Search request logs

GET/v1/logs

Search request logs.

Returns one page of log rows matching every filter given, newest first unless sort_by says otherwise, with the total count. A row covers one request: chat, Claude Messages, embeddings, audio, images, files, batches, Responses API, Decisions and System One calls. Request and response bodies are not in the rows; read one log for them. Text filters match exactly unless noted.

With workspace_id, requires access to that workspace: an organization owner or admin, a member of the workspace, an organization-scoped va_ key with the admin role, or a va_ key scoped to the workspace. Without it, requires an organization owner or admin, or an organization-scoped va_ key with the admin role. Errors: 400 for a date-time that is not ISO 8601 with an offset, or an unknown sort_by.

Authorization

AuthorizationBearer va_
A platform API key (va_...), created under Settings > API keys. Scoped to an organization or one workspace, with an admin or member role. Used by the management API.

Query parameters

providerstring
only requests to this provider, as stored: upper case with underscores, e.g. OPENAI
modelstring
only requests for this model id, as sent
organizationstring
the organization's slug
workspace_idstring<uuid>
only this workspace's requests (UUID)
user_idstring
only requests attributed to this end user (the request's user field)
virtual_key_slugstring
only requests made with this virtual key (vk-...)
config_slugstring
only requests routed through this routing config
guardrail_slugstring
only requests this guardrail applied to (gr-...), alone or with others
guardrail_statusstring
only requests whose combined guardrail result is this: PASS, MONITOR, REDACT, BLOCK or BYPASS
statusstring
success (status code below 400) or error (400 and above); anything else is ignored
cache_statusstring
only requests with this cache result: HIT, SEMANTIC_HIT, MISS, BYPASS, DISABLED or REFRESH
start_timestring
only requests at or after this time (ISO 8601 date-time with offset)
end_timestring
only requests before this time (ISO 8601 date-time with offset)
searchstring
text to find anywhere in the stored request or response body, case-insensitive; only matches logs whose bodies were stored
min_costnumber<double>
only requests that cost at least this much, in USD
max_costnumber<double>
only requests that cost at most this much, in USD
min_duration_msinteger<int64>
only requests that took at least this long, in milliseconds
max_duration_msinteger<int64>
only requests that took at most this long, in milliseconds
min_tokensinteger<int32>
only requests with at least this many total tokens
max_tokensinteger<int32>
only requests with at most this many total tokens
sort_bystring
the field to sort by: created_at (default), provider, model, status_code, duration_ms, prompt_tokens, completion_tokens, total_tokens, cost, cache_status, retry_count, fallback_provider, virtual_key_slug, config_slug, guardrail_slug, guardrail_status, organization, workspace_id, user_id or streaming
sort_dirstring
asc or desc (default)
pageinteger<int32>
the page to return, starting at 1 (default 1)
page_sizeinteger<int32>
rows per page, default 50, at most 100 (a larger value is lowered to 100)
request_idstring
only the request with this request id (the X-Request-ID it ran under)
operation_typestring
only this kind of operation, e.g. embedding, transcription, translation, image_generation, image_edit, file_upload, batch_create, response_create, message_batch, system_one or openai_decision (chat and Claude Messages requests have none)
tool_scan_statusstring
only requests whose tool-result scan had this result: PASS, MONITOR, REDACT or BLOCK; FLAGGED matches any of the last three

Responses

200application/json

totalinteger<int64>
How many logs match the search, over all pages.
pageinteger<int32>
This page's number, starting at 1.
page_sizeinteger<int32>
The most logs a page holds.
dataobject[]
The logs on this page.
Child attributes
- idstring
  The log's id (UUID). Use it to read the log in full.
- request_idstring
  The request id the call ran under: the caller's X-Request-ID, or one the gateway generated. For a Message Batch request, <batch id>/<custom_id>.
- created_atstring<date-time>
  When the request was logged (ISO 8601 date-time).
- providerstring
  The provider the request went to, upper case with underscores, e.g. OPENAI or VERTEX_AI.
- modelstring
  The model id the request asked for.
- status_codeinteger<int32>
  The HTTP status the request ended with. 499 means the caller disconnected during a stream.
- duration_msinteger<int64>
  How long the request took at the gateway, in milliseconds.
- streamingboolean
  Whether the response was streamed.
- organizationstring
  The slug of the organization the request belongs to.
- user_idstring
  The end user the caller attributed the request to (the request's user field, or metadata.user_id on Claude Messages); null when none was sent.
- prompt_tokensinteger<int32>
  Input tokens, cached input included. Null when the provider reported no usage.
- completion_tokensinteger<int32>
  Output tokens. Null when the provider reported no usage.
- total_tokensinteger<int32>
  Input plus output tokens. Null when the provider reported no usage.
- costnumber<double>
  What the request cost, in USD, including the platform markup on a PLATFORM virtual key. 0 on a cache hit; null when the model has no price.
- cost_saved_usdnumber<double>
  What a cache hit would have cost, in USD (cost is 0 on a hit); null on every other request.
- cache_statusstring
  The response cache result: HIT, SEMANTIC_HIT, MISS, BYPASS, DISABLED or REFRESH.
- retry_countinteger<int32>
  How many times the gateway retried the call (see x-retry-count).
- fallback_providerstring
  The fallback provider that served the request when the primary failed; null when no fallback was used.
- virtual_key_slugstring
  The virtual key the request was made with (vk-...); null for direct provider credentials.
- workspace_idstring<uuid>
  The workspace the request belongs to (UUID); null when it has none.
- config_slugstring
  The routing config the request was routed through; null when none was used.
- guardrail_slugstring
  The guardrails that applied to the request, comma-separated (gr-...); null when none did.
- guardrail_statusstring
  The combined guardrail result over the input and output checks: PASS, MONITOR, REDACT, BLOCK or BYPASS; null when no guardrail applied.
- tool_scan_statusstring
  The result of the scan of the tool results the request fed back to the model: PASS, MONITOR, REDACT or BLOCK; null when the request carried no tool results.
- initiatorstring
  MCP_SERVER when the MCP Gateway made this request on behalf of an MCP server's sampling request; null otherwise.
- mcp_server_slugstring
  The MCP server whose sampling request this request answered; null otherwise.
- regionstring
  The AWS region that served the request and holds this log, e.g. us-east-2; null on logs written before it was recorded.
- tool_call_namesstring[]
  The function names the model asked to call, in the order it asked; null or empty when it asked for none.
- tool_result_countinteger<int32>
  How many tool results (tool-role messages) the request carried.
- declared_tool_countinteger<int32>
  How many tools the caller declared on the request.

VectorAxis is built by GK Labs, a member of NVIDIA Inception.