Metadata-Version: 2.4
Name: sentient-agent-framework
Version: 0.3.0
Summary: This python package provides an agent framework that can be used to build agents that serve Sentient Chat events.
Author-email: Chris Aubin <christopher@sentient.xyz>
Maintainer-email: Chris Aubin <christopher@sentient.xyz>
License-Expression: Apache-2.0
Project-URL: Homepage, https://github.com/sentient-agi/Sentient-Agent-Framework
Project-URL: Issues, https://github.com/sentient-agi/Sentient-Agent-Framework/issues
Project-URL: Website, https://sentient.xyz
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: Programming Language :: Python :: 3.9
Requires-Python: >=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: cuid2==2.0.1
Requires-Dist: fastapi==0.115.12
Requires-Dist: python-ulid==3.0.0
Requires-Dist: uvicorn==0.34.0
Dynamic: license-file

<!-- Sentient Banner -->
<p align="center">
  <img src="banner.png"/>
</p>

<!-- Socials -->
<p align="center">
      <a href="https://sentient.xyz/" target="_blank" style="margin: 2px;">
    <img alt="Homepage" src="https://img.shields.io/badge/Website-Sentient.xyz-%23EAEAEA?logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%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%2BPC9zdmc%2B&link=https%3A%2F%2Fhuggingface.co%2FSentientagi"/>
  </a>
    <!-- Twitter -->
    <a href="https://x.com/SentientAGI">
        <img alt="Twitter Follow" src="https://img.shields.io/badge/Twitter-SentientAGI-white?logo=x"/>
    </a>
    <!-- Discord -->
    <a href="https://discord.gg/sentientfoundation">
        <img alt="Discord" src="https://img.shields.io/badge/Discord-SentientAGI-7289da?logo=discord&logoColor=white&color=7289da"/>
    </a>
    <!-- Hugging face -->
    <a href="https://huggingface.co/Sentientagi">
        <img src="https://img.shields.io/badge/Hugging_Face-SentientAGI-yellow?style=sociak&logo=huggingface"/>
    </a>
</p>

<!-- PyPI Package Info -->
<p align="center">
    <!-- Release -->
    <a href="https://pypi.org/project/sentient-agent-framework/">
        <img src="https://img.shields.io/pypi/v/sentient-agent-framework">
    </a>
    <!-- License -->
    <a href="https://pypi.org/project/sentient-agent-framework/">
        <img alt="License" src="https://img.shields.io/pypi/dm/sentient-agent-framework">
    </a>
</p>


<h1 align="center">Sentient Agent Framework</h1>

> [!WARNING]
> **This python package is currently in beta and will likely change. It is not yet ready for production use.**

In addition to supporting OpenAI API compatible agents, Sentient Chat supports a custom, open source event system for agent responses. These events can be rendered in Sentient Chat to provide a richer user experience. This particularly useful for streaming responses from an AI agent, when you might want to show the agent's work while the response is being generated, rather than having the user wait for the final response. **This python package provides an agent framework that can be used to build agents that serve Sentient Chat events.**

Examples of agents that use this framework/package can be found [here](https://github.com/sentient-agi/Sentient-Agent-Framework-Examples). A client for testing agents built with this framework is available [here](https://github.com/chris-aubin-sentient/Sentient-Agent-Client).

## Installation
```bash
pip install sentient-agent-framework
```

## Usage
The simplest way to use this framework is to import and use the `AbstractAgent` class and the `DefaultServer` class.

#### AbstractAgent
The `AbstractAgent` class is lightweight and extensible. To use it, simply subclass the class and implement the `assist()` method. Use the `ResponseHandler` object passed to the `assist()` method to emit events to the client.

#### DefaultServer
The `DefaultServer` class is designed to be used with the `AbstractAgent` class. A concrete implementation of the `AbstractAgent` class is passed into the `DefaultServer` constructor. The `DefaultServer` provides SSE server with `/assist` endpoint and automatically streams events emitted in the `assist()` method to the client.

> [!NOTE]  
> The snippet below comes from the [Sentient Agent Framework Examples](https://github.com/sentient-agi/Sentient-Agent-Framework-Examples) repository.

#### Example
```python
import logging
import os
from dotenv import load_dotenv
from src.search_agent.providers.model_provider import ModelProvider
from src.search_agent.providers.search_provider import SearchProvider
from sentient_agent_framework import (
    AbstractAgent,
    DefaultServer,
    Session,
    Query,
    ResponseHandler)
from typing import AsyncIterator


load_dotenv()
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)


class SearchAgent(AbstractAgent):
    def __init__(
            self,
            name: str
    ):
        super().__init__(name)

        model_api_key = os.getenv("MODEL_API_KEY")
        if not model_api_key:
            raise ValueError("MODEL_API_KEY is not set")
        self._model_provider = ModelProvider(api_key=model_api_key)

        search_api_key = os.getenv("TAVILY_API_KEY")
        if not search_api_key:
            raise ValueError("TAVILY_API_KEY is not set") 
        self._search_provider = SearchProvider(api_key=search_api_key)


    # Implement the assist method as required by the AbstractAgent class
    async def assist(
            self,
            session: Session,
            query: Query,
            response_handler: ResponseHandler
    ):
        """Search the internet for information."""
        # Search for information
        await response_handler.emit_text_block(
            "SEARCH", "Searching internet for results..."
        )
        search_results = await self._search_provider.search(query.prompt)
        if len(search_results["results"]) > 0:
            # Use response handler to emit JSON to the client
            await response_handler.emit_json(
                "SOURCES", {"results": search_results["results"]}
            )
        if len(search_results["images"]) > 0:
            # Use response handler to emit JSON to the client
            await response_handler.emit_json(
                "IMAGES", {"images": search_results["images"]}
            )

        # Process search results
        # Use response handler to create a text stream to stream the final 
        # response to the client
        final_response_stream = response_handler.create_text_stream(
            "FINAL_RESPONSE"
            )
        async for chunk in self.__process_search_results(search_results["results"]):
            # Use the text stream to emit chunks of the final response to the client
            await final_response_stream.emit_chunk(chunk)
        # Mark the text stream as complete
        await final_response_stream.complete()
        # Mark the response as complete
        await response_handler.complete()
    

    async def __process_search_results(
            self, 
            search_results: dict
    ) -> AsyncIterator[str]:
        """Process the search results."""
        process_search_results_query = f"Summarise the following search results: {search_results}"
        async for chunk in self._model_provider.query_stream(process_search_results_query):
            yield chunk


if __name__ == "__main__":
    # Create an instance of a SearchAgent
    agent = SearchAgent(name="Search Agent")
    # Create a server to handle requests to the agent
    server = DefaultServer(agent)
    # Run the server
    server.run()
```


## Emitting events
Whether using the `AbstractAgent` or the `DefaultResponseHandler`, a `ResponseHandler` is created for every agent query and is used to emit events to the client. 

#### Emitting text events
Text events are used to send single, complete messages to the client:
```python
await response_handler.emit_text_block(
    "PLAN", "Rephrasing user query..."
)
```

#### Emitting JSON events
JSON events are used to send JSON objects to the client:
```python
await response_handler.emit_json(
    "SOURCES", {"results": search_results["results"]}
)
```

#### Emitting error events
Error events are used to send error messages to the client:
```python
await response_handler.emit_error(
    "ERROR", {"message": "An error occurred"}
)
```

#### Completing a response
At the end of a response, `response_handler.complete()` is called to signal the end of the response (this will emit a `DoneEvent` using the `Hook`):
```python
await response_handler.complete()
```

#### Emitting a stream of text chunks
To stream a longer response one chunk at a time, use the `response_handler.create_text_stream` method. This returns a `StreamEventEmitter` that can be used to stream text to the client using the `emit_chunk` method:
```python
final_response_stream = response_handler.create_text_stream(
    "FINAL_RESPONSE"
    )
for chunk in self.__process_search_results(search_results["results"]):
    await final_response_stream.emit_chunk(chunk)
```

#### Completing a stream
At the end of the stream, `final_response_stream.complete()` is called to signal the end of the stream (this will emit a `TextChunkEvent` with `is_complete=True`):
```python
await final_response_stream.complete()
```

## Documentation
- [Interface Documentation](./src/sentient_agent_framework/interface/README.md)
- [Implementation Documentation](./src/sentient_agent_framework/implementation/README.md)
