docs: update changelog and add robust dev-up startup script

This commit is contained in:
2026-02-23 14:22:17 -05:00
parent bb562a91ca
commit 37a9584d0c
2 changed files with 115 additions and 1 deletions

58
scripts/dev-up.ps1 Normal file
View File

@@ -0,0 +1,58 @@
param(
[switch]$Build,
[switch]$BackendOnly,
[int]$HealthTimeoutSeconds = 60
)
$ErrorActionPreference = 'Stop'
function Write-Step([string]$msg) {
Write-Host "`n=== $msg ===" -ForegroundColor Cyan
}
function Wait-Http200([string]$url, [int]$timeoutSeconds) {
$sw = [System.Diagnostics.Stopwatch]::StartNew()
while ($sw.Elapsed.TotalSeconds -lt $timeoutSeconds) {
try {
$resp = Invoke-WebRequest -Uri $url -UseBasicParsing -TimeoutSec 5
if ($resp.StatusCode -eq 200) { return $true }
} catch {}
Start-Sleep -Seconds 2
}
return $false
}
Write-Step 'Starting containers'
$composeArgs = if ($BackendOnly) { 'up -d backend' } else { 'up -d' }
if ($Build) { $composeArgs = $composeArgs.Replace('up -d', 'up -d --build') }
# Use cmd wrapper to safely allow merged output while preserving exit code semantics.
cmd /c "docker compose $composeArgs 2>&1"
if ($LASTEXITCODE -ne 0) {
Write-Error "docker compose failed with exit code $LASTEXITCODE"
exit $LASTEXITCODE
}
Write-Step 'Container status'
docker compose ps
Write-Step 'Health checks'
$backendOk = Wait-Http200 -url 'http://localhost:8000/openapi.json' -timeoutSeconds $HealthTimeoutSeconds
$frontendOk = if ($BackendOnly) { $true } else { Wait-Http200 -url 'http://localhost:3000/' -timeoutSeconds $HealthTimeoutSeconds }
if (-not $backendOk) {
Write-Error 'Backend health check failed: http://localhost:8000/openapi.json did not return HTTP 200 in time.'
exit 1
}
if (-not $frontendOk) {
Write-Error 'Frontend health check failed: http://localhost:3000/ did not return HTTP 200 in time.'
exit 1
}
Write-Step 'All good'
Write-Host 'Backend: http://localhost:8000/openapi.json (OK)' -ForegroundColor Green
if (-not $BackendOnly) {
Write-Host 'Frontend: http://localhost:3000/ (OK)' -ForegroundColor Green
}
exit 0

View File

@@ -38,4 +38,60 @@
| With logged-in users | N/A | **43** (real names only) | | With logged-in users | N/A | **43** (real names only) |
| OS detected | None | **Windows 10** (inferred from hostnames) | | OS detected | None | **Windows 10** (inferred from hostnames) |
| Deduplication | None (same host 20 datasets) | **Full** (by FQDN/ClientId) | | Deduplication | None (same host 20 datasets) | **Full** (by FQDN/ClientId) |
| System account filtering | None | **DWM-*, UMFD-*, LOCAL/NETWORK SERVICE removed** | | System account filtering | None | **DWM-*, UMFD-*, LOCAL/NETWORK SERVICE removed** |
## 2026-02-23: Agent Execution Controls, Learning Mode, and Dev Startup Hardening
### Agent Assist: Explicit Execution + Learning Controls
- **Problem**: Agent behavior was partly implicit (intent-triggered execution only), with no analyst override to force/disable execution and no explicit "learning mode" explainability toggle.
- **Solution**:
- Added `execution_preference` to assist requests (`auto | force | off`).
- Added `learning_mode` flag for analyst-friendly explanations and rationale.
- Preserved deterministic execution path for policy-domain scans while allowing explicit override.
#### Backend Updates
- `backend/app/api/routes/agent_v2.py`
- Extended `AssistRequest` with `execution_preference` and `learning_mode`.
- Added `_should_execute_policy_scan(request)` helper:
- `off`: advisory-only (never execute scan)
- `force`: execute scan regardless of query phrasing
- `auto`: existing intent-based policy execution behavior
- Wired `learning_mode` into agent context calls.
- `backend/app/agents/core_v2.py`
- Extended `AgentContext` with `learning_mode: bool`.
- Prompt construction now adds analyst-teaching/explainability guidance when enabled.
#### Frontend Updates
- `frontend/src/api/client.ts`
- Extended `AssistRequest` with `execution_preference` and `learning_mode`.
- Extended `AssistResponse` with optional `execution` payload.
- `frontend/src/components/AgentPanel.tsx`
- Added Execution selector (`Auto`, `Force execute`, `Advisory only`).
- Added `Learning mode` switch.
- Added execution results accordion (scope, datasets, top domains, hit count, elapsed).
- Cleaned stream update logic to avoid loop-closure lint warnings.
#### Tests and Validation
- `backend/tests/test_agent_policy_execution.py`
- Added regression tests for:
- `execution_preference=off` (stays advisory)
- `execution_preference=force` (executes scanner)
- Validation:
- Backend tests: `test_agent_policy_execution.py` passed.
- Frontend build: clean compile after warning cleanup.
### Frontend Warning Cleanup
- `frontend/src/components/AnalysisDashboard.tsx`
- Removed unused `DeleteIcon` import.
- `frontend/src/components/MitreMatrix.tsx`
- Fixed `useCallback` dependency warning by including `huntList`.
### Dev Reliability: Docker Compose Startup on PowerShell
- **Problem**: Intermittent `docker compose up -d 2>&1` exit code `1` despite healthy/running containers.
- **Root Cause**: PowerShell `2>&1` handling can surface `NativeCommandError` for compose stderr/progress output (false failure signal).
- **Solution**:
- Added `scripts/dev-up.ps1` startup helper to:
- run compose with stable output handling,
- show container status,
- verify backend/frontend readiness,
- return actionable exit codes.
- Updated backend liveness probe to `http://localhost:8000/openapi.json` (current app does not expose `/health`).