agno 2.3.21__py3-none-any.whl → 2.3.23__py3-none-any.whl

agno/os/routers/knowledge/schemas.py

@@ -156,7 +156,7 @@ class VectorSearchRequestSchema(BaseModel):
     class Meta(BaseModel):
         """Inline metadata schema for pagination."""
 
-        limit: int = Field(20, description="Number of results per page", ge=1, le=100)
+        limit: int = Field(20, description="Number of results per page", ge=1)
         page: int = Field(1, description="Page number", ge=1)
 
     query: str = Field(..., description="The search query text")

agno/os/routers/memory/memory.py

@@ -256,8 +256,8 @@ def attach_routes(router: APIRouter, dbs: dict[str, list[Union[BaseDb, AsyncBase
         team_id: Optional[str] = Query(default=None, description="Filter memories by team ID"),
         topics: Optional[List[str]] = Depends(parse_topics),
         search_content: Optional[str] = Query(default=None, description="Fuzzy search within memory content"),
-        limit: Optional[int] = Query(default=20, description="Number of memories to return per page"),
-        page: Optional[int] = Query(default=1, description="Page number for pagination"),
+        limit: Optional[int] = Query(default=20, description="Number of memories to return per page", ge=1),
+        page: Optional[int] = Query(default=1, description="Page number for pagination", ge=0),
         sort_by: Optional[str] = Query(default="updated_at", description="Field to sort memories by"),
         sort_order: Optional[SortOrder] = Query(default="desc", description="Sort order (asc or desc)"),
         db_id: Optional[str] = Query(default=None, description="Database ID to query memories from"),
@@ -558,8 +558,8 @@ def attach_routes(router: APIRouter, dbs: dict[str, list[Union[BaseDb, AsyncBase
     )
     async def get_user_memory_stats(
         request: Request,
-        limit: Optional[int] = Query(default=20, description="Number of user statistics to return per page"),
-        page: Optional[int] = Query(default=1, description="Page number for pagination"),
+        limit: Optional[int] = Query(default=20, description="Number of user statistics to return per page", ge=1),
+        page: Optional[int] = Query(default=1, description="Page number for pagination", ge=0),
         db_id: Optional[str] = Query(default=None, description="Database ID to query statistics from"),
         table: Optional[str] = Query(default=None, description="Table to query statistics from"),
     ) -> PaginatedResponse[UserStatsSchema]:

agno/os/routers/session/session.py

@@ -107,8 +107,8 @@ def attach_routes(router: APIRouter, dbs: dict[str, list[Union[BaseDb, AsyncBase
         ),
         user_id: Optional[str] = Query(default=None, description="Filter sessions by user ID"),
         session_name: Optional[str] = Query(default=None, description="Filter sessions by name (partial match)"),
-        limit: Optional[int] = Query(default=20, description="Number of sessions to return per page"),
-        page: Optional[int] = Query(default=1, description="Page number for pagination"),
+        limit: Optional[int] = Query(default=20, description="Number of sessions to return per page", ge=1),
+        page: Optional[int] = Query(default=1, description="Page number for pagination", ge=0),
         sort_by: Optional[str] = Query(default="created_at", description="Field to sort sessions by"),
         sort_order: Optional[SortOrder] = Query(default="desc", description="Sort order (asc or desc)"),
         db_id: Optional[str] = Query(default=None, description="Database ID to query sessions from"),

agno/os/routers/teams/router.py

@@ -95,7 +95,7 @@ async def team_response_streamer(
         )
         yield format_sse_event(error_response)
 
-    except Exception as e:
+    except BaseException as e:
         import traceback
 
         traceback.print_exc()
@@ -169,11 +169,11 @@ def get_team_router(
         kwargs = await get_request_kwargs(request, create_team_run)
 
         if hasattr(request.state, "user_id") and request.state.user_id is not None:
-            if user_id:
+            if user_id and user_id != request.state.user_id:
                 log_warning("User ID parameter passed in both request state and kwargs, using request state")
             user_id = request.state.user_id
         if hasattr(request.state, "session_id") and request.state.session_id is not None:
-            if session_id:
+            if session_id and session_id != request.state.session_id:
                 log_warning("Session ID parameter passed in both request state and kwargs, using request state")
             session_id = request.state.session_id
         if hasattr(request.state, "session_state") and request.state.session_state is not None:
@@ -325,7 +325,7 @@ def get_team_router(
         if team is None:
             raise HTTPException(status_code=404, detail="Team not found")
 
-        cancelled = team.cancel_run(run_id=run_id)
+        cancelled = await team.acancel_run(run_id=run_id)
         if not cancelled:
             raise HTTPException(status_code=500, detail="Failed to cancel run - run not found or already completed")
 

agno/os/routers/traces/traces.py

@@ -126,8 +126,8 @@ def attach_routes(router: APIRouter, dbs: dict[str, list[Union[BaseDb, AsyncBase
             default=None,
             description="Filter traces ending before this time (ISO 8601 format with timezone, e.g., '2025-11-19T11:00:00Z' or '2025-11-19T16:30:00+05:30'). Times are converted to UTC for comparison.",
         ),
-        page: int = Query(default=1, description="Page number (1-indexed)", ge=1),
-        limit: int = Query(default=20, description="Number of traces per page", ge=1, le=100),
+        page: int = Query(default=1, description="Page number (1-indexed)", ge=0),
+        limit: int = Query(default=20, description="Number of traces per page", ge=1),
         db_id: Optional[str] = Query(default=None, description="Database ID to query traces from"),
     ):
         """Get list of traces with optional filters and pagination"""
@@ -455,7 +455,7 @@ def attach_routes(router: APIRouter, dbs: dict[str, list[Union[BaseDb, AsyncBase
             description="Filter sessions with traces created before this time (ISO 8601 format with timezone, e.g., '2025-11-19T11:00:00Z' or '2025-11-19T16:30:00+05:30'). Times are converted to UTC for comparison.",
         ),
         page: int = Query(default=1, description="Page number (1-indexed)", ge=1),
-        limit: int = Query(default=20, description="Number of sessions per page", ge=1, le=100),
+        limit: int = Query(default=20, description="Number of sessions per page", ge=1),
         db_id: Optional[str] = Query(default=None, description="Database ID to query statistics from"),
     ):
         """Get trace statistics grouped by session"""

agno/os/routers/workflows/router.py

@@ -626,11 +626,11 @@ def get_workflow_router(
         kwargs = await get_request_kwargs(request, create_workflow_run)
 
         if hasattr(request.state, "user_id") and request.state.user_id is not None:
-            if user_id:
+            if user_id and user_id != request.state.user_id:
                 log_warning("User ID parameter passed in both request state and kwargs, using request state")
             user_id = request.state.user_id
         if hasattr(request.state, "session_id") and request.state.session_id is not None:
-            if session_id:
+            if session_id and session_id != request.state.session_id:
                 log_warning("Session ID parameter passed in both request state and kwargs, using request state")
             session_id = request.state.session_id
         if hasattr(request.state, "session_state") and request.state.session_state is not None:
@@ -721,7 +721,7 @@ def get_workflow_router(
         if workflow is None:
             raise HTTPException(status_code=404, detail="Workflow not found")
 
-        cancelled = workflow.cancel_run(run_id=run_id)
+        cancelled = await workflow.acancel_run(run_id=run_id)
         if not cancelled:
             raise HTTPException(status_code=500, detail="Failed to cancel run - run not found or already completed")
 

agno/os/schema.py

@@ -549,7 +549,7 @@ class SortOrder(str, Enum):
 
 class PaginationInfo(BaseModel):
     page: int = Field(0, description="Current page number (0-indexed)", ge=0)
-    limit: int = Field(20, description="Number of items per page", ge=1, le=100)
+    limit: int = Field(20, description="Number of items per page", ge=1)
     total_pages: int = Field(0, description="Total number of pages", ge=0)
     total_count: int = Field(0, description="Total count of items", ge=0)
     search_time_ms: float = Field(0, description="Search execution time in milliseconds", ge=0)

agno/reasoning/deepseek.py

@@ -8,7 +8,17 @@ from agno.utils.log import logger
 
 
 def is_deepseek_reasoning_model(reasoning_model: Model) -> bool:
-    return reasoning_model.__class__.__name__ == "DeepSeek" and reasoning_model.id.lower() == "deepseek-reasoner"
+    """Check if the model is a DeepSeek reasoning model.
+
+    Matches:
+    - deepseek-reasoner
+    - deepseek-r1 and variants (deepseek-r1-distill-*, etc.)
+    """
+    if reasoning_model.__class__.__name__ != "DeepSeek":
+        return False
+
+    model_id = reasoning_model.id.lower()
+    return "reasoner" in model_id or "r1" in model_id
 
 
 def get_deepseek_reasoning(reasoning_agent: "Agent", messages: List[Message]) -> Optional[Message]:  # type: ignore  # noqa: F821

agno/reasoning/gemini.py

@@ -13,9 +13,13 @@ def is_gemini_reasoning_model(reasoning_model: Model) -> bool:
     if not is_gemini_class:
         return False
 
-    # Check if it's a Gemini 2.5+ model (supports thinking)
+    # Check if it's a Gemini model with thinking support
+    # - Gemini 2.5+ models support thinking
+    # - Gemini 3+ models support thinking (including DeepThink variants)
     model_id = reasoning_model.id.lower()
-    has_thinking_support = "2.5" in model_id
+    has_thinking_support = (
+        "2.5" in model_id or "3.0" in model_id or "3.5" in model_id or "deepthink" in model_id or "gemini-3" in model_id
+    )
 
     # Also check if thinking parameters are set
     # Note: thinking_budget=0 explicitly disables thinking mode per Google's API docs

agno/reasoning/groq.py

@@ -8,7 +8,12 @@ from agno.utils.log import logger
 
 
 def is_groq_reasoning_model(reasoning_model: Model) -> bool:
-    return reasoning_model.__class__.__name__ == "Groq" and "deepseek" in reasoning_model.id.lower()
+    return reasoning_model.__class__.__name__ == "Groq" and (
+        "deepseek" in reasoning_model.id.lower()
+        or "openai/gpt-oss-20b" in reasoning_model.id.lower()
+        or "openai/gpt-oss-120b" in reasoning_model.id.lower()
+        or "qwen/qwen3-32b" in reasoning_model.id.lower()
+    )
 
 
 def get_groq_reasoning(reasoning_agent: "Agent", messages: List[Message]) -> Optional[Message]:  # type: ignore  # noqa: F821
@@ -95,7 +100,7 @@ def get_groq_reasoning_stream(
     reasoning_content: str = ""
 
     try:
-        for event in reasoning_agent.run(input=messages, stream=True, stream_intermediate_steps=True):
+        for event in reasoning_agent.run(input=messages, stream=True, stream_events=True):
             if hasattr(event, "event"):
                 if event.event == RunEvent.run_content:
                     # Check for reasoning_content attribute first (native reasoning)
@@ -146,7 +151,7 @@ async def aget_groq_reasoning_stream(
     reasoning_content: str = ""
 
     try:
-        async for event in reasoning_agent.arun(input=messages, stream=True, stream_intermediate_steps=True):
+        async for event in reasoning_agent.arun(input=messages, stream=True, stream_events=True):
             if hasattr(event, "event"):
                 if event.event == RunEvent.run_content:
                     # Check for reasoning_content attribute first (native reasoning)

agno/reasoning/openai.py

@@ -21,6 +21,8 @@ def is_openai_reasoning_model(reasoning_model: Model) -> bool:
             or ("o1" in reasoning_model.id)
             or ("4.1" in reasoning_model.id)
             or ("4.5" in reasoning_model.id)
+            or ("5.1" in reasoning_model.id)
+            or ("5.2" in reasoning_model.id)
         )
     ) or (isinstance(reasoning_model, OpenAILike) and "deepseek-r1" in reasoning_model.id.lower())
 

agno/remote/base.py

@@ -2,7 +2,7 @@ import time
 from abc import abstractmethod
 from dataclasses import dataclass, field
 from datetime import date
-from typing import TYPE_CHECKING, Any, AsyncIterator, Dict, List, Optional, Sequence, Tuple, Union
+from typing import TYPE_CHECKING, Any, AsyncIterator, Dict, List, Literal, Optional, Sequence, Tuple, Union
 
 from pydantic import BaseModel
 
@@ -18,6 +18,8 @@ if TYPE_CHECKING:
     from fastapi import UploadFile
 
     from agno.client import AgentOSClient
+    from agno.client.a2a import A2AClient
+    from agno.client.a2a.schemas import AgentCard
     from agno.os.routers.evals.schemas import EvalSchema
     from agno.os.routers.knowledge.schemas import (
         ConfigResponseSchema,
@@ -64,7 +66,7 @@ class RemoteDb:
         """Create a RemoteDb instance from an AgentResponse/TeamResponse/WorkflowResponse and ConfigResponse.
 
         Args:
-            response: The agent, team, or workflow response containing the db_id.
+            db_id (str): The id of the remote database
             client: The AgentOSClient for remote operations.
             config: The ConfigResponse containing database table information.
 
@@ -344,31 +346,52 @@ class RemoteKnowledge:
 class BaseRemote:
     # Private cache for OS config with TTL: (config, timestamp)
     _cached_config: Optional[Tuple["ConfigResponse", float]] = field(default=None, init=False, repr=False)
+    # Private cache for agent card with TTL: (agent_card, timestamp)
+    _cached_agent_card: Optional[Tuple[Optional["AgentCard"], float]] = field(default=None, init=False, repr=False)
 
     def __init__(
         self,
         base_url: str,
         timeout: float = 60.0,
+        protocol: Literal["agentos", "a2a"] = "agentos",
+        a2a_protocol: Literal["json-rpc", "rest"] = "rest",
         config_ttl: float = 300.0,
     ):
         """Initialize BaseRemote for remote execution.
 
+        Supports two protocols:
+        - "agentos": Agno's proprietary AgentOS REST API (default)
+        - "a2a": A2A (Agent-to-Agent) protocol for cross-framework communication
+
         For local execution, provide agent/team/workflow instances.
         For remote execution, provide base_url.
 
         Args:
             base_url: Base URL for remote instance (e.g., "http://localhost:7777")
             timeout: Request timeout in seconds (default: 60)
+            protocol: Communication protocol - "agentos" (default) or "a2a"
+            a2a_protocol: For A2A protocol only - Whether to use JSON-RPC or REST protocol.
             config_ttl: Time-to-live for cached config in seconds (default: 300)
         """
         self.base_url = base_url.rstrip("/")
         self.timeout: float = timeout
+        self.protocol = protocol
+        self.a2a_protocol = a2a_protocol
         self.config_ttl: float = config_ttl
         self._cached_config = None
+        self._cached_agent_card = None
+
+        self.agentos_client = None
+        self.a2a_client = None
 
-        self.client = self.get_client()
+        if protocol == "agentos":
+            self.agentos_client = self.get_os_client()
+        elif protocol == "a2a":
+            self.a2a_client = self.get_a2a_client()
+        else:
+            raise ValueError(f"Invalid protocol: {protocol}")
 
-    def get_client(self) -> "AgentOSClient":
+    def get_os_client(self) -> "AgentOSClient":
         """Get an AgentOSClient for fetching remote configuration.
 
         This is used internally by AgentOS to fetch configuration from remote
@@ -384,11 +407,31 @@ class BaseRemote:
             timeout=self.timeout,
         )
 
+    def get_a2a_client(self) -> "A2AClient":
+        """Get an A2AClient for A2A protocol communication.
+
+        Returns cached client if available, otherwise creates a new one.
+        This method provides lazy initialization of the A2A client.
+
+        Returns:
+            A2AClient: Client configured for A2A protocol communication
+        """
+        from agno.client.a2a import A2AClient
+
+        return A2AClient(
+            base_url=self.base_url,
+            timeout=int(self.timeout),
+            protocol=self.a2a_protocol,
+        )
+
     @property
-    def _config(self) -> "ConfigResponse":
+    def _config(self) -> Optional["ConfigResponse"]:
         """Get the OS config from remote, cached with TTL."""
         from agno.os.schema import ConfigResponse
 
+        if self.protocol == "a2a":
+            return None
+
         current_time = time.time()
 
         # Check if cache is valid
@@ -398,15 +441,15 @@ class BaseRemote:
                 return config
 
         # Fetch fresh config
-        config: ConfigResponse = self.client.get_config()  # type: ignore
+        config: ConfigResponse = self.agentos_client.get_config()  # type: ignore
         self._cached_config = (config, current_time)
         return config
 
-    def refresh_os_config(self) -> "ConfigResponse":
+    async def refresh_os_config(self) -> "ConfigResponse":
         """Force refresh the cached OS config."""
         from agno.os.schema import ConfigResponse
 
-        config: ConfigResponse = self.client.get_config()
+        config: ConfigResponse = await self.agentos_client.aget_config()  # type: ignore
         self._cached_config = (config, time.time())
         return config
 
@@ -437,6 +480,60 @@ class BaseRemote:
             return {"Authorization": f"Bearer {auth_token}"}
         return None
 
+    def get_agent_card(self) -> Optional["AgentCard"]:
+        """Get agent card for A2A protocol agents, cached with TTL.
+
+        Fetches the agent card from the standard /.well-known/agent.json endpoint
+        to populate agent metadata (name, description, etc.) for A2A agents.
+
+        Returns None for non-A2A protocols or if the server doesn't support agent cards.
+        """
+        if self.protocol != "a2a":
+            return None
+
+        current_time = time.time()
+
+        # Check if cache is valid
+        if self._cached_agent_card is not None:
+            agent_card, cached_at = self._cached_agent_card
+            if current_time - cached_at < self.config_ttl:
+                return agent_card
+
+        try:
+            agent_card = self.a2a_client.get_agent_card()  # type: ignore
+            self._cached_agent_card = (agent_card, current_time)
+            return agent_card
+        except Exception:
+            self._cached_agent_card = (None, current_time)
+            return None
+
+    async def aget_agent_card(self) -> Optional["AgentCard"]:
+        """Get agent card for A2A protocol agents, cached with TTL.
+
+        Fetches the agent card from the standard /.well-known/agent.json endpoint
+        to populate agent metadata (name, description, etc.) for A2A agents.
+
+        Returns None for non-A2A protocols or if the server doesn't support agent cards.
+        """
+        if self.protocol != "a2a":
+            return None
+
+        current_time = time.time()
+
+        # Check if cache is valid
+        if self._cached_agent_card is not None:
+            agent_card, cached_at = self._cached_agent_card
+            if current_time - cached_at < self.config_ttl:
+                return agent_card
+
+        try:
+            agent_card = await self.a2a_client.aget_agent_card()  # type: ignore
+            self._cached_agent_card = (agent_card, current_time)
+            return agent_card
+        except Exception:
+            self._cached_agent_card = (None, current_time)
+            return None
+
     @abstractmethod
     def arun(  # type: ignore
         self,
@@ -480,5 +577,5 @@ class BaseRemote:
         raise NotImplementedError("acontinue_run method must be implemented by the subclass")
 
     @abstractmethod
-    async def cancel_run(self, run_id: str) -> bool:
+    async def acancel_run(self, run_id: str) -> bool:
         raise NotImplementedError("cancel_run method must be implemented by the subclass")

agno/skills/__init__.py

@@ -0,0 +1,17 @@
+from agno.skills.agent_skills import Skills
+from agno.skills.errors import SkillError, SkillParseError, SkillValidationError
+from agno.skills.loaders import LocalSkills, SkillLoader
+from agno.skills.skill import Skill
+from agno.skills.validator import validate_metadata, validate_skill_directory
+
+__all__ = [
+    "Skills",
+    "LocalSkills",
+    "SkillLoader",
+    "Skill",
+    "SkillError",
+    "SkillParseError",
+    "SkillValidationError",
+    "validate_metadata",
+    "validate_skill_directory",
+]