CaseFlow.

Most case management systems are either oversized (Jira, ServiceNow — good products, but you're paying for an organisation that isn't yours) or undersized (a shared inbox, a spreadsheet, a Trello board — fine until an auditor asks for a history).

The mid-segment — businesses handling cases, internal requests or regulatory matters in teams of 5–50 — gets stuck between the two. Per-seat pricing becomes unreasonable, off-the-shelf systems are over-engineered, and in-house development gets deferred until someone proves there's another way.

CaseFlow is a lightweight case queue that does one thing well: create a case → assign it → comment, attach, change status → see every change forever. No per-user licence fees. No vendor lock-in.

What it does

• Cases — create, assign, comment, attach files, change status (New → In progress → Done → Closed).
• Soft-delete only — nothing is ever truly deleted; everything is restorable inside the retention window.
• Audit trail on everything — every write logged immutably with Before/After diff, automatically captured via an EF Core SaveChanges interceptor.
• Idempotency on writes — every API call requires an Idempotency-Key header (Stripe model).
• Fail-open AI assistance — auto-categorisation, summarisation, reply suggestions via a self-hosted LLM. If the AI is down, cases keep working.
• Search via SQLite FTS5 over case title and description. The demo environment resets to seed baseline every night at 03:00.
• Two roles — Admin (full system access) and User (own and assigned cases).
• Notifications — email and push on assignment, comment, status change.

What sets it apart

Backend-first as a principle, not a slogan. All audit, soft-delete, idempotency and authorisation logic enforced server-side. The frontend is a thin view layer. SQLite on NFS is an architectural choice, not a compromise — single-tenant usage on a nightly-reset cadence, k6 stress tests in CI prove correctness under concurrency.

No hard delete exists. Data volume is not the problem; auditability is.