agno 2.1.2__tar.gz → 2.1.4__tar.gz

{agno-2.1.2 → agno-2.1.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agno
-Version: 2.1.2
+Version: 2.1.4
 Summary: Agno: a lightweight library for building Multi-Agent Systems
 Author-email: Ashpreet Bedi <ashpreet@agno.com>
 Project-URL: homepage, https://agno.com

{agno-2.1.2 → agno-2.1.4}/agno/agent/agent.py

@@ -957,7 +957,6 @@ class Agent:
         dependencies: Optional[Dict[str, Any]] = None,
         response_format: Optional[Union[Dict, Type[BaseModel]]] = None,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
         yield_run_response: bool = False,
         debug_mode: Optional[bool] = None,
         **kwargs: Any,
@@ -973,7 +972,8 @@ class Agent:
         6. Optional: Save output to file if save_response_to_file is set
         7. Add the RunOutput to the Agent Session
         8. Update Agent Memory
-        9. Save session to storage
+        9. Create the run completed event
+        10. Save session to storage
         """
 
         # Register run for cancellation tracking
@@ -1034,7 +1034,7 @@ class Agent:
         try:
             # Start the Run by yielding a RunStarted event
             if stream_intermediate_steps:
-                yield self._handle_event(create_run_started_event(run_response), run_response, workflow_context)
+                yield self._handle_event(create_run_started_event(run_response), run_response)
 
             # 3. Reason about the task if reasoning is enabled
             yield from self._handle_reasoning_stream(run_response=run_response, run_messages=run_messages)
@@ -1050,7 +1050,6 @@ class Agent:
                     run_messages=run_messages,
                     response_format=response_format,
                     stream_intermediate_steps=stream_intermediate_steps,
-                    workflow_context=workflow_context,
                 ):
                     raise_if_cancelled(run_response.run_id)  # type: ignore
                     yield event
@@ -1066,7 +1065,6 @@ class Agent:
                     run_messages=run_messages,
                     response_format=response_format,
                     stream_intermediate_steps=stream_intermediate_steps,
-                    workflow_context=workflow_context,
                 ):
                     raise_if_cancelled(run_response.run_id)  # type: ignore
                     if isinstance(event, RunContentEvent):
@@ -1084,7 +1082,6 @@ class Agent:
                     run_response=run_response,
                     run_messages=run_messages,
                     stream_intermediate_steps=stream_intermediate_steps,
-                    workflow_context=workflow_context,
                 ):
                     raise_if_cancelled(run_response.run_id)  # type: ignore
                     yield event
@@ -1115,9 +1112,6 @@ class Agent:
             # 5. Calculate session metrics
             self._update_session_metrics(session=session, run_response=run_response)
 
-            completed_event = self._handle_event(
-                create_run_completed_event(from_run_response=run_response), run_response, workflow_context
-            )
             # 6. Optional: Save output to file if save_response_to_file is set
             self.save_run_response_to_file(
                 run_response=run_response,
@@ -1134,7 +1128,12 @@ class Agent:
                 run_response=run_response, run_messages=run_messages, session=session, user_id=user_id
             )
 
-            # 9. Save session to storage
+            # 9. Create the run completed event
+            completed_event = self._handle_event(
+                create_run_completed_event(from_run_response=run_response), run_response
+            )
+
+            # 10. Save session to storage
             self.save_session(session=session)
 
             if stream_intermediate_steps:
@@ -1158,7 +1157,6 @@ class Agent:
             yield self._handle_event(
                 create_run_cancelled_event(from_run_response=run_response, reason=str(e)),
                 run_response,
-                workflow_context,
             )
 
             # Add the RunOutput to Agent Session even when cancelled
@@ -1302,9 +1300,6 @@ class Agent:
         )
         add_history = add_history_to_context if add_history_to_context is not None else self.add_history_to_context
 
-        # Extract workflow context from kwargs if present
-        workflow_context = kwargs.pop("workflow_context", None)
-
         # Initialize Knowledge Filters
         effective_filters = knowledge_filters
 
@@ -1380,7 +1375,6 @@ class Agent:
                         dependencies=run_dependencies,
                         response_format=response_format,
                         stream_intermediate_steps=stream_intermediate_steps,
-                        workflow_context=workflow_context,
                         yield_run_response=yield_run_response,
                         debug_mode=debug_mode,
                         **kwargs,
@@ -1651,7 +1645,6 @@ class Agent:
         dependencies: Optional[Dict[str, Any]] = None,
         response_format: Optional[Union[Dict, Type[BaseModel]]] = None,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
         yield_run_response: Optional[bool] = None,
         debug_mode: Optional[bool] = None,
         **kwargs: Any,
@@ -1667,7 +1660,8 @@ class Agent:
         6. Calculate session metrics
         7. Add RunOutput to Agent Session
         8. Update Agent Memory
-        9. Save session to storage
+        9. Create the run completed event
+        10. Save session to storage
         """
 
         # 1. Resolving here for async requirement
@@ -1731,7 +1725,7 @@ class Agent:
         try:
             # Start the Run by yielding a RunStarted event
             if stream_intermediate_steps:
-                yield self._handle_event(create_run_started_event(run_response), run_response, workflow_context)
+                yield self._handle_event(create_run_started_event(run_response), run_response)
 
             # 4. Reason about the task if reasoning is enabled
             async for item in self._ahandle_reasoning_stream(run_response=run_response, run_messages=run_messages):
@@ -1749,7 +1743,6 @@ class Agent:
                     run_messages=run_messages,
                     response_format=response_format,
                     stream_intermediate_steps=stream_intermediate_steps,
-                    workflow_context=workflow_context,
                 ):
                     raise_if_cancelled(run_response.run_id)  # type: ignore
                     yield event
@@ -1765,7 +1758,6 @@ class Agent:
                     run_messages=run_messages,
                     response_format=response_format,
                     stream_intermediate_steps=stream_intermediate_steps,
-                    workflow_context=workflow_context,
                 ):
                     raise_if_cancelled(run_response.run_id)  # type: ignore
                     if isinstance(event, RunContentEvent):
@@ -1782,7 +1774,6 @@ class Agent:
                     session=session,
                     run_response=run_response,
                     run_messages=run_messages,
-                    workflow_context=workflow_context,
                     stream_intermediate_steps=stream_intermediate_steps,
                 ):
                     raise_if_cancelled(run_response.run_id)  # type: ignore
@@ -1811,10 +1802,6 @@ class Agent:
             if run_response.metrics:
                 run_response.metrics.stop_timer()
 
-            completed_event = self._handle_event(
-                create_run_completed_event(from_run_response=run_response), run_response, workflow_context
-            )
-
             # 6. Calculate session metrics
             self._update_session_metrics(session=session, run_response=run_response)
 
@@ -1835,7 +1822,12 @@ class Agent:
             ):
                 yield event
 
-            # 9. Save session to storage
+            # 9. Create the run completed event
+            completed_event = self._handle_event(
+                create_run_completed_event(from_run_response=run_response), run_response
+            )
+
+            # 10. Save session to storage
             self.save_session(session=session)
 
             if stream_intermediate_steps:
@@ -1859,7 +1851,6 @@ class Agent:
             yield self._handle_event(
                 create_run_cancelled_event(from_run_response=run_response, reason=str(e)),
                 run_response,
-                workflow_context,
             )
 
             # Add the RunOutput to Agent Session even when cancelled
@@ -1999,9 +1990,6 @@ class Agent:
         )
         add_history = add_history_to_context if add_history_to_context is not None else self.add_history_to_context
 
-        # Extract workflow context from kwargs if present
-        workflow_context = kwargs.pop("workflow_context", None)
-
         effective_filters = knowledge_filters
         # When filters are passed manually
         if self.knowledge_filters or knowledge_filters:
@@ -2074,7 +2062,6 @@ class Agent:
                         metadata=metadata,
                         response_format=response_format,
                         stream_intermediate_steps=stream_intermediate_steps,
-                        workflow_context=workflow_context,
                         yield_run_response=yield_run_response,
                         dependencies=run_dependencies,
                         debug_mode=debug_mode,
@@ -2093,7 +2080,6 @@ class Agent:
                         metadata=metadata,
                         response_format=response_format,
                         stream_intermediate_steps=stream_intermediate_steps,
-                        workflow_context=workflow_context,
                         yield_run_response=yield_run_response,
                         dependencies=run_dependencies,
                         debug_mode=debug_mode,
@@ -2484,11 +2470,12 @@ class Agent:
         Steps:
         1. Handle any updated tools
         2. Generate a response from the Model
-        3. Update Agent Memory
-        4. Calculate session metrics
-        5. Save output to file if save_response_to_file is set
-        6. Add RunOutput to Agent Session
-        7. Save session to storage
+        3. Calculate session metrics
+        4. Save output to file if save_response_to_file is set
+        5. Add the run to memory
+        6. Update Agent Memory
+        7. Create the run completed event
+        8. Save session to storage
         """
 
         if dependencies is not None:
@@ -2523,8 +2510,6 @@ class Agent:
 
         run_response.status = RunStatus.completed
 
-        completed_event = self._handle_event(create_run_completed_event(run_response), run_response)
-
         # Set the run duration
         if run_response.metrics:
             run_response.metrics.stop_timer()
@@ -2542,7 +2527,10 @@ class Agent:
             run_response=run_response, run_messages=run_messages, session=session, user_id=user_id
         )
 
-        # 7. Save session to storage
+        # 7. Create the run completed event
+        completed_event = self._handle_event(create_run_completed_event(run_response), run_response)
+
+        # 8. Save session to storage
         self.save_session(session=session)
 
         if stream_intermediate_steps:
@@ -2882,11 +2870,12 @@ class Agent:
         Steps:
         1. Handle any updated tools
         2. Generate a response from the Model
-        3. Add the run to memory
-        4. Update Agent Memory
-        5. Calculate session metrics
-        6. Save output to file if save_response_to_file is set
-        7. Save session to storage
+        3. Calculate session metrics
+        4. Save output to file if save_response_to_file is set
+        5. Add the run to memory
+        6. Update Agent Memory
+        7. Create the run completed event
+        8. Save session to storage
         """
         # Resolve dependencies
         if dependencies is not None:
@@ -2923,8 +2912,6 @@ class Agent:
 
         run_response.status = RunStatus.completed
 
-        completed_event = self._handle_event(create_run_completed_event(run_response), run_response)
-
         # Set the run duration
         if run_response.metrics:
             run_response.metrics.stop_timer()
@@ -2943,7 +2930,10 @@ class Agent:
         ):
             yield event
 
-        # 7. Save session to storage
+        # 7. Create the run completed event
+        completed_event = self._handle_event(create_run_completed_event(run_response), run_response)
+
+        # 8. Save session to storage
         self.save_session(session=session)
 
         if stream_intermediate_steps:
@@ -3608,7 +3598,6 @@ class Agent:
         run_messages: RunMessages,
         response_format: Optional[Union[Dict, Type[BaseModel]]] = None,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
     ) -> Iterator[RunOutputEvent]:
         self.model = cast(Model, self.model)
 
@@ -3642,7 +3631,6 @@ class Agent:
                 reasoning_state=reasoning_state,
                 parse_structured_output=self.should_parse_structured_output,
                 stream_intermediate_steps=stream_intermediate_steps,
-                workflow_context=workflow_context,
             )
 
         # Determine reasoning completed
@@ -3685,7 +3673,6 @@ class Agent:
         run_messages: RunMessages,
         response_format: Optional[Union[Dict, Type[BaseModel]]] = None,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
     ) -> AsyncIterator[RunOutputEvent]:
         self.model = cast(Model, self.model)
 
@@ -3721,7 +3708,6 @@ class Agent:
                 reasoning_state=reasoning_state,
                 parse_structured_output=self.should_parse_structured_output,
                 stream_intermediate_steps=stream_intermediate_steps,
-                workflow_context=workflow_context,
             ):
                 yield event
 
@@ -3766,7 +3752,6 @@ class Agent:
         reasoning_state: Optional[Dict[str, Any]] = None,
         parse_structured_output: bool = False,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
     ) -> Iterator[RunOutputEvent]:
         if isinstance(model_response_event, tuple(get_args(RunOutputEvent))) or isinstance(
             model_response_event, tuple(get_args(TeamRunOutputEvent))
@@ -3830,7 +3815,6 @@ class Agent:
                             content_type=content_type,
                         ),
                         run_response,
-                        workflow_context=workflow_context,
                     )
                 elif (
                     model_response_event.content is not None
@@ -3849,7 +3833,6 @@ class Agent:
                             model_provider_data=model_response_event.provider_data,
                         ),
                         run_response,
-                        workflow_context=workflow_context,
                     )
 
                 # Process audio
@@ -3910,7 +3893,6 @@ class Agent:
                             response_audio=run_response.response_audio,
                         ),
                         run_response,
-                        workflow_context=workflow_context,
                     )
 
                 if model_response_event.images is not None:
@@ -3920,7 +3902,6 @@ class Agent:
                             image=model_response_event.images[-1],
                         ),
                         run_response,
-                        workflow_context=workflow_context,
                     )
 
                     if model_response.images is None:
@@ -6892,7 +6873,6 @@ class Agent:
         run_response: RunOutput,
         run_messages: RunMessages,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
     ):
         """Parse the model response using the output model."""
         from agno.utils.events import (
@@ -6916,7 +6896,6 @@ class Agent:
                 run_response=run_response,
                 model_response=model_response,
                 model_response_event=model_response_event,
-                workflow_context=workflow_context,
                 stream_intermediate_steps=stream_intermediate_steps,
             )
 
@@ -6945,7 +6924,6 @@ class Agent:
         run_response: RunOutput,
         run_messages: RunMessages,
         stream_intermediate_steps: bool = False,
-        workflow_context: Optional[Dict] = None,
     ):
         """Parse the model response using the output model."""
         from agno.utils.events import (
@@ -6971,7 +6949,6 @@ class Agent:
                 run_response=run_response,
                 model_response=model_response,
                 model_response_event=model_response_event,
-                workflow_context=workflow_context,
                 stream_intermediate_steps=stream_intermediate_steps,
             ):
                 yield event
@@ -6986,14 +6963,7 @@ class Agent:
         # Update the RunResponse metrics
         run_response.metrics = self._calculate_run_metrics(messages_for_run_response)
 
-    def _handle_event(self, event: RunOutputEvent, run_response: RunOutput, workflow_context: Optional[Dict] = None):
-        if workflow_context:
-            event.workflow_id = workflow_context.get("workflow_id")
-            event.workflow_run_id = workflow_context.get("workflow_run_id")
-            event.step_id = workflow_context.get("step_id")
-            event.step_name = workflow_context.get("step_name")
-            event.step_index = workflow_context.get("step_index")
-
+    def _handle_event(self, event: RunOutputEvent, run_response: RunOutput):
         # We only store events that are not run_response_content events
         events_to_skip = [event.value for event in self.events_to_skip] if self.events_to_skip else []
         if self.store_events and event.event not in events_to_skip:

{agno-2.1.2 → agno-2.1.4}/agno/db/dynamo/dynamo.py

@@ -1454,17 +1454,17 @@ class DynamoDb(BaseDb):
         """
         import json
 
-        item = {}
+        item: Dict[str, Any] = {}
         for key, value in data.items():
             if value is not None:
                 if isinstance(value, bool):
-                    item[key] = {"BOOL": str(value)}
+                    item[key] = {"BOOL": value}
                 elif isinstance(value, (int, float)):
                     item[key] = {"N": str(value)}
                 elif isinstance(value, str):
                     item[key] = {"S": str(value)}
                 elif isinstance(value, (dict, list)):
-                    item[key] = {"S": json.dumps(str(value))}
+                    item[key] = {"S": json.dumps(value)}
                 else:
                     item[key] = {"S": str(value)}
         return item
@@ -1803,14 +1803,16 @@ class DynamoDb(BaseDb):
 
             if filter_type is not None:
                 if filter_type == EvalFilterType.AGENT:
-                    filter_expressions.append("agent_id IS NOT NULL")
+                    filter_expressions.append("attribute_exists(agent_id)")
                 elif filter_type == EvalFilterType.TEAM:
-                    filter_expressions.append("team_id IS NOT NULL")
+                    filter_expressions.append("attribute_exists(team_id)")
                 elif filter_type == EvalFilterType.WORKFLOW:
-                    filter_expressions.append("workflow_id IS NOT NULL")
+                    filter_expressions.append("attribute_exists(workflow_id)")
 
             if filter_expressions:
                 scan_kwargs["FilterExpression"] = " AND ".join(filter_expressions)
+
+            if expression_values:
                 scan_kwargs["ExpressionAttributeValues"] = expression_values  # type: ignore
 
             # Execute scan

{agno-2.1.2 → agno-2.1.4}/agno/db/dynamo/schemas.py

@@ -73,6 +73,7 @@ USER_MEMORY_TABLE_SCHEMA = {
         {"AttributeName": "user_id", "AttributeType": "S"},
         {"AttributeName": "agent_id", "AttributeType": "S"},
         {"AttributeName": "team_id", "AttributeType": "S"},
+        {"AttributeName": "workflow_id", "AttributeType": "S"},
         {"AttributeName": "updated_at", "AttributeType": "S"},
     ],
     "GlobalSecondaryIndexes": [
@@ -123,7 +124,6 @@ EVAL_TABLE_SCHEMA = {
     "AttributeDefinitions": [
         {"AttributeName": "run_id", "AttributeType": "S"},
         {"AttributeName": "eval_type", "AttributeType": "S"},
-        {"AttributeName": "eval_input", "AttributeType": "S"},
         {"AttributeName": "agent_id", "AttributeType": "S"},
         {"AttributeName": "team_id", "AttributeType": "S"},
         {"AttributeName": "workflow_id", "AttributeType": "S"},
@@ -176,18 +176,9 @@ KNOWLEDGE_TABLE_SCHEMA = {
     "KeySchema": [{"AttributeName": "id", "KeyType": "HASH"}],
     "AttributeDefinitions": [
         {"AttributeName": "id", "AttributeType": "S"},
-        {"AttributeName": "name", "AttributeType": "S"},
-        {"AttributeName": "description", "AttributeType": "S"},
-        {"AttributeName": "metadata", "AttributeType": "S"},
         {"AttributeName": "type", "AttributeType": "S"},
-        {"AttributeName": "size", "AttributeType": "N"},
-        {"AttributeName": "linked_to", "AttributeType": "S"},
-        {"AttributeName": "access_count", "AttributeType": "N"},
         {"AttributeName": "status", "AttributeType": "S"},
-        {"AttributeName": "status_message", "AttributeType": "S"},
         {"AttributeName": "created_at", "AttributeType": "N"},
-        {"AttributeName": "updated_at", "AttributeType": "N"},
-        {"AttributeName": "external_id", "AttributeType": "S"},
     ],
     "GlobalSecondaryIndexes": [
         {

{agno-2.1.2 → agno-2.1.4}/agno/db/dynamo/utils.py

@@ -23,7 +23,7 @@ def serialize_to_dynamo_item(data: Dict[str, Any]) -> Dict[str, Any]:
         A DynamoDB-ready dict with the serialized data
 
     """
-    item = {}
+    item: Dict[str, Any] = {}
     for key, value in data.items():
         if value is not None:
             if isinstance(value, (int, float)):
@@ -31,7 +31,7 @@ def serialize_to_dynamo_item(data: Dict[str, Any]) -> Dict[str, Any]:
             elif isinstance(value, str):
                 item[key] = {"S": value}
             elif isinstance(value, bool):
-                item[key] = {"BOOL": str(value)}
+                item[key] = {"BOOL": value}
             elif isinstance(value, (dict, list)):
                 item[key] = {"S": json.dumps(value)}
             else:

{agno-2.1.2 → agno-2.1.4}/agno/db/firestore/firestore.py

@@ -682,9 +682,7 @@ class FirestoreDb(BaseDb):
             log_error(f"Error deleting memories: {e}")
             raise e
 
-    def get_all_memory_topics(
-        self, create_collection_if_not_found: Optional[bool] = True
-    ) -> List[str]:
+    def get_all_memory_topics(self, create_collection_if_not_found: Optional[bool] = True) -> List[str]:
         """Get all memory topics from the database.
 
         Returns:

{agno-2.1.2 → agno-2.1.4}/agno/db/gcs_json/gcs_json_db.py

@@ -472,7 +472,8 @@ class GcsJsonDb(BaseDb):
 
             # Filter out the memory, with optional user_id verification
             memories = [
-                m for m in memories
+                m
+                for m in memories
                 if not (m.get("memory_id") == memory_id and (user_id is None or m.get("user_id") == user_id))
             ]
 
@@ -499,7 +500,8 @@ class GcsJsonDb(BaseDb):
 
             # Filter out memories, with optional user_id verification
             memories = [
-                m for m in memories
+                m
+                for m in memories
                 if not (m.get("memory_id") in memory_ids and (user_id is None or m.get("user_id") == user_id))
             ]
 
@@ -642,7 +644,7 @@ class GcsJsonDb(BaseDb):
                         user_stats[memory_user_id] = {
                             "user_id": memory_user_id,
                             "total_memories": 0,
-                            "last_memory_updated_at": 0
+                            "last_memory_updated_at": 0,
                         }
                     user_stats[memory_user_id]["total_memories"] += 1
                     updated_at = memory.get("updated_at", 0)

{agno-2.1.2 → agno-2.1.4}/agno/db/in_memory/in_memory_db.py

@@ -359,8 +359,7 @@ class InMemoryDb(BaseDb):
             # If user_id is provided, verify ownership before deleting
             if user_id is not None:
                 self._memories = [
-                    m for m in self._memories
-                    if not (m.get("memory_id") == memory_id and m.get("user_id") == user_id)
+                    m for m in self._memories if not (m.get("memory_id") == memory_id and m.get("user_id") == user_id)
                 ]
             else:
                 self._memories = [m for m in self._memories if m.get("memory_id") != memory_id]
@@ -388,8 +387,7 @@ class InMemoryDb(BaseDb):
             # If user_id is provided, verify ownership before deleting
             if user_id is not None:
                 self._memories = [
-                    m for m in self._memories
-                    if not (m.get("memory_id") in memory_ids and m.get("user_id") == user_id)
+                    m for m in self._memories if not (m.get("memory_id") in memory_ids and m.get("user_id") == user_id)
                 ]
             else:
                 self._memories = [m for m in self._memories if m.get("memory_id") not in memory_ids]
@@ -532,7 +530,11 @@ class InMemoryDb(BaseDb):
 
                 if memory_user_id:
                     if memory_user_id not in user_stats:
-                        user_stats[memory_user_id] = {"user_id": memory_user_id, "total_memories": 0, "last_memory_updated_at": 0}
+                        user_stats[memory_user_id] = {
+                            "user_id": memory_user_id,
+                            "total_memories": 0,
+                            "last_memory_updated_at": 0,
+                        }
                     user_stats[memory_user_id]["total_memories"] += 1
                     updated_at = memory.get("updated_at", 0)
                     if updated_at > user_stats[memory_user_id]["last_memory_updated_at"]:

{agno-2.1.2 → agno-2.1.4}/agno/knowledge/embedder/fastembed.py

@@ -23,7 +23,7 @@ class FastEmbedEmbedder(Embedder):
     """Using BAAI/bge-small-en-v1.5 model, more models available: https://qdrant.github.io/fastembed/examples/Supported_Models/"""
 
     id: str = "BAAI/bge-small-en-v1.5"
-    dimensions: int = 384
+    dimensions: Optional[int] = 384
 
     def get_embedding(self, text: str) -> List[float]:
         model = TextEmbedding(model_name=self.id)

{agno-2.1.2 → agno-2.1.4}/agno/models/aws/bedrock.py

@@ -360,7 +360,7 @@ class AwsBedrock(Model):
             formatted_messages, system_message = self._format_messages(messages)
 
             tool_config = None
-            if tools: 
+            if tools:
                 tool_config = {"tools": self._format_tools_for_request(tools)}
 
             body = {

agno-2.1.4/agno/models/openrouter/openrouter.py

@@ -0,0 +1,66 @@
+from dataclasses import dataclass, field
+from os import getenv
+from typing import Any, Dict, List, Optional, Type, Union
+
+from pydantic import BaseModel
+
+from agno.models.openai.like import OpenAILike
+from agno.run.agent import RunOutput
+
+
+@dataclass
+class OpenRouter(OpenAILike):
+    """
+    A class for using models hosted on OpenRouter.
+
+    Attributes:
+        id (str): The model id. Defaults to "gpt-4o".
+        name (str): The model name. Defaults to "OpenRouter".
+        provider (str): The provider name. Defaults to "OpenRouter".
+        api_key (Optional[str]): The API key.
+        base_url (str): The base URL. Defaults to "https://openrouter.ai/api/v1".
+        max_tokens (int): The maximum number of tokens. Defaults to 1024.
+        fallback_models (Optional[List[str]]): List of fallback model IDs to use if the primary model
+            fails due to rate limits, timeouts, or unavailability. OpenRouter will automatically try
+            these models in order. Example: ["anthropic/claude-sonnet-4", "deepseek/deepseek-r1"]
+    """
+
+    id: str = "gpt-4o"
+    name: str = "OpenRouter"
+    provider: str = "OpenRouter"
+
+    api_key: Optional[str] = field(default_factory=lambda: getenv("OPENROUTER_API_KEY"))
+    base_url: str = "https://openrouter.ai/api/v1"
+    max_tokens: int = 1024
+    models: Optional[List[str]] = None  # Dynamic model routing https://openrouter.ai/docs/features/model-routing
+
+    def get_request_params(
+        self,
+        response_format: Optional[Union[Dict, Type[BaseModel]]] = None,
+        tools: Optional[List[Dict[str, Any]]] = None,
+        tool_choice: Optional[Union[str, Dict[str, Any]]] = None,
+        run_response: Optional[RunOutput] = None,
+    ) -> Dict[str, Any]:
+        """
+        Returns keyword arguments for API requests, including fallback models configuration.
+
+        Returns:
+            Dict[str, Any]: A dictionary of keyword arguments for API requests.
+        """
+        # Get base request params from parent class
+        request_params = super().get_request_params(
+            response_format=response_format, tools=tools, tool_choice=tool_choice, run_response=run_response
+        )
+
+        # Add fallback models to extra_body if specified
+        if self.models:
+            # Get existing extra_body or create new dict
+            extra_body = request_params.get("extra_body") or {}
+
+            # Merge fallback models into extra_body
+            extra_body["models"] = self.models
+
+            # Update request params
+            request_params["extra_body"] = extra_body
+
+        return request_params