stack = None
+
+    # Add updated CORS
+    app.add_middleware(
+        CORSMiddleware,
+        allow_origins=final_origins,
+        allow_credentials=True,
+        allow_methods=["*"],
+        allow_headers=["*"],
+        expose_headers=["*"],
+    )

{agno-2.0.6 → agno-2.0.7}/agno/run/agent.py

@@ -96,6 +96,7 @@ class RunContentEvent(BaseAgentRunEvent):
     content: Optional[Any] = None
     content_type: str = "str"
     reasoning_content: Optional[str] = None
+    model_provider_data: Optional[Dict[str, Any]] = None
     citations: Optional[Citations] = None
     response_audio: Optional[Audio] = None  # Model audio response
     image: Optional[Image] = None  # Image attached to the response
@@ -119,6 +120,7 @@ class RunCompletedEvent(BaseAgentRunEvent):
     content_type: str = "str"
     reasoning_content: Optional[str] = None
     citations: Optional[Citations] = None
+    model_provider_data: Optional[Dict[str, Any]] = None
     images: Optional[List[Image]] = None  # Images attached to the response
     videos: Optional[List[Video]] = None  # Videos attached to the response
     audio: Optional[List[Audio]] = None  # Audio attached to the response
@@ -383,6 +385,8 @@ class RunOutput:
     reasoning_steps: Optional[List[ReasoningStep]] = None
     reasoning_messages: Optional[List[Message]] = None
 
+    model_provider_data: Optional[Dict[str, Any]] = None
+
     model: Optional[str] = None
     model_provider: Optional[str] = None
     messages: Optional[List[Message]] = None

{agno-2.0.6 → agno-2.0.7}/agno/run/team.py

@@ -98,6 +98,7 @@ class RunContentEvent(BaseTeamRunEvent):
     content: Optional[Any] = None
     content_type: str = "str"
     reasoning_content: Optional[str] = None
+    model_provider_data: Optional[Dict[str, Any]] = None
     citations: Optional[Citations] = None
     response_audio: Optional[Audio] = None  # Model audio response
     image: Optional[Image] = None  # Image attached to the response
@@ -121,6 +122,7 @@ class RunCompletedEvent(BaseTeamRunEvent):
     content_type: str = "str"
     reasoning_content: Optional[str] = None
     citations: Optional[Citations] = None
+    model_provider_data: Optional[Dict[str, Any]] = None
     images: Optional[List[Image]] = None  # Images attached to the response
     videos: Optional[List[Video]] = None  # Videos attached to the response
     audio: Optional[List[Audio]] = None  # Audio attached to the response
@@ -382,7 +384,7 @@ class TeamRunOutput:
     reasoning_content: Optional[str] = None
 
     citations: Optional[Citations] = None
-
+    model_provider_data: Optional[Dict[str, Any]] = None
     metadata: Optional[Dict[str, Any]] = None
 
     references: Optional[List[MessageReferences]] = None