A
AI Creator PlatformHelp Center
PLAYBOOKS · FAQ · DEPLOYMENT

Run the platform without decoding the fucking Rosetta Stone.

Customer guidance, Production Team workflows, shadow deployment steps and operator safety rules in one clean place.

01
Customer FAQUsing AI Fans and interacting with creators
What is an AI creator?

An AI creator is a fictional interactive personality with a persistent identity, media, chat, voice, video and creator-specific experiences. The product clearly presents creators as AI-generated rather than real humans.

How do I open a creator profile?

Choose a creator from the featured carousel or full roster. Select View creator to open the profile, gallery, membership and interaction controls.

How does live video work?

Select Enter live room or Video chat. The browser joins a private LiveKit room and receives the creator's avatar, voice and chat response tracks through the existing AI Fans runtime.

What do memberships unlock?

Membership tiers can unlock member feeds, priority chat, voice, video and opt-in communication. The exact benefits are shown before checkout.

Can a creator contact me?

Only through channels you explicitly consent to. SMS, email and call consent are stored per fan, creator and channel and can be revoked.

02
Working with the Production TeamBrief, review, approval and iteration
What does the Production Team do?

The team converts a bounded creator objective into a strategy, content slate, critic verdict, production requests and lineage-ready evidence. It does not automatically publish or deploy.

What should a good content request contain?

Name the creator, business goal, audience, required deliverables, timing, available worlds or resources, brand constraints and the metric you want to improve.

What is the Strategist responsible for?

The Content Strategist proposes the campaign thesis, audience promise, experiment hypothesis, content pillars, success metrics and an ordered slate tied to supplied evidence.

What is the Critic responsible for?

The Critic independently scores brand fit, evidence quality, production clarity, novelty, audience intent, authority safety and slate coherence. It can recommend human review, revisions or a safety block.

What does human approval mean?

Approval advances work only to the next bounded stage. Approving a strategy does not silently authorize Factory execution, GPU rendering, deployment or public publishing.

03
Operator RunbookModel, lineage and Overwatch pilot
Which secrets are required for the observed pilot?

Add CLOUDFLARE_API_TOKEN, CLOUDFLARE_ACCOUNT_ID, MODEL_RUNTIME_UPSTREAM_TOKEN, MODEL_GATEWAY_CAPABILITY_TOKEN, PRODUCTION_STAFF_CAPABILITY_TOKEN, LINEAGE_CAPABILITY_TOKEN and OVERWATCH_OBSERVER_TOKEN to the protected ai-creator-platform-shadow environment.

What is MODEL_RUNTIME_UPSTREAM_TOKEN?

Use the same secret value currently used as AI_FANS_RUNTIME_TOKEN for the authenticated https://aifans-runtime.xyz-labs.xyz/chat boundary. It is not a Cloudflare API token.

How do I run the safe dry run?

Open GitHub Actions, choose Deploy EILA Observed Shadow Pilot, select dry-run, leave confirmation blank and run from main. Review the generated Wrangler plans and artifact.

How do I run the live observed pilot?

Run the same workflow with deploy-and-observe and the exact confirmation phrase DEPLOY_AND_RUN_CHUCK_OBSERVED_SHADOW. The workflow deploys only the model gateway, Production Staff, correlation ledger and read-only Overwatch.

What evidence should I inspect?

Review the live plan response, lineage append results, deterministic replay, Overwatch snapshot, latency, model-call order, terminal status and output fingerprint. Secret values are excluded.

04
Deployment FAQShadow first, production last
Can I deploy the dashboard preview now?

After the UI pull request is merged, run Deploy AI Creator Platform UI Preview. The default mode is dry-run. Live preview deployment requires the exact confirmation phrase DEPLOY_UI_PREVIEW.

What does the preview workflow change?

Only the static Pages preview project ai-creator-platform-shadow. It does not deploy Workers, create routes, provision D1 or R2, run models, invoke Factory, start Unreal or publish social content.

Why are rendering and publishing disabled?

Those actions require separate authenticated capabilities, operator approval, proof artifacts, budget limits, lineage correlation and rollback behavior. The UI may not expose an authority before the backend has earned it.

When does legacy AI Fans retire?

Last. The complete shadow stack must pass creator, account, checkout, chat, video, planning, rendering, draft publication, analytics and rollback tests before routes move. Legacy services remain available through a stability window.

Where should Ollama run?

On the private inference host behind the authenticated EILA model gateway. Do not expose the native Ollama port publicly. The Production Team talks to the gateway, not directly to the machine.