Build with Primex
Start with a working operation, then add the features your application needs. Use Getting Started for your first text request or choose a specialized workload below. All use your Primex API key and account.
| Workload | Start here | Continue with |
|---|---|---|
| Text and reasoning | Chat Completions | Streaming, reasoning, structured outputs |
| Agents and tools | Tool calling | Responses, Messages, MCP |
| Images and video | Images | Image recipe, video jobs |
| Voice and audio | Audio | Transcription, speech and sound effects; transcribe and translate |
| Documents and vision | Files | OCR, document extraction and image analysis, vision inputs |
| Retrieval and search | Embeddings | Reranking, RAG recipe, provider-run search |
| Translation and language | Language APIs | Translator catalogue, text analysis |
| Content safety | Moderation | Current safety offerings |
From an experiment to an application
- Pick a currently compatible offering from the Catalog.
- Test the operation in Playground where supported.
- Use Setup for production request examples.
- Handle the documented response, errors and any file/job lifecycle.
- Inspect request activity in Usage, then review production guidance before expanding traffic.
Model and service capabilities vary. Use the capability map and individual endpoint references to confirm features and units. Client-run functions and provider-run tools have different execution and billing contracts.
Run upload → document extraction for private PDF inputs and structured results.