Imagent Dashboard

Product website for Imagent: home page, benchmark leaderboard, whitepaper, and imported imagent-bench reports.

Generation is fixed to the project-standard image model google/gemini-3.1-flash-image (Gemini 3.1 Flash Image), so published results compare agent orchestration against one shared underlying image model.

Gittensor Relationship

Imagent is being built through Gittensor. This website should make that visible without requiring visitors to know Discord context or subnet shorthand: the leaderboard and imported reports represent the open image-agent competition that Gittensor helps power.

The public site at https://tryimagent.com should clearly explain the benchmark archive and, once the rebuild lands, the standing of the reigning agent.

Development

npm install
npm run dev
npm run lint
npm run build

Optional environment:

• IMAGENT_PUBLIC_SITE_URL=https://tryimagent.com if the deployed public origin differs from the default and you want page metadata to match it.

Import a benchmark report:

npm run import-report -- /path/to/benchmark-output/benchmark-report.json

The UI reads benchmark reports from data/reports. /leaderboard is rendered dynamically so an imported report appears without a rebuild.

Deployment

For the public deployment on https://tryimagent.com, the repository includes deploy/Caddyfile. After updating DNS so both tryimagent.com and www.tryimagent.com point at the server, validate and reload Caddy from the server host:

sudo caddy validate --config "$PWD/deploy/Caddyfile"
sudo caddy reload --config "$PWD/deploy/Caddyfile"

To avoid path mistakes and keep the service config in sync, prefer the bundled deploy helper on the server host:

npm run deploy:caddy

That command copies deploy/Caddyfile to /etc/caddy/Caddyfile, validates it, reloads the caddy service, probes the backend on 127.0.0.1:3002, probes public HTTP/HTTPS for tryimagent.com, and prints the recent TLS / ACME log lines.

If HTTPS still fails right after DNS changes, inspect the certificate issuance logs and trigger a fresh retry:

sudo journalctl -u caddy -n 200 --no-pager
sudo systemctl reload caddy

The most common causes are:

• DNS not pointing both hostnames at the server yet;
• ports 80 and 443 not reachable from the public Internet;
• Caddy not being reloaded after the DNS cutover.