oice-to-Action Support Snippet Agent — Devlog Update (Hackathon Build)
Shipping a production-like, judge-safe Support Automation Command Center that converts voice transcript or text into deterministic structured outputs with guardrails + telemetry.
What’s live • Cloud Run: https://voice-to-action-agent-zbluqfbniq-ew.a.run.app • Health: https://voice-to-action-agent-zbluqfbniq-ew.a.run.app/api/health • Guardian: https://voice-to-action-agent-zbluqfbniq-ew.a.run.app/api/guardian • Metrics: https://voice-to-action-agent-zbluqfbniq-ew.a.run.app/api/metrics • Firebase: https://chatgpt-ops.web.app • Health: https://chatgpt-ops.web.app/health.json • Guardian: https://chatgpt-ops.web.app/api/guardian • Metrics: https://chatgpt-ops.web.app/api/metrics
Core output contract (contract-first AI)
Single request → consistent response schema: • transcript • executive summary • action items • email draft • audit trail • meta diagnostics (requestId, model, latency, validation, fallback)
Reliability + safety (demo-safe) • JSON Schema constrained Gemini output + server-side Zod validation • Deterministic safety_check + quality scoring + audit notes • Runtime Guardian loop → healthy | degraded | critical • Security shield: rate limits, payload caps, abuse signals & temporary blocking
Reproducible proof (for judges)
Quickstart (≈5 min): npm ci cp .env.local.example .env.local npm run dev
Verification commands: npm run lint npm run typecheck npm run test npm run eval npm run build npm run scan npm run judge:verify
Ops automation shipped • GitHub Actions deploy: .github/workflows/deploy-gcp.yml • Cloud Run deploy scripts: scripts/deploy.sh, scripts/precheck-cloudrun.sh • Firebase deploy: scripts/deploy-firebase.sh • Judge bundle + verification: scripts/release-bundle.sh, scripts/judge-verify.sh • Infra: cloudbuild.yaml, infra/main.tf
Log in or sign up for Devpost to join the conversation.