Support topic explorations
One row per normalized topic the deep-generation pipeline has looked
at. Doubles as the concurrency lock (``pending`` + unique ``topic_key``
dedupes across gunicorn workers/containers) and as the cache of verdicts:
``done`` points to the guide it produced, ``no_match`` is the negative
cache that stops the pipeline from re-exploring a topic that really does
not exist in the ERP on every question. Central ``admin`` DB, no FKs —
same pattern as SupportGuide.
| Field | Description |
|---|---|
| Topic_key | |
| Question | |
| Status | Options:
|
| Resolved_model_keys_json | |
| Guide_id | |
| Attempts | |
| Cost_usd | |
| Error_note |
Support topic explorations It has the following related modules and may be of interest to you: