[feat] Model and vector supplier increases Baidu Qianfan

cookbook/models/qianfan/README.md

@@ -0,0 +1,45 @@
+# Qianfan Cookbook
+
+> Note: Fork and clone this repository if needed
+
+### 1. Create and activate a virtual environment
+
+```shell
+python3 -m venv ~/.venvs/aienv
+source ~/.venvs/aienv/bin/activate
+```
+
+### 2. Export your `QIANFAN_API_KEY`
+
+```shell
+export QIANFAN_API_KEY=***
+```
+
+### 3. Install libraries
+
+```shell
+pip install -U openai agno
+```
+
+### 4. Run basic Agent
+
+- Streaming on
+
+```shell
+python cookbook/models/qianfan/basic_stream.py
+```
+
+- Streaming off
+
+```shell
+python cookbook/models/qianfan/basic.py
+```
+
+### 5. Run Agent with Tools
+
+
+- DuckDuckGo search
+
+```shell
+python cookbook/models/qianfan/tool_use.py
+```

cookbook/models/qianfan/async_basic.py

@@ -0,0 +1,12 @@
+"""
+Basic async example using Qianfan.
+"""
+
+import asyncio
+
+from agno.agent import Agent
+from agno.models.qianfan import Qianfan
+
+agent = Agent(model=Qianfan(id="deepseek-v3"), markdown=True)
+
+asyncio.run(agent.aprint_response("Share a 2 sentence horror story"))

cookbook/models/qianfan/async_basic_stream.py

@@ -0,0 +1,12 @@
+"""
+Basic streaming async example using Qianfan.
+"""
+
+import asyncio
+
+from agno.agent import Agent
+from agno.models.qianfan import Qianfan
+
+agent = Agent(model=Qianfan(id="deepseek-v3"), markdown=True)
+
+asyncio.run(agent.aprint_response("Share a 2 sentence horror story", stream=True))

cookbook/models/qianfan/async_tool_use.py

@@ -0,0 +1,18 @@
+"""
+Async example using Mistral with tool calls.
+"""
+
+import asyncio
+
+from agno.agent import Agent
+from agno.models.qianfan import Qianfan
+from agno.tools.duckduckgo import DuckDuckGoTools
+
+agent = Agent(
+    model=Qianfan(id="deepseek-v3"),
+    tools=[DuckDuckGoTools()],
+    show_tool_calls=True,
+    markdown=True,
+)
+
+asyncio.run(agent.aprint_response("Whats happening in France?", stream=True))

cookbook/models/qianfan/basic.py

@@ -0,0 +1,11 @@
+from agno.agent import Agent, RunResponse  # noqa
+from agno.models.qianfan import Qianfan
+
+agent = Agent(model=Qianfan(id="deepseek-v3"), 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/qianfan/basic_stream.py

@@ -0,0 +1,12 @@
+from agno.agent import Agent, RunResponseEvent  # noqa
+from agno.models.qianfan import Qianfan
+
+agent = Agent(model=Qianfan(id="deepseek-v3"), markdown=True)
+
+# Get the response in a variable
+# run_response: Iterator[RunResponseEvent] = 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/qianfan/tool_use.py

@@ -0,0 +1,16 @@
+"""Run `pip install duckduckgo-search` to install dependencies."""
+
+import os
+
+from agno.agent import Agent
+from agno.models.qianfan import Qianfan
+from agno.tools.duckduckgo import DuckDuckGoTools
+
+agent = Agent(
+    model=Qianfan(id="deepseek-v3"),
+    tools=[DuckDuckGoTools()],
+    show_tool_calls=True,
+    markdown=True,
+)
+
+agent.print_response("Whats happening in France?", stream=True)

libs/agno/agno/embedder/google.py

@@ -35,7 +35,7 @@ def client(self):
 
         _client_params: Dict[str, Any] = {}
         vertexai = self.vertexai or getenv("GOOGLE_GENAI_USE_VERTEXAI", "false").lower() == "true"
-        
+
         if not vertexai:
             self.api_key = self.api_key or getenv("GOOGLE_API_KEY")
             if not self.api_key:
@@ -46,9 +46,9 @@ def client(self):
             _client_params["vertexai"] = True
             _client_params["project"] = self.project_id or getenv("GOOGLE_CLOUD_PROJECT")
             _client_params["location"] = self.location or getenv("GOOGLE_CLOUD_LOCATION")
-        
+
         _client_params = {k: v for k, v in _client_params.items() if v is not None}
-        
+
         if self.client_params:
             _client_params.update(self.client_params)
 

libs/agno/agno/embedder/qianfan.py

@@ -0,0 +1,11 @@
+from agno.embedder.openai import OpenAIEmbedder
+
+
+class QianfanEmbedder(OpenAIEmbedder):
+    id: str = "tao-8k"
+    base_url = "https://qianfan.baidubce.com/v2"
+    dimensions = 1024
+
+    def __post_init__(self):
+        if self.dimensions is None:
+            self.dimensions = 1024 if self.id == "tao-8k" else 1536

libs/agno/agno/models/qianfan/__init__.py

@@ -0,0 +1,5 @@
+from agno.models.qianfan.chat import Qianfan
+
+__all__ = [
+    "Qianfan",
+]

libs/agno/agno/models/qianfan/chat.py

@@ -0,0 +1,36 @@
+from dataclasses import dataclass
+from os import getenv
+from typing import Optional
+
+from agno.models.openai import OpenAILike
+
+
+@dataclass
+class Qianfan(OpenAILike):
+    """
+    A class for interacting with Nvidia models.
+
+    Attributes:
+        id (str): The id of the Nvidia model to use. Default is "nvidia/llama-3.1-nemotron-70b-instruct".
+        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 = "deepseek-v3"
+    name: str = "Qianfan"
+    provider: str = "Qianfan"
+
+    api_key: Optional[str] = getenv("QIANFAN_API_KEY")
+    base_url: str = "https://qianfan.baidubce.com/v2"
+
+    supports_native_structured_outputs: bool = True
+
+    role_map = {
+        "system": "system",
+        "user": "user",
+        "assistant": "assistant",
+        "tool": "tool",
+        "model": "assistant",
+    }

libs/agno/tests/integration/models/qianfan/test_basic.py

@@ -0,0 +1,165 @@
+import pytest
+from pydantic import BaseModel, Field
+
+from agno.agent import Agent, RunResponse
+from agno.models.qianfan import Qianfan
+from agno.storage.sqlite import SqliteStorage
+
+
+def _assert_metrics(response: RunResponse):
+    input_tokens = response.metrics.get("input_tokens", [])
+    output_tokens = response.metrics.get("output_tokens", [])
+    total_tokens = response.metrics.get("total_tokens", [])
+
+    assert sum(input_tokens) > 0
+    assert sum(output_tokens) > 0
+    assert sum(total_tokens) > 0
+    assert sum(total_tokens) == sum(input_tokens) + sum(output_tokens)
+
+
+def _get_model():
+    return Qianfan(id="deepseek-v3")
+
+
+def test_basic():
+    agent = Agent(model=_get_model(), markdown=True, telemetry=False, monitoring=False)
+
+    # Print the response in the terminal
+    response: RunResponse = agent.run("Share a 2 sentence horror story")
+
+    assert response.content is not None
+    assert len(response.messages) == 3
+    assert [m.role for m in response.messages] == ["system", "user", "assistant"]
+
+    _assert_metrics(response)
+
+
+def test_basic_stream():
+    agent = Agent(model=_get_model(), markdown=True, telemetry=False, monitoring=False)
+
+    response_stream = agent.run("Share a 2 sentence horror story", stream=True)
+
+    # Verify it's an iterator
+    assert hasattr(response_stream, "__iter__")
+
+    responses = list(response_stream)
+    assert len(responses) > 0
+    for response in responses:
+        assert response.content is not None
+
+    _assert_metrics(agent.run_response)
+
+
+@pytest.mark.asyncio
+async def test_async_basic():
+    agent = Agent(model=_get_model(), markdown=True, telemetry=False, monitoring=False)
+
+    response = await agent.arun("Share a 2 sentence horror story")
+
+    assert response.content is not None
+    assert len(response.messages) == 3
+    assert [m.role for m in response.messages] == ["system", "user", "assistant"]
+    _assert_metrics(response)
+
+
+@pytest.mark.asyncio
+async def test_async_basic_stream():
+    agent = Agent(model=_get_model(), markdown=True, telemetry=False, monitoring=False, debug_mode=True)
+
+    response_stream = await agent.arun("Share a 2 sentence horror story", stream=True)
+
+    async for response in response_stream:
+        assert response.content is not None
+
+    _assert_metrics(agent.run_response)
+
+
+def test_with_memory():
+    agent = Agent(
+        model=_get_model(),
+        add_history_to_messages=True,
+        num_history_responses=5,
+        markdown=True,
+        telemetry=False,
+        monitoring=False,
+    )
+
+    # First interaction
+    response1 = agent.run("My name is John Smith")
+    assert response1.content is not None
+
+    # Second interaction should remember the name
+    response2 = agent.run("What's my name?")
+    assert "John Smith" in response2.content
+
+    # Verify memories were created
+    messages = agent.get_messages_for_session()
+    assert len(messages) == 5
+    assert [m.role for m in messages] == ["system", "user", "assistant", "user", "assistant"]
+
+    # Test metrics structure and types
+    _assert_metrics(response2)
+
+
+def test_response_model():
+    class MovieScript(BaseModel):
+        title: str = Field(..., description="Movie title")
+        genre: str = Field(..., description="Movie genre")
+        plot: str = Field(..., description="Brief plot summary")
+
+    agent = Agent(
+        model=_get_model(),
+        markdown=True,
+        telemetry=False,
+        monitoring=False,
+        response_model=MovieScript,
+    )
+
+    response = agent.run("Create a movie about time travel")
+
+    # Verify structured output
+    assert isinstance(response.content, MovieScript)
+    assert response.content.title is not None
+    assert response.content.genre is not None
+    assert response.content.plot is not None
+
+
+def test_json_response_mode():
+    class MovieScript(BaseModel):
+        title: str = Field(..., description="Movie title")
+        genre: str = Field(..., description="Movie genre")
+        plot: str = Field(..., description="Brief plot summary")
+
+    agent = Agent(
+        model=_get_model(),
+        use_json_mode=True,
+        telemetry=False,
+        monitoring=False,
+        response_model=MovieScript,
+    )
+
+    response = agent.run("Create a movie about time travel")
+
+    # Verify structured output
+    assert isinstance(response.content, MovieScript)
+    assert response.content.title is not None
+    assert response.content.genre is not None
+    assert response.content.plot is not None
+
+
+def test_history():
+    agent = Agent(
+        model=_get_model(),
+        storage=SqliteStorage(table_name="agent_sessions", db_file="tmp/agent_storage.db"),
+        add_history_to_messages=True,
+        telemetry=False,
+        monitoring=False,
+    )
+    agent.run("Hello")
+    assert len(agent.run_response.messages) == 2
+    agent.run("Hello 2")
+    assert len(agent.run_response.messages) == 4
+    agent.run("Hello 3")
+    assert len(agent.run_response.messages) == 6
+    agent.run("Hello 4")
+    assert len(agent.run_response.messages) == 8