Privacy and compliance · 2026-09-25

The difference between an EVREN request record and a prompt record

The two traces a request to EVREN can leave: a record of the call itself and a record of the content sent. What is known about each, what is not, and what LLMTR keeps, with sources.

Diagram splitting a request to EVREN into a request record of time, model and token count on one side and a content record of text and response on the other.

Short answer

An API call can leave two separate traces. One is a record of the request itself: when it arrived, which model it went to, how many tokens it used and which status code it ended with. The other is a record of the content: the text, document or image that was sent and the answer the model gave. Privacy discussions often treat the two as one thing, but the existence of one does not show the existence of the other.

We recommend using EVREN through LLMTR: you register your own EVREN key once in Settings, and the same code then reaches both EVREN models and the rest of the catalog. The key stays yours, and EVREN's own logging and retention policy still applies unchanged.

What a request record tells you

A request record is the information kept to operate a service. Billing, quota, rate limiting and debugging all depend on it; no paid or quota-limited API can work without it. It contains none of the text a user wrote, but it does show who used how much, and when.

That EVREN keeps such a record is clear from its own schema. The /v1/quota endpoint returns your account's token cap over a sliding window and its credit balance, so EVREN counts usage per account. A separate endpoint reports whether a given request completed and whether its charge was finalised.

What a prompt record tells you

A prompt record is the content itself. System instructions, conversation history, attached files, text extracted from images and the model's answer can all end up in it. If personal data, a trade secret or a customer's contract was sent, this is where the real risk sits.

With EVREN it is also worth remembering that such a record can be tied to an identity. You sign in to the platform with e-Devlet, and the API key is created on that personal account. If the content of a prompt is logged, the log can be linked to a verified person. EVREN's schema also shows a content moderation layer running on generative endpoints; that layer sees the content, but whether it keeps records of its own is not public.

What EVREN's request-status endpoint shows

EVREN's OpenAPI schema documents an endpoint called /v1/requests/{request_id}. It returns the execution and billing state of each request: whether it completed, whether the charge was finalised, and the credit charged and pending. The schema states separately that the completed response body is not returned from this endpoint.

That shows a per-request record exists. It says nothing about whether the response body is stored anywhere else; an endpoint that does not return content does not prove that content is kept nowhere. The endpoint belongs to EVREN and is called directly with your own EVREN key; LLMTR does not serve it.

What each side keeps

When you use EVREN through LLMTR, a request passes through two separate systems. The table separates what is known on each side today. The LLMTR column describes our own system; the EVREN column describes what we could find in public sources.

Request and content records: LLMTR and EVREN (checked 25 September 2026)
RecordLLMTREVREN
Time, model, status code, durationKept in the usage logA per-request status record exists
Token counts and costKept; cost is $0 on EVREN rowsQuota and credit are charged to the account
Prompt and response textNot written to the databaseNo public information found
Retention period and training useNot applicable, content is not writtenNo public information found
Your EVREN API keyStored encrypted, only the last four characters are shownBound to your e-Devlet account

What this means before you send sensitive data

Because there is no public retention commitment for EVREN on the content side, the safe assumption is that the provider may keep logs. The data notice on the model cards says the same. This is not an accusation; it is the way to treat something unknown as unknown.

In practice it means making content smaller before it leaves: removing names and numbers from a task that does not need identity, and using sample data rather than real customer data while testing. EVREN's terms can be read with an API key from the /v1/terms/text endpoint. For organisational use, looking for the retention and training clauses there and, where needed, asking EVREN for a written answer is the only way to close the gap. The sensitive-data article in this series covers redaction techniques in detail.

Frequently asked questions

Should I use EVREN directly or through LLMTR?

We recommend using EVREN through LLMTR. Your requests still go to EVREN on your own EVREN key and LLMTR charges nothing for them; the difference is that the same code also reaches the other models in the catalog. EVREN's logging and retention policy is not affected by the choice.

Does a request record prove that my prompt is retained?

No. A request record shows that a call happened and how it ended. Whether the text sent was retained is a separate question, and for EVREN there is no public answer to it.

Does EVREN's request-status endpoint return the response text?

No. According to the OpenAPI schema it returns execution and billing state; the completed response body is not returned from this endpoint. It does not say whether the body is stored elsewhere.

Can I call EVREN's request-status endpoint through LLMTR?

No. It is one of EVREN's own account endpoints and is called directly against EVREN with your own EVREN key. The EVREN rows served through LLMTR are for model calls.

Does LLMTR store request content?

No. Request and response bodies are not written to the LLMTR database. The usage log keeps the model, token counts, cost, status code and duration.

Related posts