Kyber Command

A localized multi-agent coordination hub. Power specialized droids with local LLMs (Ollama). Monitor their activity via Chainlit and Arize Phoenix. Human-in-the-loop safeguards ensure no agent modifies your files without clearance.

Architecture

• Supervisor: Routes queries to Researcher (general info) or Coder (Python/filesystem)
• Researcher: Answers factual questions, no tools
• Coder: Handles code and files with read_file, write_file, list_directory—file writes require approval
• Persistence: SQLite (state.db) via LangGraph SqliteSaver
• Observability: Phoenix at http://localhost:6006 for tracing

Prerequisites

• Python 3.11+
• Ollama running at http://localhost:11434
• Model: llama3.1:8b (or configure in config.yaml)

Quick Start

# 1. Create virtual environment
python -m venv .venv
.venv\Scripts\activate   # Windows
# source .venv/bin/activate  # macOS/Linux

# 2. Install dependencies
pip install -r requirements.txt

# 3. Ensure Ollama is running and has a model
ollama pull llama3.1:8b

# 4. (Optional) Start Phoenix for observability
phoenix serve
# Then open http://localhost:6006

# 5. Run the hub
chainlit run app.py --port 8080

Open http://localhost:8080 in your browser.

Usage

1. Ask a question – The Supervisor routes to Researcher or Coder.
2. File operations – If the Coder proposes a file write, you'll see Approve, Modify, or Reject buttons before it runs.
3. Sessions – Conversations persist in state.db across restarts.

Configuration

Edit config.yaml to:

• Change Ollama base URL and default model
• Customize agent system prompts
• Set SQLite path and Phoenix endpoint

Project Structure

kyber-command/
├── agents/
│   ├── state.py        # Pydantic/TypedDict state
│   ├── supervisor.py   # Router agent
│   ├── researcher.py   # General info agent
│   └── coder.py        # Code/files agent (HITL tools)
├── config.yaml         # Dynamic configuration
├── graph_engine.py     # LangGraph definition + SqliteSaver
├── app.py              # Chainlit entry (chat + HITL UI)
├── observability.py    # Phoenix instrumentation
├── requirements.txt
└── README.md

Observability

With Phoenix running (phoenix serve), all agent "thought chains" are traced at http://localhost:6006. The app uses OpenInference instrumentation for LangChain/LangGraph.