AI Providers¶

Supported providers¶

• openai
• gemini (AI Studio)
• vertex_gemini (Vertex AI)
• openrouter
• perplexity

Capability matrix (v0.0.4 release line)¶

Unsupported operation/provider pairs throw AstAiCapabilityError.

Script property keys¶

OpenAI¶

• OPENAI_API_KEY
• OPENAI_MODEL

Gemini (AI Studio)¶

• GEMINI_API_KEY
• GEMINI_MODEL

Vertex Gemini¶

• VERTEX_PROJECT_ID
• VERTEX_LOCATION
• VERTEX_GEMINI_MODEL
• VERTEX_SERVICE_ACCOUNT_JSON (optional; enables service-account auth mode)
• VERTEX_AUTH_MODE (optional: oauth | service_account | auto, default auto)

Vertex auth modes:

• oauth: require OAuth token path (auth.oauthToken / auth.accessToken or ScriptApp.getOAuthToken()).
• service_account: require service account JSON (auth.serviceAccountJson or VERTEX_SERVICE_ACCOUNT_JSON) and exchange JWT for access token.
• auto (default): prefer service-account JSON when present, otherwise fallback to OAuth.

Service-account token_uri is validated against an allowlist: - https://oauth2.googleapis.com/token - https://www.googleapis.com/oauth2/v4/token

OpenRouter¶

• OPENROUTER_API_KEY
• OPENROUTER_MODEL
• OPENROUTER_HTTP_REFERER (optional)
• OPENROUTER_X_TITLE (optional)

Perplexity¶

• PERPLEXITY_API_KEY
• PERPLEXITY_MODEL

Precedence for auth/model values¶

For each required key:

1. per-call request.auth.* (or request.model)
2. runtime config from ASTX.AI.configure(...)
3. script properties
4. throw AstAiAuthError

For Vertex service-account JSON specifically:

1. request.auth.serviceAccountJson (object or JSON string)
2. ASTX.AI.configure({ VERTEX_SERVICE_ACCOUNT_JSON: ... })
3. script property VERTEX_SERVICE_ACCOUNT_JSON
4. fallback to OAuth path when authMode=auto

Notes¶

• providerOptions are passed through to provider payloads for advanced use.
• Keep provider/model mappings explicit in production automations for deterministic behavior.