Skip to main content

Interface: AIGatewayInferenceClientOptions

Defined in: src/ai-gateway/inference-client.ts:73

Runtime connection options; no SCM OAuth credentials are accepted.

Example​

`new AIGatewayInferenceClient({ endpoint: 'https://gateway.example/v1', apiKey: process.env.PANW_AI_GW_INFERENCE_API_KEY });`

Properties​

endpoint?​

optional endpoint?: string;

Defined in: src/ai-gateway/inference-client.ts:75

Full runtime base URL including its API prefix. Falls back to PANW_AI_GW_INFERENCE_ENDPOINT.

apiKey?​

optional apiKey?: string;

Defined in: src/ai-gateway/inference-client.ts:77

Runtime key sent as x-portkey-api-key. Falls back to PANW_AI_GW_INFERENCE_API_KEY.

timeoutMs?​

optional timeoutMs?: number;

Defined in: src/ai-gateway/inference-client.ts:79

Default total per-attempt deadline, including stream consumption. Default: 60,000 ms.

numRetries?​

optional numRetries?: number;

Defined in: src/ai-gateway/inference-client.ts:81

Explicit replay budget, 0–5. Default: 0 to avoid duplicate billable generations.

maxEventBytes?​

optional maxEventBytes?: number;

Defined in: src/ai-gateway/inference-client.ts:83

Maximum SSE event size, default 1 MiB.

fetch?​

optional fetch?: (input, init?) => Promise<Response>;

Defined in: src/ai-gateway/inference-client.ts:85

Optional caller-owned fetch implementation for explicit proxy/network integration.

Parameters​

ParameterType
inputstring | URL | Request
init?RequestInit

Returns​

Promise<Response>