# Resistance Openness DB Smoke

Date: 2026-06-28

## Command

```powershell
C:\Users\encep\AppData\Local\Programs\Python\Python311\python.exe scripts/smoke-resistance-openness-db.py --api-base-url http://127.0.0.1:8000 --timeout 240 --out docs/ops/resistance-openness-db-smoke-2026-06-28.json
```

## Result

The smoke created two fresh P1 learner sessions through the real API path:

- dev-login
- onboarding and consent
- `POST /sessions`
- five `POST /sessions/{id}/turn` calls
- direct DB read from `app.session_state`, `app.turns`, and `app.turn_client_state`

## Evidence

| Track | Session | API Curve | DB Final State |
|---|---|---|---|
| Empathy | `a6c4d731-875e-4023-a369-b07de65f511e` | `0.0 -> 0.0 -> 0.095 -> 0.133 -> 0.171`, stage `탐색` from turn 3 | stage `탐색`, `turn_seq=5`, `effective_openness=0.17100003361701965`, `rapport_credit=0.8100000023841858`, `resistance=0.5199999213218689` |
| Advice jump | `5bf60d80-12f5-45ef-99fd-1856d02f6bb4` | `0.0 -> 0.0 -> 0.0 -> 0.0 -> 0.0`, stage `라포` throughout | stage `라포`, `turn_seq=5`, `effective_openness=0.0`, `rapport_credit=0.0`, `resistance=0.9580000042915344` |

The DB also stored fast-loop `turn_client_state` rows for both sessions. The full JSON output is `docs/ops/resistance-openness-db-smoke-2026-06-28.json`.

## Verdict

PASS. The deterministic resistance engine opens the P1 session under empathic responses and keeps the session closed under advice-jump responses. The result is backed by both API turn responses and DB state.