Skip to main content

A structured framework bringing uniformity to agent orchestration!

Project description

License: MIT Docs Tests

What is Grapheteria?

A structured agentic workflow framework bringing uniformity to agent orchestration:

  • 🎨 Design complex agent topologies with standardized visual primitives and a consistent API
  • ⏰ Debug and deploy with deterministic state tracking and minimal abstraction overhead
  • 🔌 Seamlessly integrate with ecosystem innovations like MCP tool interfaces, Agent2Agent communication, and any foundation model

"Oh no 😵‍💫, not ANOTHER workflow library! " - I hear you. But being fashionably late means we've learned from everyone else's mistakes and cherry-picked their best ideas, while adding our own secret sauce.

How is it different?

While agent creation and tool-calling have standards, workflow creation is still a mess. Grapheteria fixes that with a clean framework that gives you what you need while letting you extend things however you want.

Problems with Existing Tools

Code-based workflow builders make your brain hurt:

  • "Wait, what does this wrapper do again?"
  • "How many layers of inheritance am I dealing with?"
  • "I just want to see what's happening!"

UI-only tools hit walls quickly:

  • Can't customize for complex scenarios
  • Multi-agent setups? Forget about it
  • You eventually crawl back to code anyway

The Vision: Best of Both Worlds

Grapheteria blends code and UI seamlessly. Bounce between visual design and code customization without compromises. Get all the power of code with the clarity of visual debugging.

Code-UI Sync

Standout Features

Clean, Simple Code

Write workflows without a PhD in API-ology:

start_node = InputNode(id="get_name")
process_node = ProcessNode(id="greet")
output_node = OutputNode(id="display")

start_node > process_node > output_node

Powerful Agentic Patterns

Build sophisticated agent architectures with proven patterns:

  • Chain-of-Thought: Break down complex reasoning into sequential steps
  • Verification Loops: Self-check outputs with critic agents
  • ReAct Framework: Integrate reasoning and action loops seamlessly
  • Human-in-the-Loop: Add supervision exactly where needed
%%{init: {'theme': 'neutral' }}%%
graph TD
    %% Chain of Thought
    subgraph "Chain of Thought"
    A1[Task Input] --> B1[Reasoning Step 1]
    B1 --> C1[Reasoning Step 2]
    C1 --> D1[Reasoning Step 3]
    D1 --> E1[Final Output]
    end
    
    %% Verification Loop
    subgraph "Verification Loop"
    A2[Input] --> B2[Generator]
    B2 --> C2[Critic]
    C2 -->|Pass| D2[Output]
    C2 -->|Fail| B2
    end
    
    %% ReAct Framework
    subgraph "ReAct Framework"
    A4[Problem] --> B4[Reason]
    B4 --> C4[Act]
    C4 --> D4[Observe]
    D4 -->|Repeat| B4
    D4 -->|Complete| E4[Solution]
    end
    
    %% Human-in-the-Loop
    subgraph "Human-in-the-Loop"
    A5[Input] --> B5[Agent Process]
    B5 -->|Uncertain| C5[Human Review]
    C5 --> D5[Incorporate Feedback]
    D5 --> B5
    B5 -->|Confident| E5[Output]
    end

Visual Workflow Design

Edit visually or with code - they stay perfectly in sync!

  • Drop nodes right in the UI
  • Reroute your flow with a few clicks
  • Debug in real-time while tweaking code
  • Switch between code and UI whenever you want

Workflow Editor

Time-Travel Debugging

Made a mistake? No sweat:

  • Step backward in time
  • Fix the bug
  • Step forward
  • Keep going from exactly where you left off

Time Travel Debug

Built-in Essentials

  • Super detailed logging
  • Auto-saves your state
  • Easy resume after stopping

Logging Demo

Production-Ready Path

From your laptop to the cloud with minimal changes:

  • Scale to distributed execution
  • Track everything automatically
  • Handle errors like a pro

Scaling Diagram

Vibe-Coding Compatible

Describe what you want, then fine-tune:

  • Generate workflows with AI
  • Tweak instead of starting from scratch
  • Build complex flows in minutes, not hours

LLM Generation

Ready to Try It?

Installation

# Create a virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install Grapheteria
pip install grapheteria

Note: Grapheteria requires Python 3.6 or higher.

Launch the UI

Once installed, fire up the UI with:

grapheteria

This starts the Grapheteria interface and automatically syncs with your code.

Project details


Download files

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

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

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

grapheteria-0.0.0-1f-py3-none-any.whl (2.7 MB view details)

Uploaded Python 3

grapheteria-0.0.0-1e-py3-none-any.whl (263.8 kB view details)

Uploaded Python 3

grapheteria-0.0.0-1d-py3-none-any.whl (263.6 kB view details)

Uploaded Python 3

File details

Details for the file grapheteria-0.0.0-1f-py3-none-any.whl.

File metadata

  • Download URL: grapheteria-0.0.0-1f-py3-none-any.whl
  • Upload date:
  • Size: 2.7 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.11.11

File hashes

Hashes for grapheteria-0.0.0-1f-py3-none-any.whl
Algorithm Hash digest
SHA256 1282c679e2e1833023fe1bf08aeb882a648d7ccf9aa695b99325b36e6a65e8b0
MD5 596a2ebf759cb4d1cce84396cac9b431
BLAKE2b-256 8a6090963d49a3dc2e484bc9acd8b809c21d4e32a64cc5a7e9378c85bc4191da

See more details on using hashes here.

File details

Details for the file grapheteria-0.0.0-1e-py3-none-any.whl.

File metadata

  • Download URL: grapheteria-0.0.0-1e-py3-none-any.whl
  • Upload date:
  • Size: 263.8 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.10.16

File hashes

Hashes for grapheteria-0.0.0-1e-py3-none-any.whl
Algorithm Hash digest
SHA256 40759b026393bd15f10fc5a44682dc0484d8d7f707b3b8b2a24a4056a24516cf
MD5 44fb7c7f3a3e13d84ff0ff8a20b19c94
BLAKE2b-256 5ad771bbdaf6ba9df188f7a967acfb627b58c067fd3ba8fe6cdbf22018c337fa

See more details on using hashes here.

File details

Details for the file grapheteria-0.0.0-1d-py3-none-any.whl.

File metadata

  • Download URL: grapheteria-0.0.0-1d-py3-none-any.whl
  • Upload date:
  • Size: 263.6 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.10.16

File hashes

Hashes for grapheteria-0.0.0-1d-py3-none-any.whl
Algorithm Hash digest
SHA256 5ce13c5aa66a4dbd717418297eefbb0aa3ff3bae197aebfb071e3e03c1d90596
MD5 98a86dfd2cad2e9e72dc235eb63065f5
BLAKE2b-256 bb774f7a6dd30338117d62dcf1175174bbb9fa059aad94468065cb124c6695c9

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