---
title: "Get endpoint logs (structured, paginated) [READ]"
method: GET
path: "/v3/endpoint/{namespace}/{name}/logs"
tags: ["Endpoints"]
---

# Get endpoint logs (structured, paginated) [READ]

`GET /v3/endpoint/{namespace}/{name}/logs`

Cursor-based pagination split into two concerns:
- **Time window** (`since` / `until`, RFC 3339, inclusive): immutable filter carried across all paginated requests
  for a given view.
- **Cursor** (`after` / `before`): opaque pagination cursor produced by the API in the `Link` header. Clients MUST
  treat it as opaque.

`order` controls which side of the window `limit` cuts from (`desc` keeps the
newest, `asc` keeps the oldest). Response entries are always returned in
chronological (ascending) order regardless of `order`.

All log entries sharing a boundary millisecond are always included, even if
that exceeds `limit`.

Always returns `application/json`. For a kubectl-style line-per-log rendering,
use `/v3/endpoint/{namespace}/{name}/logs/download`.

## Path parameters

- `namespace` string, required
- `name` string, required

## Query parameters

- `replica` string
- `level` string
- `search` string
- `since` string
- `until` string
- `after` string
- `before` string
- `order` string
- `limit` integer
- `line_max_length` integer

## Response `200`

Log entries returned successfully. Link header contains rel=next, rel=prev and rel=next_sse cursors.

## Other responses

- `501` — Log search not available on this deployment

---

[API](https://skmtc.net/huggingface/apis/hf-inference-endpoints-api.md) · [All operations](https://skmtc.net/huggingface/apis/hf-inference-endpoints-api/llms.txt) · [OpenAPI document](https://skmtc-service-staging.skmtc.workers.dev/v1/apis/huggingface/hf-inference-endpoints-api/versions/52e3f4cdbf86/schema)