chore: sync smartem-decisions OpenAPI spec to latest main

[vredchenko]

Summary

One-shot regeneration of `docs/api/smartem/swagger.json` from smartem-decisions main. Relates to smartem-decisions#253.

The checked-in spec had been frozen at `0.1.dev276+g1b7ed28d6.d20250818` (August 2025) because there is no automation wired up between backend releases and this copy. Eight months of endpoint and schema additions have been invisible to every consumer that pulls this file — the frontend's Orval client regen (`npm run api:update` in `packages/api`) among them.

Regenerated from smartem-decisions@`2c75a79` (fix: populate ATLAS_CREATED event id from atlas_id, not name), which includes all of the recent work:

• #248 — post-commit SELECT storm fix
• #250 — processing-feedback publish endpoints
• #249 — bulk gridsquare creation endpoint
• #254 — ATLAS_CREATED event id field fix
• Plus ~8 months of prior work not tracked in the above issues.

Delta

	Old	New	Change
Paths	25	58	+33 added, 0 removed
Schemas	30	51	+21 added, 0 removed

Zero removals — this is an additive-only sync, so existing generated clients will not break on regeneration.

New path families

• Agent streaming + instruction lifecycle: `/agent/{agent_id}/session/{session_id}/heartbeat`, `.../instructions/stream`, `.../instructions/{instruction_id}/ack`
• Debug surface for session/instruction management: `/debug/agent-connections`, `/debug/session/{session_id}/`, `/debug/sessions/`
• Quality-prediction model surface: `/prediction_models`, `/quality_predictions`, `/quality_metrics`, `/prediction_model/{name}/...`
• Image endpoints: `/grids/{grid_uuid}/atlas_image`, `/gridsquares/{gridsquare_uuid}/gridsquare_image`
• Atlas tile gridsquare positions: `/atlas-tiles/{tile_uuid}/gridsquares`
• Bulk gridsquare creation: `/grids/{grid_uuid}/gridsquares/batch` (#249)
• Processing feedback: `/micrographs/{uuid}/{motion_correction|ctf_estimation}/{completed|registered}` (#250)

New schema highlights

`GridSquareBatchCreateRequest`, `GridSquareBatchCreateResponse`, `MotionCorrectionRequest`, `CtfEstimationRequest`, `ProcessingFeedbackPublishResponse`, `AgentInstructionAcknowledgement`, `AgentInstructionAcknowledgementResponse`, various quality/prediction models and responses.

Generation command

```
SKIP_DB_INIT=true uv run python -c "
import logging, json, pathlib
logging.disable(logging.CRITICAL)
import smartem_backend.api_server as s
pathlib.Path('docs/api/smartem/swagger.json').write_text(
json.dumps(s.app.openapi(), indent=2, sort_keys=True) + '\n'
)
"
```

(Run from a smartem-decisions checkout at the desired commit. Uses `sort_keys=True` to match the existing file's sorted-key convention, so future diffs stay readable.)

Test plan

• `python -m json.tool docs/api/smartem/swagger.json` — valid JSON.
• Delta check: 0 removals (additive-only).
• Consumer smoke (to do manually after merge): in smartem-frontend, `cd packages/api && npm run api:update` — expect a clean regen with many new endpoints and types added, no breakage.

Next step

Automating this sync via a GitHub Actions workflow (part of smartem-decisions#253) will be a follow-up PR. Intent to open it for discussion before implementation.