fix(ai): repair resource search, add AI tools and console commands

[roncodes]

Why

Auditing the Fleetbase AI extension against a production log export turned up a hard failure on this side: fleet-ops.search_resources threw on every call because Sensor and the resource search still referenced sensor_type, a column an earlier migration had dropped. The full SQL error, company UUID included, was handed to the model provider, and the turns were still recorded as answered.

Fixes

• sensor_type → type in Sensor (slug, activity log, filter params, scopeByType, alert context) and in the resource search.
• Each resource type is searched in its own try/catch, so one broken type no longer kills the whole search; it is reported as an unavailable search instead.
• Search terms are built from reference-like tokens only, with stopwords: no more LIKE against status, type or uuid, and the search is skipped when nothing useful is left. (The old behavior matched half the database on words like "the".)
• Order amount thresholds are converted from dollars to minor units using the currency's exponent, and insights expose the currency.
• applyDirectivesForPermissions is applied to the order insights, asset status and search queries, so AI answers stay inside what the user is allowed to see.

New surface for the AI extension

• propose_create_order and fleetops_search tools (registered only when the AI extension's tool interface is present).
• FleetOpsAiConsoleCommands: open, create, view and import commands for orders, drivers, vehicles, contacts, customers, places, fleets, issues, work orders and maintenances, plus the Fleet-Ops settings pages. They are built on the existing *-actions services. The model can only propose one; the user confirms and the server re-authorizes before anything runs.
• CreateOrderPreviewCapability::buildDraft accepts a tool-supplied draft and an existing draft, instead of only parsing the prompt.

Testing

2882 tests, 0 failures, 8 errors. The 8 errors are the local SQLite build lacking CONCAT and fail the same way on main; they pass on MySQL in CI. Run one file per process, as this suite requires. php-cs-fixer is clean on the changed files.

Coverage for the new files has not been measured, and this repo's CI gate is 100%, so that may need a follow-up pass.

Related

The AI extension side is in fleetbase/ai; fleetbase/iam-engine#33 and fleetbase/dev-engine#46 add the resource-action services the console commands call into.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.99%. Comparing base (14601ca) to head (9b72307).
⚠️ Report is 6 commits behind head on main.

Additional details and impacted files

@@              Coverage Diff              @@
##                main     #330      +/-   ##
=============================================
- Coverage     100.00%   99.99%   -0.01%     
- Complexity     11887    11990     +103     
=============================================
  Files            582      586       +4     
  Lines          44715    45132     +417     
=============================================
+ Hits           44715    45131     +416     
- Misses             0        1       +1     

Flag	Coverage Δ
backend	99.99% <100.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.