LLM Proxy Logs

Find the saved record for a model request and see its usage, cost, and routing decision.

Read one request

GET /api/llm-proxy-logs/{logId} returns the request recorded by Archestra; it does not change the provider response sent to your client.

bash

curl -H "Authorization: arch_..." \
  https://archestra.example.com/api/llm-proxy-logs/<log-id>

The response includes request and response payloads, attribution, cost, and usage. These usage fields are an excerpt from a log response:

json

{
  "inputTokens": 12,
  "inputTokensEstimated": false,
  "outputTokens": null,
  "cacheReadTokens": 0,
  "cacheWriteTokens": null
}

These fields distinguish a known zero from a missing count.

Field	Meaning
`inputTokens`, `outputTokens`	Recorded input and output tokens. `null` means unavailable; `0` is a recorded zero.
`inputTokensEstimated`	Whether Archestra estimated the input count locally instead of using a provider count.
`cacheReadTokens`, `cacheWriteTokens`	Tokens read from or written to the provider cache. `null` means unavailable or not applicable; `0` is a recorded zero.
`cacheWrite1hTokens`	Tokens written to a one-hour provider cache, when the provider reports them.

A numeric count can still be partial if a stream ends early. The endpoint's response schema below contains the complete field list.