DWS AI Gateway API
Use AI Gateway to call managed AI models through your Nutrient DWS account. Send prompts to stable Nutrient model aliases and pay for token usage in USD.
Choose a model
| Alias | Starting point |
|---|---|
nutrient-s | Lower-cost generation and everyday prompts. |
nutrient-m | A balance of capability and cost for general workloads. |
nutrient-l | More demanding analysis and complex tasks. |
nutrient-document-decision | Yes/no, choice, and score questions over text or images, through the decision API. Recommended. |
nutrient-systemone | Yes/no, choice, and score questions through the decision API. Text-only. |
Evaluate the model with your own inputs before selecting it for production. See the token rate table for current model pricing.
Choose an API
Choose the API for your task:
| Task | Endpoint |
|---|---|
| Generation and reasoning with function tools | POST /responses |
| Existing Chat Completions integrations | POST /chat/completions |
| Yes/no, choice, and score questions | POST /systemone |
Use Responses for new generation integrations and the decision API for structured questions. Existing Chat Completions integrations can keep calling /chat/completions with the same model aliases. See getting started for SDK configuration and the API reference for request parameters and supported capabilities.
The /models endpoint supports SDK model discovery. Use the model table above and the token rate table to choose a model and compare prices.
Essential guides
Get an API key and send your first request with curl, JavaScript, or Python.
Ask yes/no, choice, or score questions and receive probabilities.
Understand token billing, included usage, test keys, and pay-as-you-go billing.
Use AI Gateway with documents
For document parsing with page coordinates, citations, or schema-shaped extraction, use the DWS Data Extraction API. Send the extracted text or Markdown to AI Gateway for summarization or other model tasks. Each API is billed separately.
Live AI Gateway keys also support model-compatible image and file inputs. These are model inputs, not the Data Extraction API’s structured extraction response.
Next steps
- API reference — Request and response schemas, limits, and error codes.
- Errors and limits — Handle authentication, quota, and rate-limit errors.
- Security and privacy — Review key handling, provider processing, and retention.
- Agent skill — Give AI coding agents like Claude Code and Codex up-to-date knowledge to build an AI Gateway integration.