🔧 Build Agentic AI Application with py-libp2p, SeaLion, LiteLLM, and EtherCalc Integration

[seetadev]

Description:

This issue proposes building an agentic AI application leveraging decentralized and composable components, with the following stack:

py-libp2p: For peer-to-peer communication between AI agents and services.

SeaLion: A framework for building agentic workflows using LLMs, designed for automation and orchestration of reasoning tasks.

LiteLLM: Lightweight abstraction layer for using OpenAI-compatible LLM APIs locally or remotely.

EtherCalc: Web-based collaborative spreadsheet, to serve as a UI for displaying or modifying agent-generated data.

Goals:

• Set up a py-libp2p node for P2P messaging between agents.
• Integrate SeaLion to manage the AI agent orchestration flow.
• Use LiteLLM to connect with LLM APIs or self-hosted models to power reasoning.
• Embed EtherCalc to allow spreadsheet-based data collaboration (e.g. agent updates results into EtherCalc sheet).
• Implement a sample task like:
  “An agent receives a prompt from another peer, queries the LLM, logs the result into EtherCalc, and sends a summary response to the sender.”

Stretch Goals:

• Support for decentralized task scheduling and propagation using libp2p PubSub.
• Capability for multiple agents to collaborate on spreadsheets in real-time via CRDT-backed data sharing.
• Add QUIC transport support if needed via py-libp2p (WIP).
• Package the app into a Dockerized or minimal binary deployment for OSS demos.

References:

SeaLion Docs: https://github.com/aisingapore/sealion

LiteLLM: https://github.com/BerriAI/litellm

EtherCalc: https://github.com/audreyt/ethercalc

py-libp2p issue for QUIC: libp2p#463

[GautamBytes]

@seetadev commenting to get assigned on this. Thanks!