Sentient Labs develops the foundation for open-source Artificial General Intelligence (AGI) through research in reasoning, autonomy, and learning. They provide frameworks like ROMA for building recursive AI agents and SERA for reliable tool-calling in specialized domains like crypto analysis. The company fosters an ecosystem called GRID where builders collaborate to advance open-source frontier AI.
Funding
Funding not disclosed
Founders
Product
Problem
Developing AI agents that can effectively interact with users in chat environments requires significant engineering effort to manage real-time communication, event handling, and data streaming. Existing frameworks often lack the necessary tools for building agents that can provide a rich, interactive user experience within platforms like Sentient Chat.
Solution
Sentient Agent Framework is a Python package designed to streamline the creation of AI agents that can seamlessly integrate with the Sentient Chat platform. The framework provides developers with an `AbstractAgent` class for defining agent logic and a `DefaultServer` class for handling real-time communication with the chat interface. By leveraging the `ResponseHandler` object, agents can emit various event types—text, JSON, errors, and streamed text chunks—to deliver a dynamic and informative user experience. The framework simplifies the process of building agents that can respond to user queries, provide intermediate updates, and stream final responses, enhancing user engagement and satisfaction.
Target Audience
The primary users are AI developers and engineers who want to build and deploy AI agents within the Sentient Chat ecosystem, particularly those seeking to create interactive and engaging user experiences.
Features
- `AbstractAgent` class for defining custom agent logic through the `assist()` method
- `DefaultServer` class for managing real-time communication and serving agent responses via SSE
- `ResponseHandler` object for emitting different event types (text, JSON, errors) to the client
- Support for streaming text chunks using `create_text_stream()` and `emit_chunk()` methods
- Asynchronous event handling for concurrent request processing
- Integration with Sentient Chat for rendering events and providing a richer user experience
- Example agents and client available for reference and testing