Start with a product or guide

  • Start here

↑↓to moveto openEscto close
DesktopData and evidence

Data, governance, and evidence

Mezair Desktop stores workspace state, repository metadata, connector settings, reviews and workflow history in a local SQLite database. This page explains how contributors should handle that data and keep review records attached to the work.

SQLite persistence

SQLite is the desktop persistence layer. Schema, current version, and migrations live in src/main/db/schema.ts; startup and migration execution live in src/main/db/database.ts.

When persistence changes:

  • add a new migration
  • preserve old migrations
  • update service behavior
  • test the affected path where practical
  • document any user-visible data migration risk

Workspace state

Workspace state coordinates:

  • active repositories
  • provider connections
  • selected roles
  • preferences
  • chat/session history
  • review and security findings
  • launch intents
  • companion pairing state

Use the workspace's existing settings and state. A second global copy can become stale or apply one workspace's choices to another.

Delivery decisions

Governance features should make lifecycle state visible:

  • work item readiness
  • acceptance criteria gaps
  • impact analysis
  • risk notes
  • compliance checks
  • release readiness
  • handover packs
  • follow-up ownership

Each gate should identify the decision it controls and the information needed to make it.

Evidence lifecycle

Evidence should move with the work:

  1. Planning identifies files, assumptions, and risks.
  2. Implementation changes code and records validation.
  3. Review and security sessions produce findings.
  4. Fixes update evidence and rerun checks.
  5. Handover summarizes changed behavior, passed checks, skipped checks, and remaining risk.

The useful artifact is not the chat transcript. It is the reviewable summary with enough source references and command context to trust.

Connector data

Connector configuration can involve provider tokens, issue metadata, repository state, and documentation systems. Treat those values as credentials or sensitive workflow data.

The renderer should not own secret handling. Main-process services and the preload bridge should keep privileged access behind explicit APIs.

Suggest an edit to this page on GitHub