src/llama_suite/: core Python code.src/llama_suite/watchers/:llama-swaplaunch/restart helpers.src/llama_suite/eval/,src/llama_suite/bench/: evaluation and benchmarking tools.src/llama_suite/webui/: FastAPI Web UI, schema, and static assets.configs/: base config plus machine-specific overrides.deploy/: container, Compose, Helm, and marketplace packaging.
- Install:
python tools/scripts/install.py --dev-extras - Update deps/tools:
./.venv/bin/python tools/scripts/update.py --dev-extras - Run Web UI:
./.venv/bin/python -m llama_suite.webui.server - Run tests:
./.venv/bin/python -m pytest -q - Lint:
./.venv/bin/python -m ruff check src tools - Type check:
./.venv/bin/python -m mypy src/llama_suite
- Follow PEP 8.
- Prefer type hints.
- Keep scripts cross-platform.
- Use
snake_casefor functions/variables andPascalCasefor classes.
- Do not commit local runtime data or generated artifacts.
- Keep secrets out of YAML and git; use environment variables or CLI flags.
- Treat
configs/config.base.yamlas the shared baseline and keep machine-specific changes inconfigs/overrides/.
- At the start of every session, run:
~/.codex/skills/self-improving-agent-1.0.5/scripts/activator.sh - After any failed shell command, run:
~/.codex/skills/self-improving-agent-1.0.5/scripts/error-detector.sh