AI Endpoints Overview

PickLayer provides a unified OpenAI-compatible API surface that works with multiple AI providers. All endpoints require a model field with a provider prefix (e.g., openai/gpt-4o, openrouter/anthropic/claude-3-opus, azure/gpt-4o-prod).

Base URLs & Routing

Canonical path shape


https://{HOST}/api/org/:orgSlug/app/:appSlug/:env?/v1/<openai_path>

:env defaults to prod if omitted
Everything after /v1/ matches the OpenAI path exactly

Example:


https://api.picklayer.dev/api/org/acme/app/customer-support/v1/chat/completions

Custom domains

Custom domains strip the org/app prefix; PickLayer infers tenancy from DNS metadata:


https://api.mychatapp.com/v1/chat/completions

Available Endpoints

Endpoint	Method	Description
Chat Completions	`POST /v1/chat/completions`	Generate chat responses (OpenAI-compatible)
Responses API	`POST /v1/responses`	Multimodal responses (OpenAI Responses format)
Messages API	`POST /v1/messages`	Anthropic Claude native format
Embeddings	`POST /v1/embeddings`	Generate vector embeddings
Audio Transcription	`POST /v1/audio/transcriptions`	Speech-to-text
Text-to-Speech	`POST /v1/audio/speech`	Text-to-speech synthesis
Models	`GET /v1/models`	List available models

Provider Translation Warnings

When sending provider-specific parameters to providers that don’t support them, warnings are returned in metadata.warnings:


{
  "warnings": {
    "unsupported": ["modalities", "audio"],
    "ignored": ["instructions", "parallel_tool_calls"]
  }
}

Enable verbose mode with "metadata": { "verbose_warnings": true } to see all dropped fields.