Home / Ops / Setup and Activation

Activation how-to (tonight and this week)

Updated Jun 17, 2026 · Affirmology_Activation_HowTo_v1.md

Activation how-to (tonight and this week)

You buy and click; Claude Code does the deploys, builds, and the corpus run. Order for tonight: buy everything, rotate the OpenRouter key, kick the corpus overnight run, start the cloud move, get on Higgsfield.

A. Corpus, overnight upgrade + nightly crawl

After adding OpenRouter credit (~$10-20) and ROTATING the old key: 1. Paste to Claude Code: "Run the Nemotron quality gate (about 75 human-design docs, scored against the Haiku-structured ones) and show me the comparison. On my OK, run the overnight backlog (numerology, human_design, gene_keys, vedic) with --backend openrouter and a cost cap." 2. It structures overnight; read the morning report. This week, make it automatic: 3. Paste to Claude Code: "Set up a nightly launchd job on this Mac that runs the corpus crawl + structure (Nemotron backend) + the morning quality report, and logs each run." Then the corpus keeps crawling and improving nightly on its own.

B. Cloud move, the agent system off your laptop (do this BEFORE Atlas)

B1. Set up R2 (Cloudflare storage), click by click

1. Cloudflare dashboard, left sidebar, R2. Click to enable it (add a payment method; R2's free tier is 10GB storage with no egress fees, so it is ~free at your scale).
2. Create bucket: R2, Create bucket, name it affirmology-media, location Automatic, Create.
3. Create an API token: in R2, open "Manage R2 API Tokens" (or API Tokens), Create API Token, permission "Object Read & Write," scope it to the affirmology-media bucket, Create. Copy and save securely: the Access Key ID, the Secret Access Key, your Account ID, and the S3 endpoint (it looks like https://<account-id>.r2.cloudflarestorage.com). These are secrets; do not put them in any file or repo.
4. Leave the bucket private. The Studio reads and writes it with those keys and serves media through the app. (Optional later: connect a custom domain like media.affirmology.ai to the bucket for direct public delivery.)

B2. Set up Render and deploy

1. render.com, sign up, connect GitHub, New Web Service from the affirmology-studio repo (it has render.yaml + the Dockerfile). Pick the Standard instance (2GB RAM, ~$25/mo).
2. In Render, set env secrets: ELEVENLABS_API_KEY, ANTHROPIC_API_KEY, STUDIO_AUTH_TOKEN, STUDIO_REQUIRE_AUTH=true, and for storage: STUDIO_STORAGE=r2 plus the R2 values (account id, access key id, secret access key, bucket=affirmology-media, endpoint). (Claude Code will confirm the exact R2 variable names the Studio's storage.py expects.)
3. Paste to Claude Code: "Deploy affirmology-studio to Render from render.yaml + the Dockerfile; switch storage to R2; ALSO get the corpus DB to the cloud (export /Volumes/Affirmology/corpus.db, upload it to R2 or the Render disk, and point AFFIRMOLOGY_CORPUS_DB at it) so cloud generation stays corpus-grounded; then run one real render in the cloud to verify, and show me the result."
4. Repoint the domain (Cloudflare, you): change studio.affirmology.ai to hit the new Render URL instead of the laptop Tailscale funnel. Claude Code gives you the exact one-line change (the worker's STUDIO_FUNNEL_URL or the route); you confirm it in the dashboard.
5. Verify: studio.affirmology.ai loads from the cloud, a render runs, media lands in R2. Then you can stop the laptop server and funnel; the Studio is now 24/7. (The corpus's nightly crawl stays on the laptop SSD and syncs its DB up after runs.)

C. Atlas, right after the cloud move

1. Paste to Claude Code: "Build Atlas per Affirmology_Atlas_AppSpec_v1.md on the cloud Studio: make it an installable PWA (manifest + service worker + icon), mobile-optimized; add the Hermes chat mode (chart-grounded, reads jpl-knowledge) as the default with a one-tap toggle to the full Studio; add voicenote capture stored in R2, and team messaging."
2. Install it: on your phone, open studio.affirmology.ai and Add to Home Screen; on your Mac, Install from the browser address bar. Same login (Cloudflare Access). Note: Atlas needs the cloud up first, so it follows the cloud move (likely a day or two), not literally tonight.

D. Higgsfield, tonight

1. Subscribe to the mid tier at higgsfield.ai (start mid, upgrade or buy a top-up pack if you hit the ceiling).
2. Open your Affirmology_HiggsfieldPromptKit docs.
3. Generate ONE asset first (the meet-cute scene or a brand-story shot) to learn the flow and gauge your credit burn, then scale.

Tonight vs this week

• Tonight: buy all; rotate the OpenRouter key; kick the corpus overnight run; start the cloud move (R2 + Render deploy + corpus sync + repoint); get on Higgsfield and generate your first asset.
• This week: nightly-crawl automation, the Atlas build, final cloud verification, then retire the laptop dependency for the Studio.