AI Knowledge Heatmap
An interactive knowledge survey mapping 153 AI/ML topics across 12 domains,
scored on a 4-level scale (untouched → heard of → can explain → built/taught)
with evidence behind every cell and free learning resources for every topic.
Two parts: a static dashboard (index.html) that loads scores from a
canonical JSON profile, and an interactive quiz (quiz.html) backed by a
FastAPI + LLM evaluation engine.
How it deploys
Google Cloud Run (.github/workflows/deploy-cloudrun.yml) — builds the
Dockerfile (FastAPI backend + static files) and deploys it as a container
service on every push to main. This is the full app: dashboard, quiz, and
scoring backend. Cloud Run scales to zero when idle, so it stays within the
free tier for low-traffic personal use.
One-time setup for the Cloud Run workflow (requires a Google Cloud account):
- Create a GCP project and enable the Cloud Run, Cloud Build, and Artifact
Registry APIs.
- Create a service account with the Cloud Run Admin, Cloud Build Editor,
Artifact Registry Writer, and Service Account User roles.
- Set up Workload Identity Federation so GitHub Actions can authenticate as
that service account without a downloadable JSON key (many GCP projects
now block key creation by org policy, and WIF is the recommended
approach anyway) — see the commands in
.github/workflows/deploy-cloudrun.yml's auth step for what it expects.
- In the repo's GitHub Settings → Secrets, add:
GCP_PROJECT_ID — your GCP project IDGCP_WORKLOAD_IDENTITY_PROVIDER — full resource name of the WIF
provider (projects/<number>/locations/global/workloadIdentityPools/<pool>/providers/<provider>)GCP_SERVICE_ACCOUNT — the service account's emailOLLAMA_API_KEY — your Ollama Cloud key (or adapt the workflow/backend
to use OPENROUTER_API_KEY instead, see backend/.env.example)
.gitlab-ci.yml is a leftover static-only mirror (copies index.html to
public/ for GitLab Pages) and isn't part of the active deploy path.
Updating the content
- Scores & reasoning: Edit
assets/data/amit-profile.json directly on
GitHub — update a score, push, and the dashboard reflects it on next page
load. Format: {profileName, exportedAt, totalTopics, evaluatedCount, domainNotes, domains} with each topic carrying {score, reasoning, evidence, resources}.
- Topic list: Edit
assets/js/data.js — the DOMAINS array used by the
quiz. Each topic is [name, level, reasoning, evidence, resources].
- Learning resources: Each topic carries a
resources array of
{title, url} objects. These appear in the dashboard's detail rail as
"Learn more" links when a cell is selected.
