feat(agents): PINA multi-agent team with pydantic-graph + MLflow + A2A/AG-UI#33

Merged

bjoernbethge merged 24 commits into

mainfrom

feat/agents-pina-team

Apr 21, 2026

Commits on Apr 21, 2026

fix(agents): use OpenAIChatModel + ruff format (review fixes for Task 3)
claude
committed
feat(agents): add skill loader + build_skill_instructions callable
claude
committed
feat(agents): add MCP toolset builders for marimo + mlflow servers
claude
committed
feat(agents): add MLflowStatePersistence (snapshots as MLflow artifacts)
claude
committed
feat(agents): add RouteNode classifier with sub-node stubs
claude
committed
feat(agents): implement NotebookNode (marimo MCP + skills)
claude
committed
feat(agents): implement ProblemNode (define_problem — open-spec PINA Problem)
claude
committed
feat(agents): implement ModelNode (define_model — open-spec architecture)
claude
committed
feat(agents): implement SolverNode (define_solver — open-spec config)
claude
committed
feat(agents): implement MLflowNode (mlflow MCP + mlflow skill)
claude
committed
feat(agents): assemble graph with all six nodes (RouteNode start)
claude
committed
feat(agents): add lead agent wrapping graph + mlflow autolog
claude
committed
feat(agents): add marimo chat adapter (delta streaming)
claude
committed
feat(agents): add A2A (with AgentCard skills) + AG-UI ASGI app builders
claude
committed
feat(agents): expose public API at package root
claude
committed
feat(examples): rewrite lab.py as PINA team chat demo
claude
committed
test(agents): add end-to-end smoke test (real MLflow + TestModel)
claude
committed
fix(agents): use MlflowClient with explicit run_id; xfail upstream AG-UI bug
claude
committed
docs(agents): document PINA multi-agent team (README + CHANGELOG + CLAUDE.md)
claude
committed
style(agents): apply ruff isort to test_state.py
claude
committed