Skip to main content

🎨 Palette — Multi-Agent Orchestration Framework

Build, configure, and run multi-agent AI workflows with ease.

Palette is a configurable multi-agent orchestration framework built using Python and LangGraph. It allows users to create AI agents, organize them into layers, configure different LLM providers, and execute workflows through a simple graphical interface.

Instead of manually writing complex agent workflows, you can define your agents and their responsibilities in a YAML configuration file.


✨ Features

  • Visual GUI: Configure and run multi-agent workflows through a PyQt6 interface.
  • Dynamic Agent Creation: Create agents using configuration instead of hardcoding them.
  • Layer-Based Architecture: Organize agents into multiple execution layers.
  • Multiple LLM Providers: Supports Ollama, OpenAI, and Anthropic.
  • Flexible Model Configuration: Configure each agent's model, temperature, role, and instructions.
  • Agent Communication: Agents can access outputs generated by previous agents.
  • Final Answer Generation: A final agent can combine previous outputs into one response.
  • Dynamic LangGraph Workflow: Build graph structures automatically from YAML.
  • YAML Configuration: Save, load, and customize workflows.
  • Error Handling: Handles provider errors, missing API keys, model errors, and quota issues.

🏗️ How Palette Works

Palette follows a layer-based multi-agent architecture.

Each layer contains one or more agents that perform specific tasks.

Example workflow:

             USER INPUT
                  |
                  v
          +---------------+
          |    Layer 1    |
          |  Lead Agent   |
          +---------------+
                  |
                  v
       +---------------------+
       |      Layer 2        |
       |                     |
       |  Researcher         |
       |  Reviewer           |
       +---------------------+
                  |
                  v
          +---------------+
          |    Layer 3    |
          | Final Agent   |
          +---------------+
                  |
                  v
             FINAL OUTPUT

Layer Rules

Layer Maximum agents
Layer 1 1
Layer 2 onwards 2 per layer

Users can define the number of layers and assign agents to each layer.


🛠️ Technology Stack

Technology Purpose
Python Core programming language
LangGraph Agent workflow orchestration
LangChain LLM integration
PyQt6 Graphical user interface
PyYAML YAML configuration
Ollama Local LLM execution
OpenAI API Cloud LLM integration
Anthropic API Cloud LLM integration

📦 Installation

Option 1: Install using pip

After Palette is published on PyPI:

pip install palette-multi-agent

Launch the graphical interface:

palette

Option 2: Install from source

Clone the repository:

git clone YOUR_GITHUB_REPOSITORY_URL

Navigate to the project directory:

cd palette

Create a virtual environment:

python -m venv venv

Activate it on Windows:

.\venv\Scripts\activate

Activate it on Linux/macOS:

source venv/bin/activate

Install the project:

pip install -e .

Launch Palette:

palette

⚙️ Configuration

Palette uses YAML to define agents and workflow layers.

Example config.yaml:

task: "Explain Machine Learning to a beginner."

agents:

  head:
    provider: openai
    model: gpt-4o-mini
    temperature: 0.3
    role: "Lead AI"
    instructions: >
      Understand the user's question
      and provide useful context.

  researcher:
    provider: ollama
    model: qwen2:1.5b
    temperature: 0.7
    role: "Researcher"
    instructions: >
      Provide detailed and accurate
      information about the question.

  reviewer:
    provider: ollama
    model: qwen2:1.5b
    temperature: 0.5
    role: "Reviewer"
    instructions: >
      Review previous outputs and
      identify mistakes or missing information.

  final:
    provider: ollama
    model: qwen2:1.5b
    temperature: 0.3
    role: "Final Answer Generator"
    final: true
    instructions: >
      Combine previous outputs into
      one clear and complete answer.

layers:

  - name: layer_1
    agents:
      - head

  - name: layer_2
    agents:
      - researcher
      - reviewer

  - name: layer_3
    agents:
      - final

Agent Configuration Parameters

Parameter Description
provider LLM provider: ollama, openai, or anthropic
model Model identifier
temperature Controls response randomness
role Agent's responsibility
instructions Instructions given to the agent
final Marks the final-answer agent when set to true

The task field contains the user's question or task.


🔑 API Key Configuration

Palette does not require users to share API keys with the project developer.

Each user should configure their own API keys.

OpenAI

Set the environment variable:

Windows PowerShell:

$env:OPENAI_API_KEY="your_openai_api_key"

Linux/macOS:

export OPENAI_API_KEY="your_openai_api_key"

Anthropic

Windows PowerShell:

$env:ANTHROPIC_API_KEY="your_anthropic_api_key"

Linux/macOS:

export ANTHROPIC_API_KEY="your_anthropic_api_key"

You can also use a personal .env file with python-dotenv.

Never upload real API keys to GitHub or PyPI.


🦙 Using Ollama (Local LLM)

Ollama allows users to run supported language models locally.

Step 1: Install Ollama

Download it from:

https://ollama.com/download

Step 2: Download a model

ollama pull qwen2:1.5b

Step 3: Start the Ollama server

ollama serve

Step 4: Configure the agent

provider: ollama
model: qwen2:1.5b
temperature: 0.7

Keep the Ollama server running while using Palette.


🚀 Running Palette

Launch the GUI

palette

Run using a YAML configuration

python -m palettle.cli --config config.yaml

The GUI lets users configure agents, edit YAML, visualize workflows, and execute the configured task.


📁 Project Structure

palette/
│
├── frontend/
│   ├── __init__.py
│   └── app.py
│
├── palettle/
│   ├── __init__.py
│   ├── cli.py
│   ├── runner.py
│   ├── config.py
│   ├── model.py
│   ├── api_key.py
│   ├── agent_builder.py
│   ├── graph_builder.py
│   ├── layer_validaor.py
│   └── error.py
│
├── main.py
├── pyproject.toml
├── README.md
└── LICENSE

🧩 Error Handling

Palette provides readable error messages for common issues, including:

  • Missing or invalid API keys.
  • Unsupported LLM providers.
  • Ollama server connection failures.
  • Missing Ollama models.
  • API quota and rate-limit errors.
  • Invalid layer configurations.

🔮 Future Improvements

  • More LLM provider integrations.
  • Advanced agent communication strategies.
  • Parallel and conditional execution.
  • Agent memory and context management.
  • Workflow templates.
  • More advanced graph visualization.
  • Plugin-based agent extensions.

🤝 Contributing

Contributions, suggestions, and bug reports are welcome!

  1. Fork the repository.
  2. Create a new branch.
  3. Make your changes.
  4. Submit a pull request.

📄 License

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


👩‍💻 Author

Sejal Raykhere

GitHub: https://github.com/sejal04-hash


⭐ If you find Palette useful, consider giving the repository a star!

Metadata

Release files for palette-multi-agent 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for palette-multi-agent 0.1.0
File Size Uploaded Provenance Actions
palette_multi_agent-0.1.0.tar.gz 21.6 kB Without provenance Details for palette_multi_agent-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for palette-multi-agent 0.1.0
File Interpreter ABI Platform Actions
palette_multi_agent-0.1.0-py3-none-any.whl Python 3 none any Details for palette_multi_agent-0.1.0-py3-none-any.whl

Total release size: 42.0 kB

Release files / palette_multi_agent-0.1.0.tar.gz

Download URL palette_multi_agent-0.1.0.tar.gz
Size 21.6 kB
Tags Source
SHA-256 checksum
How to use checksums
b7daa3232ce0e01e3acc213f849e33e941b011e5711735af62f32248d0ddc673
BLAKE2b-256 checksum
How to use checksums
60b1734f3732f510dae99d9954d28df33fcb745129b3482c6049d478bd3ccd9e
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.5

Release files / palette_multi_agent-0.1.0-py3-none-any.whl

Download URL palette_multi_agent-0.1.0-py3-none-any.whl
Size 20.4 kB
Tags Python 3
SHA-256 checksum
How to use checksums
4bce678992287a1c3dbe0887a350a86aea209e33a57bdf47321fcaff06b5b8f9
BLAKE2b-256 checksum
How to use checksums
c914643966499fd3c4f88c1c4133fbaaa4237699b6d79ae038fd33199d6368c9
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.14.5

Security for palette-multi-agent 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page