INCIDENTOPS / SKELETON CREW
Two investigators.
One evidence-first foundation.
Linux logs and helpdesk tickets share normalization, local AI, evidence checks and human review.
Recorded walkthrough — not live AI. These are saved results from real local model/API calls on synthetic inputs. This page loads no model and accepts no private files. Run the complete applications locally for live analysis.
Source & local installation · Download test build · How Kiro was used · Watch 2:19 demo video
Validated local result
Explicit signals
A signal match is not proof of a root cause. This demo acknowledgment is not a signed audit trail.
Run the actual applications
Windows · Python3.11 · about2GB RAM for the CPU model · no paid API.
python -m venv .venv
.venv\Scripts\python.exe -m pip install -r requirements-lock.txt
.venv\Scripts\python.exe setup_local_model.py
.venv\Scripts\python.exe run.py both
Two separate folders: apps/linux-investigator and apps/helpdesk-investigator. Shared implementation: incidentops/. Root .kiro contains product steering and specifications.
Exact downloads, checksums and instructions
Measured, with limitations
CPU preflight:2.76s median for short cases;19.69s for one52-line report. Strict baseline9/12; the small model sometimes invents causes or malformed evidence IDs. The application rejects unsupported claims and keeps manual review mandatory.
26 automated tests plus real-model integration and browser review checks. Small synthetic tests are not production accuracy.