SCTY-Inc/mirofish-cli(opens in a new tab)
Language: Python
Multi-agent AI prediction engine - digital sandbox for scenario simulation (fork of 666ghj/MiroFish)
- stars
- 323
- watchers
- 2
- forks
- 109
- open issues
- 6
- Last updated
- 10/01/2026
Language: Python
Multi-agent AI prediction engine - digital sandbox for scenario simulation (fork of 666ghj/MiroFish)
Social simulation engine. Feed it documents and a plain-English requirement; it builds a knowledge graph, spins up AI agents that post and argue on simulated Twitter and Reddit, then writes a report, a machine-readable verdict.json, and SVG snapshots.
Fork of 666ghj/MiroFish, translated to English and reduced to a CLI that runs on the claude or codex CLI.
uv tool install git+https://github.com/SCTY-Inc/mirofish-cli
mirofish doctor # check provider CLI (claude/codex) and configRequires Python 3.11-3.12 and uv. For development: uv sync, then uv run pytest.
mirofish run \ --files policy.pdf context.md \ --requirement "Predict public reaction over 30 days" \ --json mirofish runs list --json mirofish runs status <run_id> --json mirofish runs export <run_id> --json
mirofish run options: --files (pdf/md/txt), --requirement, --platform parallel|twitter|reddit (default parallel), --max-rounds N, --output-dir PATH, --json.
--json: live progress display on stderr. With --json: JSON on stdout.Each run writes an immutable directory under uploads/runs/<run_id>/ containing manifest.json, report/verdict.json (prediction, confidence, key dynamics, signals), report/summary.json, report/report.md, and visuals/*.svg.
Set LLM_PROVIDER in .env (see .env.example). Accepted values: claude-cli (default) and codex-cli. Anything else exits 1 at startup.
AGPL-3.0