Conduct
██████╗ ██████╗ ██████╗██╗ ██╗ ██╔═══██╗██╔══██╗██╔════╝██║ ██║ ██║ ██║██████╔╝██║ ███████║ ██║ ██║██╔══██╗██║ ██╔══██║ ╚██████╔╝██║ ██║╚██████╗██║ ██║ ╚═════╝ ╚═╝ ╚═╝ ╚═════╝╚═╝ ╚═╝
Conduct is a unified command-line interface for screenplay analysis, coverage generation, model deployment, and production workflow management. It provides professional-grade tools for evaluating screenplays using the industry-standard 16-point rubric while integrating with GPU infrastructure for AI-powered analysis.
Overview
Conduct consolidates multiple specialized CLIs into a single, coherent interface. Rather than switching between different tools for screenplay analysis, model serving, and infrastructure management, users interact with one unified system that handles the full production workflow.
The system is built around the concept of encounters—deep analytical sessions with screenplays that go beyond surface-level coverage to understand the material's commercial viability, artistic merit, and production requirements.
Features
Coverage System
Professional 16-point screenplay analysis using industry-standard rubric with scoring, recommendations, and actionable insights.
Model Architecture
Deploy and manage AI models on GPU infrastructure with automatic tensor parallelism and memory optimization.
Fine-Tuning
Conversational fine-tuning pipeline for customizing models with LoRA adapters and training session management.
GPU Management
Lambda Labs cluster management with real-time monitoring, visualization, and optimization tools.
Experiments
Run multi-model experiments across screenplays with configurable analysis protocols.
TUI Interface
Rich terminal interface with dashboards, progress views, and interactive components.
Installation
From Source
# Clone the repository
git clone https://github.com/AGI-Tooling/Orchestrator.git
cd Conduct
# Build and install
make install
# Or install globally (requires sudo)
make install-globalAfter installation, both conduct and conduct commands are available.
Core Modules
| Module | Command | Description |
|---|---|---|
| Coverage | coverage | Screenplay analysis and professional coverage generation |
| Architect | architect | Model deployment and serving on GPU infrastructure |
| Producer | producer | Fine-tuning pipeline and LoRA adapter management |
| Lambda | lambda | Lambda Labs GPU cluster management and monitoring |
| Experiment | experiment | Multi-model experiment management |
| Screenplays | screenplays | Screenplay library management and import |
| Interface | interface | GPU cluster web interface and control server |
Quick Start
Analyze a Screenplay
# Full 16-point coverage analysis
conduct coverage full screenplay.pdf --output-format html --output report.html
# Individual rubric items
conduct coverage logline screenplay.pdf
conduct coverage structure screenplay.pdf
conduct coverage characters screenplay.pdfDeploy a Model
# View deployment plan
conduct architect plan jamba-1.7-fp8
# Start serving
conduct architect serve jamba-1.7-fp8See Also
- Coverage System Overview — Detailed guide to screenplay analysis
- 16-Point Professional Rubric — Complete rubric documentation
- Command Reference — Full command documentation
- Professional Coverage Guide — Industry-standard evaluation framework
This documentation is embedded in the Conduct binary and served via conduct docs. No external files or internet connection required.