Nextcloud version: 35.0.2.2
Context Agent version: 2.11.0
Affected component: Task Processing API / Context Agent
Environment: Nextcloud running in Docker behind Traefik
Severity: Medium – AI requests intermittently fail during processing.
1. Problem Description
When using Nextcloud Chat with AI, the Context Agent intermittently failed to schedule tasks through the Nextcloud Task Processing API.
The Context Agent logs showed the following error:
niquests.exceptions.HTTPError:
429 Client Error: Too Many Requests
POST https://my.domain.ltd/ocs/v1.php/taskprocessing/schedule?format=json
nc_py_api._exceptions.NextcloudException:
[429] Too Many Requests
The error was generated by the Context Agent while attempting to schedule a task.
The issue occurred repeatedly during AI request processing and prevented the expected response from being returned.
2. Environment and Diagnostic Findings
The following environment was verified:
-
Nextcloud 35.0.2.2
-
Context Agent 2.11.0
-
Context Chat and Context Chat Backend installed
-
Nextcloud deployed using Docker
-
Traefik used as reverse proxy
The following checks were performed:
-
Checked Traefik configuration for rate-limit middleware.
-
Checked Traefik logs for HTTP 429 errors.
-
Inspected the Nextcloud system configuration.
-
Checked Task Processing worker processes and task statistics.
-
Inspected the Nextcloud Task Processing API controller source code.
No evidence of a Traefik rate-limit middleware was found.
Some Context Agent tasks were successfully processed by the Task Processing worker, indicating that the worker was operational.
3. Root Cause
Inspection of the following Nextcloud source file:
core/Controller/TaskProcessingApiController.php
revealed the following rate-limit annotation on the standard task scheduling endpoint:
#[UserRateLimit(limit: 20, period: 120)]
This configures a limit of 20 requests per 120 seconds.
The affected endpoint was:
POST /ocs/v1.php/taskprocessing/schedule
The Context Agent generated enough scheduling requests to exceed this limit, resulting in HTTP 429 responses.
This rate limit appears insufficient for the request pattern generated by the Context Agent during AI processing.
4. Applied Workaround / Fix
The rate limit was increased from 20 to 100 requests per 120 seconds.
Original code:
#[UserRateLimit(limit: 20, period: 120)]
Modified code:
#[UserRateLimit(limit: 100, period: 120)]
The Nextcloud container was restarted after the modification.
5. Result
After increasing the rate limit and restarting the Nextcloud container:
-
Chat with AI started returning responses successfully.
-
The previously observed HTTP 429 error was no longer preventing AI requests from completing.
-
The workaround resolved the reported problem in this environment.
Important: The current workaround modifies a Nextcloud Core source file directly. This modification may be overwritten by future Nextcloud updates. A supported configuration option or an upstream code change would be preferable.
7. Conclusion
The problem was traced to the rate limit configured on the Nextcloud Task Processing API scheduling endpoint.
Increasing the limit from 20 to 100 requests per 120 seconds resolved the issue in the tested environment.
The recommended permanent solution is to make the rate limit configurable or adjust the default limit to accommodate the scheduling requirements of Context Agent and other AI applications.
Thank you for investigating this issue.
