Multi-model AI

Connect OpenAI, Anthropic, Google keys. Switch models per prompt. Keys never leave your server.

Try the model picker

Select a provider, add a (fake) API key, pick a model, and simulate a prompt run.

Supported providers

Setup flow

1. Get API keys from each provider’s dashboard
2. Open PromptAura → Settings → AI Providers
3. Click “Add Key” for each provider
4. Paste key → encrypted and stored server-side
5. Select default model per provider

Switching models per prompt

When you open a session, the model selector (top-right) shows all your configured models:

Model: ▼ GPT-4o (OpenAI) ▼
       ├─ GPT-4o
       ├─ GPT-4o-mini
       ├─ Claude 3.5 Sonnet (Anthropic)
       └─ Gemini 1.5 Pro (Google)

Pick one → that prompt uses that model. Your choice saves with the session.

Why multi-model?

BYOK (Bring Your Own Keys) architecture

┌─────────────┐     HTTPS      ┌──────────────┐     Encrypted     ┌─────────────┐
│  Your Browser  ─────────────→  │  PromptAura   ───────────────→  │  AI Provider │
│               │  (no keys)     │  Server       │  (decrypt key)    │  (OpenAI)    │
└─────────────┘                └──────────────┘                   └─────────────┘

1. Browser sends: POST /api/v1/run { sessionId, model: "gpt-4o", variables }
2. Server decrypts your OpenAI key
3. Server calls OpenAI API
4. Server streams response back to browser

Your browser never sees the API key. This is safer than client-side AI apps.

Rate limits & costs

PromptAura shows estimated cost before you run (based on input tokens + model pricing).

API access

Multi-model works via REST API too:

curl -X POST https://prompt-aura.ppai.web.id/api/v1/sessions/:id/run \
  -H "Authorization: Bearer pa_xxx..." \
  -H "Content-Type: application/json" \
  -d '{"model": "claude-3-5-sonnet-20241022", "variables": {"topic": "API design"}}'

See API Reference → Run Session for full schema.