Add DeepInfra model with Reasoning support for OpenAILike deepseek models

cookbook/agent_concepts/reasoning/deepinfra/9_11_or_9_9.py

@@ -0,0 +1,45 @@
+"""
+Be sure to set your DEEPSEEK_API_KEY environment variable before running this script.
+"""
+
+from agno.agent import Agent
+from agno.cli.console import console
+from agno.models.anthropic import Claude
+from agno.models.deepinfra import DeepInfra
+from agno.models.openai import OpenAIChat
+
+task = "9.11 and 9.9 -- which is bigger?"
+
+regular_agent_claude = Agent(model=Claude("claude-3-5-sonnet-20241022"))
+reasoning_agent_claude = Agent(
+    model=Claude("claude-3-5-sonnet-20241022"),
+    reasoning_model=DeepInfra(
+        id="deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
+        temperature=0.6,
+        max_tokens=1024,
+        top_p=0.95,
+    ),
+)
+
+regular_agent_openai = Agent(model=OpenAIChat(id="gpt-4o"))
+reasoning_agent_openai = Agent(
+    model=OpenAIChat(id="gpt-4o-mini"),
+    reasoning_model=DeepInfra(
+        id="deepseek-ai/DeepSeek-R1-Distill-Llama-70B",
+        temperature=0.6,
+        max_tokens=1024,
+        top_p=0.95,
+    ),
+)
+
+console.rule("[bold blue]Regular Claude Agent[/bold blue]")
+regular_agent_claude.print_response(task, stream=True)
+
+console.rule("[bold green]Claude Reasoning Agent[/bold green]")
+reasoning_agent_claude.print_response(task, stream=True)
+
+console.rule("[bold red]Regular OpenAI Agent[/bold red]")
+regular_agent_openai.print_response(task, stream=True)
+
+console.rule("[bold yellow]OpenAI Reasoning Agent[/bold yellow]")
+reasoning_agent_openai.print_response(task, stream=True)

cookbook/models/deepinfra/README.md

@@ -0,0 +1,72 @@
+# DeepInfra Cookbook
+
+> Note: Fork and clone this repository if needed
+
+> Note: DeepInfra does not appear to include models that support structured outputs.
+
+### 1. Create and activate a virtual environment
+
+```shell
+python3 -m venv ~/.venvs/aienv
+source ~/.venvs/aienv/bin/activate
+```
+
+### 2. Export your `DEEPINFRA_API_KEY`
+
+```shell
+export DEEPINFRA_API_KEY=***
+```
+
+### 3. Install libraries
+
+```shell
+pip install -U openai duckduckgo-search agno
+```
+
+### 4. Run basic Agent
+
+- Streaming on
+
+```shell
+python cookbook/models/deepinfra/basic_stream.py
+```
+
+- Streaming off
+
+```shell
+python cookbook/models/deepinfra/basic.py
+```
+
+### 5. Run Async Agent
+
+- Streaming on
+
+```shell
+python cookbook/models/deepinfra/async_basic_stream.py
+```
+
+- Streaming off
+
+```shell
+python cookbook/models/deepinfra/async_basic.py
+```
+
+### 6. Run Agent with Tools
+
+- DuckDuckGo Search
+
+```shell
+python cookbook/models/deepinfra/tool_use.py
+```
+
+- Async DuckDuckGo Search
+
+```shell
+python cookbook/models/deepinfra/async_tool_use.py
+```
+
+### 6. Run Agent that returns JSON output defined by the response model
+
+```shell
+python cookbook/models/deepinfra/json_output.py
+```

cookbook/models/deepinfra/async_basic.py

@@ -0,0 +1,18 @@
+import asyncio
+
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+
+agent = Agent(
+    model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"),
+    markdown=True,
+)
+
+# Get the response in a variable
+# def run_async() -> RunResponse:
+#     return agent.arun("Share a 2 sentence horror story")
+# response = asyncio.run(run_async())
+# print(response.content)
+
+# Print the response in the terminal
+asyncio.run(agent.aprint_response("Share a 2 sentence horror story"))

cookbook/models/deepinfra/async_basic_stream.py

@@ -0,0 +1,13 @@
+import asyncio
+from typing import Iterator  # noqa
+
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+
+agent = Agent(
+    model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"),
+    markdown=True,
+)
+
+# Print the response in the terminal
+asyncio.run(agent.aprint_response("Share a 2 sentence horror story", stream=True))

cookbook/models/deepinfra/async_tool_use.py

@@ -0,0 +1,16 @@
+"""Run `pip install duckduckgo-search` to install dependencies."""
+
+import asyncio
+
+from agno.agent import Agent  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+from agno.tools.duckduckgo import DuckDuckGoTools  # noqa
+
+agent = Agent(
+    model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"),
+    tools=[DuckDuckGoTools()],
+    show_tool_calls=True,
+    markdown=True,
+)
+
+asyncio.run(agent.aprint_response("What's the latest news about AI?", stream=True))

cookbook/models/deepinfra/basic.py

@@ -0,0 +1,15 @@
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+
+
+agent = Agent(
+    model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"),
+    markdown=True,
+)
+
+# Get the response in a variable
+# run: RunResponse = agent.run("Share a 2 sentence horror story")
+# print(run.content)
+
+# Print the response in the terminal
+agent.print_response("Share a 2 sentence horror story")

cookbook/models/deepinfra/basic_stream.py

@@ -0,0 +1,19 @@
+from typing import Iterator  # noqa
+
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+
+agent = Agent(
+    model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"),
+    markdown=True,
+)
+
+# Get the response in a variable
+run_response: Iterator[RunResponse] = agent.run(
+    "Share a 2 sentence horror story", stream=True
+)
+for chunk in run_response:
+    print(chunk.content)
+
+# Print the response in the terminal
+agent.print_response("Share a 2 sentence horror story", stream=True)

cookbook/models/deepinfra/json_output.py

@@ -0,0 +1,39 @@
+from typing import List
+
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+from pydantic import BaseModel, Field
+from rich.pretty import pprint  # noqa
+
+
+class MovieScript(BaseModel):
+    setting: str = Field(
+        ..., description="Provide a nice setting for a blockbuster movie."
+    )
+    ending: str = Field(
+        ...,
+        description="Ending of the movie. If not available, provide a happy ending.",
+    )
+    genre: str = Field(
+        ...,
+        description="Genre of the movie. If not available, select action, thriller or romantic comedy.",
+    )
+    name: str = Field(..., description="Give a name to this movie")
+    characters: List[str] = Field(..., description="Name of characters for this movie.")
+    storyline: str = Field(
+        ..., description="3 sentence storyline for the movie. Make it exciting!"
+    )
+
+
+# Agent that uses JSON mode
+agent = Agent(
+    model=DeepInfra(id="microsoft/phi-4"),
+    description="You write movie scripts.",
+    response_model=MovieScript,
+)
+
+# Get the response in a variable
+# response: RunResponse = agent.run("New York")
+# pprint(response.content)
+
+agent.print_response("New York")

cookbook/models/deepinfra/sync_basic.py

@@ -0,0 +1,11 @@
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+
+agent = Agent(model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"), markdown=True)
+
+# Get the response in a variable
+# run: RunResponse = agent.run("Share a 2 sentence horror story")
+# print(run.content)
+
+# Print the response in the terminal
+agent.print_response("Share a 2 sentence horror story")

cookbook/models/deepinfra/tool_use.py

@@ -0,0 +1,14 @@
+"""Run `pip install duckduckgo-search` to install dependencies."""
+
+from agno.agent import Agent  # noqa
+from agno.models.deepinfra import DeepInfra  # noqa
+from agno.tools.duckduckgo import DuckDuckGoTools  # noqa
+
+agent = Agent(
+    model=DeepInfra(id="meta-llama/Llama-2-70b-chat-hf"),
+    tools=[DuckDuckGoTools()],
+    show_tool_calls=True,
+    markdown=True,
+)
+
+agent.print_response("Whats happening in France?", stream=True)

libs/agno/agno/agent/agent.py

@@ -31,6 +31,7 @@
 from agno.memory.agent import AgentMemory, AgentRun
 from agno.models.base import Model
 from agno.models.message import Message, MessageReferences
+from agno.models.openai.like import OpenAILike
 from agno.models.response import ModelResponse, ModelResponseEvent
 from agno.reasoning.step import NextAction, ReasoningStep, ReasoningSteps
 from agno.run.messages import RunMessages
@@ -2768,8 +2769,10 @@ def reason(self, run_messages: RunMessages) -> Iterator[RunResponse]:
                     reasoning_steps=[ReasoningStep(result=groq_reasoning_message.content)],
                     reasoning_agent_messages=[groq_reasoning_message],
                 )
-            # Use o-3 for reasoning
-            elif reasoning_model.__class__.__name__ == "OpenAIChat" and reasoning_model.id.startswith("o3"):
+            # Use o-3 or OpenAILike with deepseek model for reasoning
+            elif (reasoning_model.__class__.__name__ == "OpenAIChat" and reasoning_model.id.startswith("o3")) or (
+                isinstance(reasoning_model, OpenAILike) and "deepseek" in reasoning_model.id.lower()
+            ):
                 from agno.reasoning.openai import get_openai_reasoning, get_openai_reasoning_agent
 
                 openai_reasoning_agent = self.reasoning_agent or get_openai_reasoning_agent(
@@ -2948,7 +2951,10 @@ async def areason(self, run_messages: RunMessages) -> Any:
                     reasoning_agent_messages=[groq_reasoning_message],
                 )
             # Use o-3 for reasoning
-            elif reasoning_model.__class__.__name__ == "OpenAIChat" and reasoning_model.id.startswith("o"):
+            elif (reasoning_model.__class__.__name__ == "OpenAIChat" and reasoning_model.id.startswith("o3")) or (
+                isinstance(reasoning_model, OpenAILike) and "deepseek" in reasoning_model.id.lower()
+            ):
+                # elif reasoning_model.__class__.__name__ == "OpenAIChat" and reasoning_model.id.startswith("o"):
                 from agno.reasoning.openai import aget_openai_reasoning, get_openai_reasoning_agent
 
                 openai_reasoning_agent = self.reasoning_agent or get_openai_reasoning_agent(

libs/agno/agno/models/deepinfra/__init__.py

@@ -0,0 +1 @@
+from agno.models.deepinfra.deepinfra import DeepInfra

libs/agno/agno/models/deepinfra/deepinfra.py

@@ -0,0 +1,28 @@
+from dataclasses import dataclass
+from os import getenv
+from typing import Optional
+
+from agno.models.openai.like import OpenAILike
+
+
+@dataclass
+class DeepInfra(OpenAILike):
+    """
+    A class for interacting with DeepInfra models.
+
+    For more information, see: https://deepinfra.com/docs/
+
+    Attributes:
+        id (str): The id of the Nvidia model to use. Default is "meta-llama/Llama-2-70b-chat-hf".
+        name (str): The name of this chat model instance. Default is "Nvidia"
+        provider (str): The provider of the model. Default is "Nvidia".
+        api_key (str): The api key to authorize request to Nvidia.
+        base_url (str): The base url to which the requests are sent.
+    """
+
+    id: str = "meta-llama/Llama-2-70b-chat-hf"
+    name: str = "DeepInfra"
+    provider: str = "DeepInfra " + id
+
+    api_key: Optional[str] = getenv("DEEPINFRA_API_KEY", None)
+    base_url: str = "https://api.deepinfra.com/v1/openai"