|
Download tasks/capterra/solution/site/implementation/clone/backend/README.md from TIGER-Lab/WebsiteBench: direct link, hf CLI and curl.
- Browser
- Download file 1.09 kB
-
https://huggingface.co/datasets/TIGER-Lab/WebsiteBench/resolve/main/tasks/capterra/solution/site/implementation/clone/backend/README.md
- Command line
-
hf download hf://datasets/TIGER-Lab/WebsiteBench/tasks/capterra/solution/site/implementation/clone/backend/README.md
-
curl -L -o README.md https://huggingface.co/datasets/TIGER-Lab/WebsiteBench/resolve/main/tasks/capterra/solution/site/implementation/clone/backend/README.md
1.09 kB
Persistent backend (R3)
Human R2 acceptance authorized the persistent backend. This package owns:
db.py— SQLite connection (WAL mode, foreign keys ON), the ordered migration list recorded inschema_migrations, deterministic fixture seeding, signed-cookie sessions, and the review draft / revision / submission domain. Migration 4 enforces one review per (reviewer, product) with a UNIQUE(account_id, product_id) index, deduplicating any pre-existing pair duplicates deterministically (lowest rowid kept) before the constraint applies.
The database file capterra.sqlite3 lives in the data directory
(env CLAWBENCH_DATA_DIR, default materials/capterra/data — a sibling of
clone/, never inside it, so the reload watcher never sees DB churn).
Admin reset: POST /__admin/reset guarded by the X-Clawbench-Admin-Token
header (env CLAWBENCH_ADMIN_TOKEN, dev default capterra-admin-dev-token).
It is never linked from any UI template.
The receipt remains truthful: it reads the stored submission row but never claims real publication on capterra.com.