Skip to content

PACKAGE_MANIFEST.md

Latest commit

 

History

History
63 lines (54 loc) · 1.77 KB

PACKAGE_MANIFEST.md

File metadata and controls

63 lines (54 loc) · 1.77 KB
Raw

Cognitive OS Distilled Runtime Snapshot

This package keeps the source needed to run the local-first Cognitive OS runtime. It intentionally excludes desktop UI shells, WebArena, ARC-AGI-3, benchmark fixtures, frozen reports, and generated artifacts.

Included source surfaces:

  • conos_cli.py
  • core/
  • decision/
  • evolution/
  • integrations/local_machine/
  • modules/
  • planner/
  • scripts/conos.py
  • scripts/local_mirror.py
  • scripts/run_local_machine.py
  • scripts/check_conos_repo_layout.py
  • scripts/check_runtime_preflight.py
  • self_model/
  • tests/
  • tools/managed_vm/
  • trace_runtime.py

Excluded by design:

  • desktop app sources and generated .app bundles
  • WebArena adapters and runners
  • ARC-AGI-3 adapters and runners
  • benchmark fixtures and closed-loop experiment runners
  • frozen phase evidence reports and external baseline artifacts
  • generated outputs under runtime/, reports/, audit/, and dist/
  • Python caches and packaging metadata generated by local installs

Minimal setup:

python3.10 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
pip install -r requirements-dev.txt
pip install -e .
conos preflight --strict-dev
conos layout
pytest -q

Primary entry points:

conos run local-machine --instruction "inspect README" --candidate README.md --max-ticks 2
conos mirror init --mirror-root runtime/mirrors/session-1
conos llm --help
conos vm report
conos supervisor --help

Execution safety note:

Local-machine command execution defaults to the Con OS managed-VM boundary. Host-local execution is a development opt-in via --execution-backend local or conos mirror exec --backend local. Mirror writes require evidence, validation, an approved sync plan, source-hash checks, and rollback checkpoints.