Skip to content

About

Interactive independent portfolio demonstration with public source and documented limitations.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

2 Commits

Folders and files

Repository files navigation

Fieldwork — interactive engineering demonstrations

Five small, original demonstrations of project-specific delivery patterns. Developed with AI assistance by Mitchell Graham. All records, prices, and evidence are synthetic. These are unsolicited concepts, not completed client engagements.

Sample Live demonstration What actually works
Order Guard Try it Simulated order and purchase deduplication, replay controls, per-tab persistence, comparison counters
Variant pricing Try it 315 variants, quantity tiers, exact-cent totals, full review CSV and draft product CSV
Scheduled reports Try it Select fixture and format, generate delimited output, download artifact/config, inspect run history
Missed-call follow-up Try it Duplicate-event handling, deterministic qualification, STOP suppression, local image filename log
Approval desk Try it Market configuration, explainable scoring, approve/reject notes, decision history, approved-only export

Run and test

Python 3: python -m http.server 8000 --bind 127.0.0.1, then open http://127.0.0.1:8000/.

Node.js 20+: node --test model.test.mjs.

No production dependencies or build process. HTML/CSS and native JavaScript modules; pure model functions are separately testable. User text is rendered with textContent. CSV exports quote cells and neutralize spreadsheet formula prefixes.

Boundaries

  • No provider accounts, API keys, network integrations, real customer information, payments, or outgoing messages.
  • The original four samples reset on reload. Order Guard uses per-tab session storage; none is a production persistence layer.
  • Product CSV is a reviewable draft, not proof of a successful Shopify import. Tier configuration and checkout pricing need platform/app validation.
  • Reports use local fixtures, not an Oracle query or GitLab pipeline. Production needs runner, Vault, artifacts, scheduling, and email integration.
  • Conversation replies are deterministic, not generated by an LLM. Production SMS/MMS requires consent handling, authenticated webhooks, carrier setup, persistence, and delivery testing.
  • Prospect signals and organizations are fictional. Production requires provider connectors, dated evidence, durable deduplication, reviewer authentication, and audit storage.

Reference interfaces

Public source is provided for inspection; no additional reuse license is granted.

About

Interactive independent portfolio demonstration with public source and documented limitations.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages