Summary

• Bump madsci.* pin from ~=0.6.0 to ~=0.8.0.
• Replace removed self.node_definition.node_name with self.node_info.node_name (v0.8 dropped the node_definition attribute).
• Regenerate pdm.lock against v0.8.
• .gitignore: ignore .madsci/ (per-machine runtime state).

Cross-references

Same v0.8-upgrade sweep as AD-SDL/bmg_module#3 (split sidecar), AD-SDL/inheco_incubator_module (lifespan migration), AD-SDL/biometra_module (straight v0.8 bump). Three upstream MADSci bugs surfaced during the sweep:

• madsci-common: psycopg2-binary is mandatory but unused (blocks 32-bit Windows installs) MADSci#343 — psycopg2-binary mandatory but unused
• event_client: double-fault when warning_category=<class> is shipped to event server MADSci#344 — event_client double-fault on warning_category=<class>
• madsci-common: filelock imported by registry/local_registry.py but not declared as a dependency MADSci#345 — filelock import not declared in deps

This module isn't blocked by any of them in practice, but the noisy event-client tracebacks in #344 do appear in this node's logs when the resource manager is unreachable.

Test plan

• pdm install resolves cleanly with v0.8.
• Node boots, serves /info returning correct NodeInfo.
• Resource template hidex.nest is created with the right node_name prefix.
• Existing actions (open, close, stop, etc.) round-trip.
• Under process-compose orchestration alongside the other v0.8 nodes.

🤖 Generated with Claude Code