Back to work
DevOps · Featured
SignalForge
Collects infra snapshots and telemetry, ranks findings, and adds an optional AI summary for triage.
Python · React · TypeScript · Observability · OpenTelemetry · LLM
Overview
SignalForge collects infra snapshots and telemetry, ranks findings, and adds an optional AI summary for triage.
- Ingest artifacts from observability and platform systems.
- Normalize them into comparable findings.
- Run one AI pass for explanation and prioritization.
- Leave remediation out of scope until the safety model is defined.
Screenshots
Design choices
- Structured findings come first; model output is secondary.
- Artifacts are normalized before any AI enrichment.
- A human approves any follow-up action.
Source
- Repository: https://github.com/Canepro/signalforge
Outcomes
Reproducible infra findings with a separate AI explanation pass. Humans approve any action.
Challenges and solutions
Challenges
- Turning messy telemetry into comparable findings without extra noise
- Making findings readable enough to act on quickly
- Keeping model output in a narrow, reviewable role
- Keeping diagnostics separate from remediation until the safety model is defined
Solutions
- Strict ingest and normalisation before any AI enrichment
- Structured findings stored separately from LLM explanation and ranking
- One optional AI pass with explicit trust boundaries
- Humans approve any follow-up action
Technology stack
Backend
Python
Frontend
React · TypeScript
AI
LLM (constrained pass)
Observability
OpenTelemetry
Platform
OTLP ingestion · Structured diagnostics artifacts