Codex: expose the analyzer through a local machine interface

[hardcoreerik]

Agent

Codex

What Changed

• Added structured local error reports under error_reports/ and wired startup/analysis failures into the report writer.
• Sanitized PCAP-derived report values so illegal spreadsheet characters do not crash XLSX generation.
• Sanitized GUI worker log/error strings before sending them through PySide.
• Added wardrive_service.py shared service functions for project summary, evidence listing, run discovery, folder scan, and analysis.
• Added wardrive_cli.py for JSON one-shot commands and a local JSONL stdio bridge for AI agents.
• Added docs/ai-bridge.md and README bridge commands.
• Added launch.cmd for local Windows startup.

Why

• Analysis errors needed durable local reports instead of transient dialogs.
• Claude/Codex need a structured interface to inspect and operate Wardrive Analyzer without scraping the GUI.
• The app should stay desktop-first while exposing safe local automation points.

Verification

• python -m py_compile wardrive_service.py wardrive_cli.py error_logger.py run_step3d_scene.py gui_step3d_scene.py project_vault.py core\analyze.py core\parser_logs.py core\parser_pcap.py core\writers.py
• git diff --check
• python wardrive_cli.py project-summary "F:\WardriveAnalytics" --import-limit 2 --compact
• python wardrive_cli.py latest-run "F:\WardriveAnalytics" --compact
• python wardrive_cli.py evidence-list "F:\WardriveAnalytics" --limit 2 --compact
• python wardrive_cli.py runs-list "F:\WardriveAnalytics" --limit 2 --compact
• JSONL stdio bridge smoke test with project_summary.
• No-evidence analyze-project returns structured JSON error.
• Full headless analysis was previously verified after the spreadsheet sanitation fix.

Handoff Notes

• The bridge is local-only. It opens no network port.
• analyze-project mutates by creating a new run folder; read-only commands should be preferred for agent inspection.
• Generated reports, app logs, and error_reports/ are ignored.

Files To Avoid Concurrently

• gui_step3d_scene.py
• project_vault.py
• core/writers.py
• wardrive_cli.py
• wardrive_service.py

Risk / Follow-Up

• Next logical step is an MCP wrapper around wardrive_service.py so agents can call named tools directly.
• Consider splitting GUI visual polish from backend/bridge work in future PRs because gui_step3d_scene.py is large.