metasystem-master

Omni-Orchestrate

Multi-AI Orchestration CLI for the Omni-Performative Engine Project

A command-line tool that coordinates research, validation, and synthesis across 5 AI services (Perplexity, Gemini, ChatGPT, Copilot, Grok) in a structured pipeline with gate validation.

Overview

PHASE 1: RESEARCH VALIDATION (Perplexity)
  ├─ Precedent verification
  ├─ Funding landscape mapping
  └─ [GATE 1: Human review]

PHASE 2: SPECIFICATION HARDENING (Gemini)
  ├─ Edge-case matrix (5×5)
  ├─ Latency/constraint validation
  └─ [GATE 2: Human review]

PHASE 3: MESSAGING SYNTHESIS (ChatGPT)
  ├─ NSF grant narrative
  ├─ NEH grant narrative
  ├─ Ars Electronica narrative
  ├─ Artist statement
  └─ [GATE 3: Human review]

PHASE 4: IMPLEMENTATION PLANNING (Copilot)
  ├─ Code architecture review
  ├─ Budget allocation
  └─ [GATE 4: Human review]

PHASE 5: VULNERABILITY AUDIT (Grok)
  ├─ Assumption critique
  ├─ Failure scenario modeling
  └─ [GATE 5: Final synthesis]

Installation

# Clone or copy this directory
cd omni-orchestrate

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # or `venv\Scripts\activate` on Windows

# Install dependencies
pip install -r requirements.txt

Configuration

1. Copy the config template

cp config.yaml.template config.yaml

2. Set API keys

Either edit config.yaml directly:

api_keys:
  perplexity: "pplx-xxxxxxxxxxxx"
  gemini: "AIzaSyxxxxxxxxxx"
  chatgpt: "sk-xxxxxxxxxxxx"
  grok: "xai-xxxxxxxxxxxx"

Or use environment variables:

export PERPLEXITY_API_KEY="pplx-xxxxxxxxxxxx"
export GEMINI_API_KEY="AIzaSyxxxxxxxxxx"
export OPENAI_API_KEY="sk-xxxxxxxxxxxx"
export GROK_API_KEY="xai-xxxxxxxxxxxx"

3. Verify setup

python src/orchestrator.py status --services all

Expected output:

Service Status:
------------------------------
  ✓ perplexity
  ✓ gemini
  ✓ chatgpt
  ✓ copilot
  ✓ grok

Usage

Run Full Pipeline

# Run all phases with gate validation
python src/orchestrator.py run --phase all --gates --output-dir ./results

# Run with human review pauses at each gate
python src/orchestrator.py run --phase all --pause-at-gate

Run Single Phase

# Run only Phase 1 (Research Validation)
python src/orchestrator.py run --phase research-validation

# Run Phase 2 with pause at gate
python src/orchestrator.py run --phase spec-hardening --pause-at-gate

Available Phases

Phase	Name	Service
1	`research-validation`	Perplexity
2	`spec-hardening`	Gemini
3	`messaging-synthesis`	ChatGPT
4	`implementation-planning`	Copilot
5	`vulnerability-audit`	Grok

Check Service Status

python src/orchestrator.py status --services all

Estimate Costs

python src/orchestrator.py estimate --phases all

Generate Synthesis Report

python src/orchestrator.py synthesis --input-dir ./results --format markdown

Output Structure

results/
├── phase1_research_validation/
│   ├── precedent_verification.json
│   ├── precedent_verification.md
│   ├── funding_landscape.json
│   ├── funding_landscape.md
│   └── gate_result.json
├── phase2_spec_hardening/
│   ├── edge_case_matrix.json
│   ├── edge_case_matrix.md
│   ├── latency_constraints.json
│   └── gate_result.json
├── phase3_messaging_synthesis/
│   ├── grant_narrative_nsf.json
│   ├── grant_narrative_nsf.md
│   ├── grant_narrative_neh.json
│   ├── grant_narrative_ars.json
│   ├── artist_statement.json
│   └── gate_result.json
├── phase4_implementation_planning/
│   ├── code_architecture.json
│   ├── budget_allocation.json
│   └── gate_result.json
├── phase5_vulnerability_audit/
│   ├── assumption_critique.json
│   └── failure_scenarios.json
├── aggregated_results.json
└── EXECUTIVE_REPORT.md

Prompt Templates

All prompt templates are in the prompts/ directory:

prompts/
├── phase1_research/
│   ├── precedent_verification.txt
│   └── funding_landscape.txt
├── phase2_specification/
│   ├── edge_case_matrix.txt
│   └── latency_constraints.txt
├── phase3_messaging/
│   ├── grant_narrative_nsf.txt
│   ├── grant_narrative_neh.txt
│   ├── grant_narrative_ars.txt
│   └── artist_statement.txt
├── phase4_implementation/
│   ├── code_architecture.txt
│   └── budget_allocation.txt
├── phase5_vulnerability/
│   ├── assumption_critique.txt
│   └── failure_scenarios.txt
└── gates/
    └── all_gates.txt

Customizing Prompts

Edit the .txt files in prompts/ to customize for your project. Use {placeholder} syntax for dynamic substitution:

You are validating research for {project_name}.

CLAIMS TO VERIFY:
{precedent_claims}

Pass context when running:

orchestrator.run_phase(phase, context={
    "project_name": "My Project",
    "precedent_claims": "..."
})

Gate Validation

Each phase (except Phase 5) has a gate validation step that checks:

Gate 1 (Research)

≥80% claims verified with sources
≥10 funding opportunities identified
≥3 near-term deadlines
No contradicted claims

Gate 2 (Specification)

25/25 edge cases populated
Critical issues have mitigations
Latency constraint achievable

Gate 3 (Messaging)

Multiple narrative variants generated
Artist statement present
Cross-narrative coherence

Gate 4 (Implementation)

Architecture reviewed
Budget includes ≥20% contingency
Timeline realistic

Gate 5 (Vulnerability)

All 6 assumption categories covered
≥10 failure scenarios
Actionable mitigations

API Cost Estimates

Service	Tasks	Est. Tokens	Est. Cost
Perplexity	2	6,000	$0.01
Gemini	2	16,000	$0.02
ChatGPT	4	16,000	$0.16
Copilot	2	6,000	$0.06
Grok	2	6,000	$0.03
Total	12	50,000	~$0.30

Actual costs depend on response length and may vary.

Development

Running Tests

pytest tests/ -v

Code Formatting

black src/

Type Checking

mypy src/

Troubleshooting

“Service not configured”

Check that the API key is set in config.yaml or environment
Run python src/orchestrator.py status to verify

“Prompt template not found”

Ensure prompt files exist in prompts/{phase_name}/
Check file extension (.txt, .md, or .prompt)

Timeout errors

Increase timeout_seconds in config.yaml
Reduce parallel_limit to avoid rate limits

Gate validation fails

Review output in results/phaseN_*/gate_result.json
Address recommendations before re-running

License

MIT License - See LICENSE file for details.

Contributing

Fork the repository
Create a feature branch
Make changes with tests
Submit a pull request

Built for the Omni-Performative Engine project

This site is open source. Improve this page.