TeamAI Runtime

TeamAI Runtime is a local-first Python runtime for running YAML-defined AI teams with structured tasks, artifacts, reviews, and inspectable runs.

The MVP focuses on:

• Teamfiles as YAML configuration
• deterministic control in Python
• planner, specialist, critic, and finalizer agents
• capability-based routing
• bounded loops and budgets
• SQLite audit persistence
• safe filesystem tools
• fully offline tests through FakeModelClient

teamai init
teamai schema --output teamfile.schema.json
teamai validate team.yaml
teamai run team.yaml --input "Analyze this workspace and produce a short report" --yes
teamai run team.yaml --input "Analyze this workspace" --json
teamai inspect <run-id>

Python usage:

from teamai import TeamRuntime

async with TeamRuntime.from_file("team.yaml") as runtime:
    result = await runtime.run(goal="Analyze this workspace and produce a short report")

print(result.final_output)

By default, the Python API rejects side-effect approvals unless an approval provider is supplied. For trusted local demos, opt in explicitly:

async with TeamRuntime.from_file("team.yaml", auto_approve=True) as runtime:
    result = await runtime.run(goal="Write a report file")

Real models use the openai_compatible provider. Set the API key in an environment variable and point base_url at your endpoint:

models:
  default:
    provider: openai_compatible
    model: gpt-4o-mini
    base_url: https://api.openai.com/v1
    api_key_env: OPENAI_API_KEY
    capabilities:
      json_mode: true
      structured_output: false
      tool_calling: false

export OPENAI_API_KEY=sk-...
teamai run team.yaml --input "Summarize this workspace" --yes

Development checks:

uv sync --extra dev --locked
uv run python scripts/check_examples.py
uv run python scripts/check_import_rules.py
uv run pytest
uv run mypy src tests
uv run ruff check .
uv build

Release artifacts are built by GitHub Actions when a v* tag is pushed. The release workflow uploads dist/* as a GitHub artifact and does not publish to PyPI automatically.

Security reports should follow SECURITY.md. CodeQL runs on pushes, pull requests, manual dispatch, and a weekly schedule.

Contribution and governance expectations are documented in CONTRIBUTING.md, CODE_OF_CONDUCT.md, and GOVERNANCE.md.