AutoRescue workflows

AutoRescue stitches Airia orchestration, Apify automations, and sponsor APIs into a post-purchase rescue loop that detects delay/payment incidents and resolves them via policy-backed agents. Start with OVERVIEW.md for the narrative and auto-rescue-usecase.md for the detailed journey log.

Repository overview

• src holds the workflow entrypoint (index.js), shared utilities, and orchestration helpers.
• apps/ contains the demo server, worker, and dashboard shells used in live walkthroughs.
• scripts/ bundles operational helpers (Kafka checks, decision reports, manual replays).
• data/ stores sample events, outbox payloads, and replay fixtures for demos.
• airia-web-apis.json is the OpenAPI 3.1 contract; lint it before committing changes.
• auto-rescue-usecase.md and docs/ capture the scenario narrative, policy notes, and outreach playbooks.
• AGENTS.md lists shipped agents, triggers, and evaluation datasets.

Getting started

1. Install Node.js 18+ and npm install dependencies from the repo root.
2. Copy the redacted .env template to .env.local, populate sandbox tokens (Shopify, Twilio, Airia), and keep production secrets outside the repo.
3. Run npm run workflow:simulate to replay the sample delay event; use SIMULATE_AIRIA=1 to mock Airia calls when offline.

Development commands

• npm run workflow:run executes the Temporal-style orchestration against live connectors.
• npm run workflow:apify invokes the Apify actor bridge for marketplace-driven rescues.
• npm run airia:setup provisions agents, triggers, and datasets defined in AGENTS.md.
• npm run server:start and npm run worker:start launch the demo API surface and background worker.
• npm run demo:e2e walks through the scripted end-to-end rescue used in demos.
• npm run dashboard:start serves the lightweight dashboard for timeline visualisations.

API contract workflow

1. Update airia-web-apis.json, keeping properties alphabetised and descriptions clear on enum values.
2. Run jq . airia-web-apis.json for a structural sanity check.
3. Run npx @redocly/cli lint airia-web-apis.json (install once with npm i -g @redocly/cli).
4. Add or refresh a sample request in the relevant Markdown section; for example:

   curl --request POST \
     --url https://sandbox.airia.ai/v1/Webhook/demo-tenant/delay-alert \
     --header 'Authorization: Bearer <api-key>' \
     --header 'Content-Type: application/json' \
     --data @data/events/sample-delay-event.json

Documentation map

• auto-rescue-usecase.md tracks the scenario narrative, linked test tables, and a changelog.
• docs/orchestration-runbook.md details Temporal workflows, retries, and incident states.
• docs/demo-playbook.md covers live demo callouts, voice IVR prompts, and fallback notes.
• Link new reference material back into the main narrative and store it under docs/ using lowercase kebab-case filenames.

Testing and verification

• Record manual test cases in Markdown tables (input, expected outcome, side-effects) and link them from the relevant section of auto-rescue-usecase.md.
• Run npx markdownlint-cli2 auto-rescue-usecase.md AGENTS.md to keep long-form docs consistent.
• Log sandbox integrations exercised (Shopify, Twilio, Airia) along with redacted tokens in .env.local.

Contributing

• Follow Conventional Commits (for example feat: add webhook retry guard) and keep subjects under 72 characters.
• Request reviews from teammates covering the touched surface area (API, orchestration narrative, outreach playbooks).
• Capture verification steps (linting, jq, manual callouts) in pull request summaries so others can reproduce them quickly.