biosage-terminal 2.0.0

BioSage - AI-Powered Medical Diagnostic Assistant TUI with ARGUS Debate Framework

BioSage Terminal

AI-Powered Medical Diagnostic Assistant - Terminal User Interface

Overview

BioSage Terminal is a sophisticated TUI (Terminal User Interface) application for medical diagnostics powered by AI. It leverages the ARGUS debate-based reasoning framework to provide evidence-based diagnostic suggestions through multiple specialist AI agents.

Features

• Full ARGUS Debate Framework Integration:

  • Moderator Agent: Orchestrates debate agenda and controls flow
  • Specialist Agents: Infectious Disease, Cardiology, Neurology, Oncology, Autoimmune, and Toxicology experts
  • Refuter Agent: Challenges weak evidence and generates rebuttals
  • Jury Agent: Renders final verdicts via Bayesian aggregation
  • CDAG: Conceptual Debate Graph tracks all propositions, evidence, and rebuttals

• Evidence-Based Reasoning: Uses Bayesian posterior computation for calibrated confidence scores

• Patient Management: Complete patient onboarding, vitals tracking, and case management

• Persistent Storage: Local JSON-based storage for all patient data and cases

• Beautiful TUI: Rich terminal interface with intuitive navigation using Textual framework

Installation

pip install biosage-terminal

# Install with preferred LLM provider
pip install biosage-terminal[gemini]   # Google Gemini (recommended)
pip install biosage-terminal[openai]   # OpenAI
pip install biosage-terminal[anthropic] # Anthropic Claude
pip install biosage-terminal[all-llms]  # All providers

Quick Start

After installation, simply run:

biosage

Or check your API configuration first:

biosage --check-api

Configuration

API keys and model names are configured via environment variables. BioSage auto-detects available providers in this order: Gemini > OpenAI > Anthropic > Groq > Mistral > Cohere > Ollama

Environment Variables

# API Keys
export GEMINI_API_KEY=your_key_here        # Google Gemini (recommended)
export OPENAI_API_KEY=your_key_here        # OpenAI
export ANTHROPIC_API_KEY=your_key_here     # Anthropic Claude
export GROQ_API_KEY=your_key_here          # Groq
export MISTRAL_API_KEY=your_key_here       # Mistral
export COHERE_API_KEY=your_key_here        # Cohere

# Optional: Custom model names
export GEMINI_MODEL=gemini-1.5-pro
export OPENAI_MODEL=gpt-4o
export ANTHROPIC_MODEL=claude-3-5-sonnet-20241022
export GROQ_MODEL=llama-3.1-70b-versatile

# For local Ollama (no API key needed)
export OLLAMA_MODEL=llama3.1

# Data storage directory (default: ~/.biosage)
export BIOSAGE_DATA_DIR=/path/to/data

You can also create a .env file in your working directory.

Navigation

Key	Action
d	Dashboard
o	Patient Onboarding (New Case)
r	Run Diagnosis
s	Specialist Grid
e	Evidence Explorer
Tab	Switch tabs / Next field
Enter	Select / Confirm
Esc	Back / Cancel
q	Quit

Data Storage

All data is stored locally in ~/.biosage/:

• patients/ - Patient records (JSON)
• cases/ - Diagnostic cases with ARGUS debate results
• evidence/ - Evidence and citations
• reports/ - Generated reports
• audit/ - Audit trail of all actions

ARGUS Debate Architecture

BioSage Terminal uses the full ARGUS multi-agent debate framework:

1. Moderator creates debate agenda for the diagnosis
2. Specialist Agents (6 domains) gather domain-specific evidence
3. Refuter Agent challenges evidence and generates rebuttals
4. CDAG (Conceptual Debate Graph) tracks:
   • Propositions (diagnosis hypotheses)
   • Evidence nodes (supporting/attacking)
   • Rebuttal nodes (challenges to evidence)
5. Bayesian Propagation computes posterior probabilities
6. Jury Agent renders final verdicts with:
   • Verdict label (supported/rejected/undecided)
   • Posterior confidence
   • Clinical reasoning

Development

git clone https://github.com/biosage/biosage-terminal
cd biosage-terminal
pip install -e .[dev]

License

MIT License