DIALOGIA

DIALOGIA is a bilingual, governed AI platform for interreligious dialogue and research. It combines an expert-approved source corpus, independently grounded Jewish/Christian/Islamic perspectives, human-controlled dialogue mediation, a narrative research atlas, and evidence-based ethics evaluation.

The product rules are defined in the product constitution, mapped back to the research proposal in the traceability matrix, and tested against the production acceptance criteria.

Local application

1. Copy .env.example to .env.local and set the browser-safe Supabase URL and anon key.
2. Keep GEMINI_API_KEY only in supabase/.env.local or Supabase Edge Function secrets.
3. Public YouTube reviews use the existing server-side AI credential. The server sends the public video URL directly for audiovisual transcription; there is no separate YouTube, OAuth, caption, or transcription-provider credential for the user to configure.

The automatic path is limited to publicly accessible videos and requires an explicit user attestation of ownership, permission, licence, or another applicable legal basis. Google's audiovisual analysis service processes the public URL temporarily; DIALOGIA does not retain the raw transcript.

Verification

• npm test runs deterministic application and live turn-boundary tests.
• npm run test:ai performs real configured-model smoke checks for structured reasoning, embeddings, speech synthesis, and live audio. It reads the local server-only environment and never prints the key.

5. Install dependencies with npm install.
6. Run the app with npm run dev.

Quality checks:

npm run lint
npm test
npm run build

Hosted backend

Apply the migrations in supabase/migrations in timestamp order, configure Edge Function secrets from supabase/.env.local, and deploy every function in supabase/functions.

After the first trusted user signs up, assign that profile the admin role through an authenticated project-owner operation. All later role changes occur through the application and are written to audit_log.

Never present generated foundation data as independent expert validation. A project owner may authorize a transparent operational foundation so the product can run, but scientific publication and research-training status still require the recorded multi-expert review defined by the corpus protocol.

Vercel deployment

The frontend is configured for Vercel as a Vite single-page application. In the Vercel project settings, set the following browser-safe environment variables:

• VITE_SUPABASE_URL
• VITE_SUPABASE_ANON_KEY

Keep GEMINI_API_KEY and all other server-only secrets in Supabase Edge Function secrets, never in Vercel or variables prefixed with VITE_. Deploy the database migrations and Edge Functions through Supabase separately, then set the deployed Vercel origin in the backend's allowed origins.

Desktop path

The renderer contains no Gemini or service-role secret. Authentication, storage, database, and model work cross the same Supabase API boundary, so a future Electron shell can reuse the web application without moving privileged credentials into the desktop client.