Skip to main content

A distributed multi-agent system framework to run AI system across multiple computers

Project description

Prompits

Prompits is a decentralized, modular framework designed to empower collaborative AI agents in executing and optimizing intelligent workflows. It supports flexible communication, distributed coordination, and evolutionary learning, making it ideal for automating complex, multi-step processes in a scalable and transparent way.

Purpose

  • Enable the creation and optimization of intelligent workflows using AI and human collaboration.
  • Facilitate decentralized problem-solving across multiple agents with reusable skills (Practices).
  • Support open experimentation and innovation with transparent evaluation and incentive mechanisms.

Benefits

  • Scalable and extensible architecture for distributed AI agents.
  • Modular and composable components for building custom workflows.
  • Decentralized discovery, evaluation, and coordination of skills and agents.
  • Token-based economy (Pondo) for rewarding useful contributions and governing the system.

Features

  • Decentralized multi-agent collaboration
  • Modular and extensible system architecture
  • Graph-based and sequential workflow management
  • AI-human hybrid optimization through Profiler and Pathway evaluations
  • Token-based economy for governance and prioritization

Core Concepts

Prompits is composed of modular components that work together to enable intelligent, decentralized multi-agent workflows. These components can be grouped into the following categories:

1. Agents and Capabilities

  • Pit – The base agent that performs actions.
  • Practice – The specific skill or function a Pit can perform.
  • Plug – Enables communication between agents or with external systems.

Relationship: Each Pit has one or more Practices (skills) and communicates via Plugs.

2. Workflow Structure

  • Pathway – A full workflow composed of steps (Posts).
  • Post – A single step in a Pathway.
  • Planner – Creates and organizes Pathways based on Proposals.
  • Pathfinder – Executes Pathways by delegating Posts to agents.
  • Pioneer – Explores and tests new Pathways or methods.

Relationship: Proposals are passed to the Planner, which builds a Pathway of Posts. The Pathfinder then assigns Posts to suitable Pits, and Pioneers can innovate new solutions.

3. Coordination and Discovery

  • Plaza – A decentralized space for agents to advertise their capabilities and discover others.
  • Proposal – A request or task needing a workflow.
  • Pact – A formal agreement among agents on how to execute tasks.

Relationship: Agents use Plazas to coordinate and respond to Proposals. Pacts form the basis of collaboration.

4. Data and Integration

  • Pool – A storage space for shared data or resources.
  • Pouch – A database interface, acting as a specialized Pool for queries.

Relationship: Agents interact with Pools for reading and writing data. Pouches enable structured query access.

5. Optimization and Incentives

  • Profiler – Scores and ranks agent performance, Practices, and Pathways.
  • Pondo – Token system used for rewards, prioritization, and governance.
  • Pricing – Mechanism for bidding on task execution using Pondo.

Relationship: Profiler tracks and ranks contributions. Pondo is earned or spent based on agent performance and pricing agreements.

Agent Structure

In Prompits, an Agent can represent a higher-level container for multiple Pits, which are the active worker components. Each Pit in turn holds a set of Practices, which define the actions or capabilities the Pit can perform. This nested structure allows for scalability and modularity in managing agent behavior and task specialization.

  • Agent – High-level container or identity and instance of Pits.
  • Pit – Basic elements in Prompits.
  • Practice – A defined skill or action a Pit can perform.
  • Plug - Communication channel between agents.
  • Plaza - Bulletin board for agents to advertise status
  • Pool - Storage or memory

The diagram below illustrates this structure:

graph TB
    PZ[[Plaza]]
    subgraph agnet1
    AG1([Agent 1])
    AG1 --> P1[Pit 1]
    AG1 --> P2[Pit 2]
    P1 --> PR1[Practice: OCR]
    P1 --> PR2[Practice: Translate]
    P2 --> PR3[Practice: Summarize]
    P2 --> PR4[Practice: Classify]
    end
    subgraph agent2
    AG2([Agent 2])
    AG2 --> P3[Pit 3]
    P3 --> PR5[Practice: Chat]
    end
    AG2 --> PZ
    AG1 --> PZ
    AG1 == Plug === AG2
    PZ --> PR6[(Pool)]

Architecture Diagram

The diagram below shows how agents (Pits) interact with central discovery hubs called Plazas. Each agent is connected to one or more Plazas, which serve as decentralized communication spaces for coordination and collaboration. Around each agent, various Practices represent the skills or capabilities they offer.

  • Plaza 1 and Plaza 2 sit at the center of the system, connecting agents together.
  • Agents A to E represent intelligent workers capable of performing tasks.
  • Practices (e.g., OCR, Translate, Classify) are modular functions that agents can execute.
  • This layout emphasizes a decentralized, collaborative system where agents can discover, cooperate, and evolve through shared tasks.
graph TD
    %% Plazas at center
    PZ1[Plaza 1]
    PZ2[Plaza 2]
    PZ1 --- PZ2

    %% Agents around plazas
    A1[Agent A] --- PZ1
    A2[Agent B] --- PZ1
    A2 --- PZ2
    A3[Agent C] --- PZ2
    A4[Agent D] --- PZ2
    A5[Agent E] --- PZ2

    %% Agent A Practices
    A1 --> A1P1[🏗 Practice: OCR]
    A1 --> A1P2[🏗 Practice: Translate]

    %% Agent B Practices
    A2 --> A2P1[🏗 Practice: Summarize]

    %% Agent C Practices
    A3 --> A3P1[🏗 Practice: Extract]

    %% Agent D Practices
    A4 --> A4P1[🏗 Practice: Classify]
    A4 --> A4P2[🏗 Practice: Validate]

    %% Agent E Practices
    A5 --> A5P1[🏗 Practice: Format]

Installation

git clone https://github.com/alvincho/prompits.git
cd prompits/python-sdk
pip install -r requirements.txt

Usage

Start a basic agent:

python create-agent.py --config agent1.json --refresh

You can also configure agents via the agent.json file or extend functionality by adding new Practices or Pathways.

Contributing

Contributions are welcome! Please open an issue or submit a pull request. See CONTRIBUTING.md for details.

License

This project is licensed under the MIT License. See the LICENSE file for more information.

Disclaimer

⚠️ This project is currently in the Proof-of-Concept (PoC) stage.
It is intended for developer review, experimentation, and early feedback only.

  • Do not use this system in production environments.
  • APIs, architecture, and components are subject to significant change.
  • Contributions and suggestions are highly welcome as the project evolves.

Roadmap

Curious about where Prompits is headed? Check out the ROADMAP.md for a detailed view of upcoming phases, features, and milestones.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

prompits-0.0.2.tar.gz (110.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

prompits-0.0.2-py3-none-any.whl (105.0 kB view details)

Uploaded Python 3

File details

Details for the file prompits-0.0.2.tar.gz.

File metadata

  • Download URL: prompits-0.0.2.tar.gz
  • Upload date:
  • Size: 110.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.8

File hashes

Hashes for prompits-0.0.2.tar.gz
Algorithm Hash digest
SHA256 c79dd7b7e441d084f2143f535407320b3fd72c36c5220294a462f3062a359511
MD5 c9a9fc8df980f527d0c6f600f5f447cf
BLAKE2b-256 b3b3fbb96a3e5b25408899c49eb602ec4fcb3b60b54d7e41bfcb6543c29f6628

See more details on using hashes here.

File details

Details for the file prompits-0.0.2-py3-none-any.whl.

File metadata

  • Download URL: prompits-0.0.2-py3-none-any.whl
  • Upload date:
  • Size: 105.0 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.8

File hashes

Hashes for prompits-0.0.2-py3-none-any.whl
Algorithm Hash digest
SHA256 8105508848717e2149e5ce27f09445cc336b3f91022667e869d70da88d7ee9b9
MD5 e5ab9103d378745f8ba1b161f32e54d8
BLAKE2b-256 d32a30f5c53765eb31dab4a62b5cec8669e5c1950ca118f24c9072bf7eeb105e

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page