> For the complete documentation index, see [llms.txt](https://docs.fastrouter.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.fastrouter.ai/explore-features/zero-data-retention.md).

# Zero Data Retention

FastRouter lets you restrict any request to Zero Data Retention (ZDR) providers by adding the :zdr suffix to the model slug.

## Introduction

Some workloads can't have prompts or completions stored by a model provider, not even temporarily for logging, abuse monitoring, or training. FastRouter lets you restrict any request to **Zero Data Retention (ZDR)** providers by adding the `:zdr` suffix to the model slug.

When you use `:zdr`, FastRouter routes your request **only** to providers whose settings and terms indicate zero data retention for your request and response data. If no ZDR provider can serve the model, the request fails instead of falling back to another provider.

***

## How it works

Append `:zdr` to any model ID:

```
<provider>/<model>:zdr
```

For example:

| Standard slug                  | ZDR slug                           |
| ------------------------------ | ---------------------------------- |
| `deepseek/deepseek-v4.1-flash` | `deepseek/deepseek-v4.1-flash:zdr` |

With the suffix in place, FastRouter:

1. Filters the model's available providers down to ZDR-compliant providers only.
2. Routes the request among those providers using your normal routing preferences (for example, `provider.sort`).
3. Returns an error if no ZDR-compliant provider is available for that model after all filters are applied.

No other changes to your request are needed. The request and response formats are identical to a standard chat completion.

***

## Supported providers

The following providers are currently treated as Zero Data Retention providers on FastRouter:

| Provider   |
| ---------- |
| Fireworks  |
| Baseten    |
| Groq       |
| Perplexity |
| DeepInfra  |
| Together   |

A model can be used with `:zdr` only if at least one of these providers serves it on FastRouter.

***

## What ZDR does not cover

`:zdr` applies only to the choice of **model provider**, routing your request to providers whose settings and terms indicate zero data retention. It doesn't turn off FastRouter features that need to store or pass along your prompts to work. If you use any of the following on a `:zdr` request, your prompts (and in some cases responses) can be stored:

| Feature                                | Why data can be stored                                                                          |
| -------------------------------------- | ----------------------------------------------------------------------------------------------- |
| **Plugins such as PDF processing**     | The plugin processes your file or prompt content as part of handling the request.               |
| **Web search tool / `:online`**        | Your query is sent to a search service to fetch results, outside the ZDR provider.              |
| **Responses API with preserved state** | Keeping conversation state across requests requires storing earlier prompts and responses.      |
| **FastRouter response caching**        | Prompts and responses are stored so that matching future requests can be served from the cache. |

If you want to minimize data retention, avoid these features on requests that carry sensitive data.

#### Disabling content logging on FastRouter

Separately from provider retention, FastRouter can log prompt and response content for your requests. You can **disable content logging at the API key level** to turn off prompt and response logging for all requests made with that key.

To reduce how much of your data is retained, you can combine these:

1. Use the `:zdr` slug so only ZDR providers serve your requests.
2. Use an API key with content logging disabled.
3. Avoid the features listed above.

***

## Example request

```bash
curl --location 'https://api.fastrouter.ai/api/v1/chat/completions' \
--header 'Authorization: Bearer KEY' \
--header 'Content-Type: application/json' \
--data '{
    "model": "deepseek/deepseek-v4.1-flash:zdr",
    "stream": false,
    "max_completion_tokens": 200,
    "messages": [
        {
            "role": "user",
            "content": "What is 2+2"
        }
    ]
}'
```

## Example response

```json
{
    "choices": [
        {
            "message": {
                "role": "assistant",
                "content": "2 + 2 = 4",
                "annotations": null
            },
            "finish_reason": "stop",
            "index": 0
        }
    ],
    "guardrails": {},
    "citations": null,
    "usage": {
        "chat_id": "fr_768d0989-2830-4f47-af04-00286e7e119c",
        "completion_tokens": 8,
        "object": "",
        "completion_tokens_details": {},
        "prompt_tokens": 10,
        "cost": 0.000006799999999999999,
        "prompt_tokens_details": {
            "cached_tokens": 0
        },
        "total_tokens": 18,
        "provider": "deepinfra"
    },
    "id": "fr_768d0989-2830-4f47-af04-00286e7e119c",
    "model": "deepseek-ai/DeepSeek-V4.1-Flash",
    "object": "chat.completion",
    "service_tier": "default"
}
```

The `usage.provider` field tells you which provider served the request. For a `:zdr` request, this will always be one of the supported ZDR providers listed above. In this example, the request was served by **DeepInfra**.

***

## Combining `:zdr` with provider routing

`:zdr` works alongside the `provider` object. ZDR filtering is applied first, and your provider preferences are then applied to the remaining ZDR providers.

For example, to route to the highest-throughput ZDR provider:

```json
{
    "model": "deepseek/deepseek-v4.1-flash:zdr",
    "messages": [
        { "role": "user", "content": "What is 2+2" }
    ],
    "provider": {
        "sort": "throughput"
    }
}
```

If you use `provider.only` to pin specific providers, include only ZDR providers. If none of the providers left after filtering are ZDR-compliant, the request is rejected.

***

## Error handling

If no ZDR-compliant provider can serve the request, FastRouter returns an error rather than routing to a non-ZDR provider.

#### Example: no ZDR provider available

Request:

```json
{
    "model": "deepseek/deepseek-v4.1-flash:zdr",
    "stream": false,
    "max_completion_tokens": 200,
    "messages": [
        {
            "role": "user",
            "content": "What is 2+2"
        }
    ],
    "provider": {
        "sort": "throughput",
        "only": [
            "non-zdr-provider"
        ]
    }
}
```

Response:

```json
{
    "error": "deepseek/deepseek-v4.1-flash is not a valid model ID or no providers found. No provider for this model supports Zero Data Retention (:zdr)"
}
```

Here, `provider.only` limits routing to a provider that isn't ZDR-compliant, so no eligible provider is left.

This error can occur when:

* The model isn't served by any of the supported ZDR providers.
* Your `provider.only` list (or other provider filters) excludes every ZDR provider serving the model.
* The model ID itself is invalid.

**To resolve it**, remove or adjust the conflicting provider filters, choose a model served by at least one ZDR provider, or check the model ID for typos.

***

## FAQ

**Does `:zdr` change the response format?** No. Requests and responses follow the standard chat completions format. The only visible difference is that `usage.provider` will always be a ZDR provider.

**Will FastRouter ever fall back to a non-ZDR provider?** No. When `:zdr` is set, requests are routed only to ZDR-compliant providers. If none are available, the request fails with an error.

**How do I confirm which provider handled my request?** Check the `usage.provider` field in the response.

**Does `:zdr` mean my prompts are never stored anywhere?** No. `:zdr` affects only which model provider is selected, based on that provider's settings and terms. Plugins such as PDF processing, the web search tool / `:online`, stateful Responses API usage, and FastRouter response caching can store prompts.

**Can I use `:zdr` with every model?** Only with models served by at least one supported ZDR provider. For other models, the request returns the error shown above.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.fastrouter.ai/explore-features/zero-data-retention.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.