com um clique
sv-deploy
// Deploy Security Verifiers environments and packages. Use when asked to deploy to Prime Intellect Environments Hub, publish to PyPI, bump versions, build wheels, or manage releases.
// Deploy Security Verifiers environments and packages. Use when asked to deploy to Prime Intellect Environments Hub, publish to PyPI, bump versions, build wheels, or manage releases.
Run and analyze Security Verifiers evaluations. Use when asked to evaluate models on E1 (network-logs) or E2 (config-verification), generate metrics reports, compare model performance, or analyze eval results.
Generate SV-Bench metrics reports (summary.json + report.md) for E1/E2 runs, validate metrics contracts, and produce comparison-friendly artifacts from outputs/evals/.
Build and manage Security Verifiers datasets. Use when asked to build E1 or E2 datasets, create test fixtures, validate data, or manage dataset files for network-logs or config-verification environments.
Development workflow for Security Verifiers. Use when asked to run tests, lint code, format files, set up the development environment, or perform CI checks on the codebase.
Manage HuggingFace datasets for Security Verifiers. Use when asked to push datasets to HuggingFace, manage metadata, configure gated access, or set up user HF repositories for E1/E2 datasets.
| name | sv-deploy |
| description | Deploy Security Verifiers environments and packages. Use when asked to deploy to Prime Intellect Environments Hub, publish to PyPI, bump versions, build wheels, or manage releases. |
| metadata | {"author":"security-verifiers","version":"1.0"} |
Deploy environments to Prime Intellect Environments Hub and publish packages to PyPI.
| Short Name | Full Package | Environment |
|---|---|---|
| network-logs | sv-env-network-logs | E1 |
| config-verification | sv-env-config-verification | E2 |
| code-vulnerability | sv-env-code-vulnerability | E3 |
| phishing-detection | sv-env-phishing-detection | E4 |
| redteam-attack | sv-env-redteam-attack | E5 |
| redteam-defense | sv-env-redteam-defense | E6 |
prime login
make test-env E=network-logs
# Default: patch version bump
make hub-deploy E=network-logs
# Minor version bump (0.x.0)
make hub-deploy E=network-logs BUMP=minor
# Major version bump (x.0.0)
make hub-deploy E=network-logs BUMP=major
# No version bump
make hub-deploy E=network-logs BUMP=none
# 1. Validate (tests + lint + build)
make hub-validate E=network-logs
# 2. Bump version
make update-version E=network-logs BUMP=patch
# 3. Deploy to Hub
make deploy E=network-logs
make deploy E=network-logs TEAM=your-team
# Build single environment
make build-env E=network-logs
# Build all environments
make build
# Build security-verifiers-utils
make build-utils
Wheels are output to environments/sv-env-*/dist/.
# Bump patch (0.0.x)
make update-version E=network-logs BUMP=patch
# Bump minor (0.x.0)
make update-version E=network-logs BUMP=minor
# Bump major (x.0.0)
make update-version E=network-logs BUMP=major
make update-utils-version BUMP=patch
This updates both pyproject.toml and __init__.py.
make pypi-publish-utils-test
Install from TestPyPI:
pip install --index-url https://test.pypi.org/simple/ security-verifiers-utils
make pypi-publish-utils
Warning: This publishes to production PyPI. Requires confirmation.
Before deploying an environment:
make test-env E=namemake lintmake build-env E=nameprime loginAfter deployment, users can run:
# Install from Hub
vf-install intertwine/sv-env-network-logs
# Run evaluation
vf-eval intertwine/sv-env-network-logs --model gpt-5-mini --num-examples 10
When updating security-verifiers-utils (sv_shared), environments that depend on it need coordinated releases:
# 1. Make changes to sv_shared/
# 2. Sync version in BOTH files (required for PyPI)
make update-utils-version BUMP=patch # Updates pyproject.toml AND __init__.py
# 3. Test environments still work
make test
# 4. Publish to PyPI first
make pypi-publish-utils
# 5. Then deploy environments that use the new utils
make hub-deploy E=network-logs
make hub-deploy E=config-verification
Important: The Hub pulls security-verifiers-utils from PyPI, so always publish utils before deploying dependent environments.
Deploy multiple environments efficiently:
# Sequential deployment (recommended for first-time)
make hub-deploy E=network-logs
make hub-deploy E=config-verification
make hub-deploy E=code-vulnerability
# Or use direct deploy if already validated
make deploy E=network-logs && make deploy E=config-verification
prime login fails: Check internet connection, try prime logout then prime login.
Wheel build fails: Ensure uv sync was run in the environment directory.
Version conflict: Check the version in pyproject.toml isn't already published.
PyPI auth: Ensure ~/.pypirc or TWINE_USERNAME/TWINE_PASSWORD are set.
Interactive login prompt: The first make deploy will open browser for authentication and prompt for team selection. This is normal.
sv_shared version mismatch: If PyPI publish fails with "version exists", ensure both sv_shared/pyproject.toml and sv_shared/__init__.py have matching, new versions.
Pre-commit hook fails: Install pre-commit with uv pip install pre-commit.
Environments are tested on the Hub via CI. If tests fail:
is_completed() signature matches verifiers APIenv_response() return typemax_turns parameter is accepted