Chat Circuit

Brief overview

🔍 Added a small feature to zoom in using mouse selection. Handy for looking at deep branches #ChatCircuit

👉 August 22, 2024

Short demos

Re-run all nodes in a branch

Chat Circuit now makes it possible to re-run a branch of your conversation with LLM with a different prompt. It supports all local LLMs running on @ollama
💾 👉 August 6, 2024

Generate Follow up questions

Implemented this idea in chat circuit. Here is a quick demo of the application along with generating follow up questions using #LLM August 20, 2024

Zoom in/out

🔍 Added a small feature to zoom in using mouse selection. Handy for looking at deep branches #ChatCircuit

👉 August 22, 2024

Minimap Support

#ChatCircuit Added a mini-map with the help of Sonnet 3.5 in @poe_platform.

Would have taken me days if not weeks to do it without any help. 🙏

~ 99% of code is written by Claude September 25, 2024

Export to JSON Canvas Document

Added option to export to #JSON Canvas document that can be imported by any supported application like @obsdmd / @KinopioClub

👉 September 26, 2024

Features

Multi-Branch Conversations Create and manage multiple conversation branches seamlessly.

Contextual Forking Fork conversation branches with accurate context retention.

Editor Features

Save and Load Diagrams

Undo and Redo

Zoom and Pan

Re-run nodes in a branch

It is possible to re-run all the nodes in a branch after changing the prompt it any node in the list.

Supported Providers

Chat Circuit supports multiple LLM providers through a flexible, provider-based architecture:

Local Providers (OpenAI-compatible APIs)

Provider	Default Endpoint	Environment Variable	Description
Ollama	http://localhost:11434	OLLAMA_API_BASE	Popular local LLM server
LMStudio	http://localhost:1234/v1	LMSTUDIO_API_BASE	Desktop app for running LLMs locally
KoboldCpp	http://localhost:5001/v1	KOBOLDCPP_API_BASE	Lightweight local inference server

Cloud Providers

Provider	Endpoint	Environment Variable	Description
OpenRouter	https://openrouter.ai/api/v1	OPENROUTER_API_KEY	Access to multiple cloud LLMs (requires API key)

All providers are auto-discovered at startup. The app will work with any combination of available providers.

Quick Start:

1. Start your preferred local provider (Ollama, LMStudio, or KoboldCpp)
2. Launch Chat Circuit - models will be automatically discovered
3. Select a model from the dropdown in any conversation node

Configuring Provider Endpoints:

You can configure provider endpoints in two ways:

1. Via Configuration Dialog (Recommended):

   • Open Configuration via menu: Configuration > Models... or press Ctrl+,
   • Enter custom endpoints for Ollama, LMStudio, and KoboldCpp
   • Settings are saved automatically and persist across sessions

2. Via Environment Variables (fallback if not set in UI):

   # Example: Run with custom endpoints via environment variables
   OLLAMA_API_BASE="http://192.168.1.100:11434" \
   LMSTUDIO_API_BASE="http://localhost:1234/v1" \
   KOBOLDCPP_API_BASE="http://localhost:5001/v1" \
   python3 main.py

See provider-config.example.sh for more detailed configuration examples.

Configuration Priority: UI Settings > Environment Variables > Defaults

Running the Application

To run this application, follow these steps:

Install dependencies

python3 -m pip install -r requirements.txt

Run application

python3 main.py

Model Discovery

This application discovers available LLM models dynamically from multiple providers:

• Ollama: Discovers local models from Ollama running at http://localhost:11434 (configurable via OLLAMA_API_BASE env var).
• LMStudio: Discovers local models from LMStudio running at http://localhost:1234 (configurable via LMSTUDIO_API_BASE env var).
• KoboldCpp: Discovers local models from KoboldCpp running at http://localhost:5001 (configurable via KOBOLDCPP_API_BASE env var).
• OpenRouter: Discovers free models from OpenRouter when OPENROUTER_API_KEY is set (via Configuration dialog or env var).

All local providers (Ollama, LMStudio, KoboldCpp) use OpenAI-compatible APIs through LiteLLM.

Provider Configuration:

You can customize provider endpoints using environment variables:

export OLLAMA_API_BASE="http://localhost:11434"      # Default Ollama endpoint
export LMSTUDIO_API_BASE="http://localhost:1234/v1"  # Default LMStudio endpoint
export KOBOLDCPP_API_BASE="http://localhost:5001/v1" # Default KoboldCpp endpoint
export OPENROUTER_API_KEY="your-api-key"              # OpenRouter API key

If a provider fails to respond (e.g., server not running), the app shows a warning but continues with models from other providers.

If no models are discovered from any provider, the app shows an error; please ensure at least one provider is running.

Running via Make

Prefer the provided make targets for development and running:

make install    # set up environment and pre-commit hooks
make check      # run linters/formatters and pre-commit checks
make run        # launch the application