# Advisory AI CI Runner Harness (DEVOPS-AIAI-31-001)

Purpose: deterministic, offline-friendly CI harness for Advisory AI service/worker. Produces warmed-cache restore, build binlog, and TRX outputs for the core test suite so downstream sprints can validate without bespoke pipelines.

Usage
- From repo root run: `ops/devops/advisoryai-ci-runner/run-advisoryai-ci.sh`
- Outputs land in `ops/devops/artifacts/advisoryai-ci/<UTC timestamp>/`:
  - `build.binlog` (solution build)
  - `tests/advisoryai.trx` (VSTest results)
  - `summary.json` (paths + hashes + durations)

Environment
- Defaults: `DOTNET_CLI_TELEMETRY_OPTOUT=1`, `DOTNET_SKIP_FIRST_TIME_EXPERIENCE=1`, `NUGET_PACKAGES=$REPO/.nuget/packages`.
- Sources default to `local-nugets` then the warmed cache; override via `NUGET_SOURCES` (semicolon-separated).
- No external services required; tests are isolated/local.

What it does
1) Warm NuGet cache from `local-nugets/` into `$NUGET_PACKAGES` for air-gap parity.
2) `dotnet restore` + `dotnet build` on `src/AdvisoryAI/StellaOps.AdvisoryAI.sln` with `/bl`.
3) Run the AdvisoryAI test project (`__Tests/StellaOps.AdvisoryAI.Tests`) with TRX output; optional `TEST_FILTER` env narrows scope.
4) Emit `summary.json` with artefact paths and SHA256s for reproducibility.

Notes
- Timestamped output folders keep ordering deterministic; consumers should sort lexicographically.
- Use `TEST_FILTER="Name~Inference"` to target inference/monitoring-specific tests when iterating.