# Implementation plan

Repository inspection: empty project directory with reserved environment directories; no existing application or repository instructions.

1. **Phase 1 (implemented):** offline PySide6 dashboard, six navigation pages, ten inventory tabs, hostname validation, background demo loading, searchable/sortable tables, session history, explicit execution selection, synthetic fixture, local demo exports and preliminary evidence/risk presentation.
2. **Phase 2:** independent read-only PowerShell modules for system/roles, services, TCP/processes, scheduled tasks, registry software, volumes/shares and public certificate metadata. Versioned JSON envelope with per-check statuses. Optional AD enrichment interface; no AD requirement.
3. **Phase 3:** Windows execution-host collector using existing authorized Invoke-Command sessions. Resolve trusted script paths; pass validated hostname as an argument to a fixed script via subprocess argument lists (never shell=True). Enforce timeouts, cancellation, process cleanup, bounded output and schema validation. Linux remains offline. No remoting setup.
4. **Phase 4:** expand transparent weighted rules, role/process mappings, service account and inbound connection correlation; distinguish observed activity from potential dependencies and unknown evidence. Review critical infrastructure role rules before live use.
5. **Phase 5:** production reports, redaction of sensitive service/task arguments, failed-check visibility, export validation and report metadata. Existing Phase 1 exports contain synthetic records only.
6. **Phase 6:** collector contract tests, PowerShell validation on Windows, failure/timeout tests, Windows packaging and signed release guidance.

## Structure and boundaries

`server_buddy/ui`: Qt views and background worker. `core`: input validation and future orchestration. `collectors`: platform-specific collection backends. `models`: future typed schema adapters. `analyzers`: evidence and heuristic rules. `reporting`: local exports. `assets`: bundled offline fixture. `powershell`: future Windows-only scripts. `schemas`: collection contract. `tests`: fixture, core and optional Qt tests.

History is session-only; timestamps are original fixture timestamps. No SQLite or remote collection in Phase 1. A missing check never means a negative result. Synthetic demo is bound to a reserved example.test hostname and documentation-only IP addresses. Export pipeline must gain production redaction before any live collection is enabled.

## Scope expansion: Linux targets and desktop startup

The target scope now includes Windows Server 2016 onward and RHEL-based Linux, with either Windows or Mint as the reporting desktop. Added a RHEL synthetic dataset, target selector, background schema-validated JSON intake, OS-neutral report labels, NFS/Samba dependency rules, and a pre-Qt diagnostic for the missing Mint xcb cursor library. Live collection remains outside the current Phase 1 implementation. See PLATFORMS.md for transport boundaries, Linux inventory mappings and the required test matrix.

## Local Linux testing implementation

Added real read-only local collection for Mint and RPM-based Linux, a direct Analyze This Linux PC action, Settings mode, cancellation, command time/output bounds and headless JSON export. Real local results have distinct banners/history/report labels. Missing service identities, user/cron tasks, share exports, disk capacity, role correlation and certificate checks remain explicit. Both package formats are covered by parser tests; full RHEL and Windows validation remains pending. No remote transport is enabled.