Issues

As your knowledge base grows, Mereon flags issues: conflicts, duplicates, gaps, and stale or misclassified knowledge. These endpoints expose them so a tool or dashboard can surface what needs attention. Reading issues needs the issue.read permission, which every role has except external; an external account gets 403, as it does in the app. People raise issues too, and a client can file one for them, or decide one that is already open.

List issues

GET /v1/issues

Most recent first, cursor-paginated. You see an issue you filed, and any other issue whose every topic or item target you can read.

Query parameters

Parameter	Type	Notes
status	string	OPEN, IN_PROGRESS, RESOLVED, or DISMISSED.
type	string	CONFLICT, DUPLICATE, MISCLASSIFICATION, STALE, CONTENT_GAP, STRUCTURE_GAP, OTHER, UPDATE_HELD (an update to a resource held because it would remove too much), or RETIRED_ITEM_REAPPEARED (an item someone retired that a new version of its source says again).
severity	string	CRITICAL, HIGH, MEDIUM, or LOW.
assignedToMe	boolean	Only issues assigned to the connected user, either directly or through a role they currently hold.
targetType	string	TOPIC, ITEM, SPACE, or CATEGORY. Use it with targetId to find the issues on one thing.
targetId	string	The id of the topic, item, space, or category the issue concerns.
search	string	Text to find in the title, the description, or a target’s title. The whole text has to appear, in that order, and case does not matter.
cursor	string	Pagination cursor from a previous response.
limit	number	1–100. Default 50.

Request

curl "https://api.mereon.ai/v1/issues?status=OPEN&severity=HIGH" \
  -H "Authorization: Bearer $MEREON_TOKEN"

Response

{
  "data": [
    {
      "id": "iss_4f9c...",
      "type": "CONFLICT",
      "severity": "HIGH",
      "status": "OPEN",
      "title": "Two documents disagree on the expense approval threshold",
      "description": "The Finance Handbook says $500; the Travel Policy says $750.",
      "assignedToName": "Alex Doe",
      "assignedToRoleName": null,
      "raisedByName": null,
      "trigger": "AI_REVIEW",
      "targets": [
        { "targetType": "TOPIC", "targetId": "top_9f2a...", "title": "Expense approval" }
      ],
      "createdAt": "2026-06-01T09:00:00.000Z",
      "resolvedAt": null
    }
  ],
  "nextCursor": null
}

Fields

Field	Type	Notes
type	string	The kind of issue (see the parameter table).
severity	string	CRITICAL, HIGH, MEDIUM, or LOW.
status	string	OPEN, IN_PROGRESS, RESOLVED, or DISMISSED.
title	string	Short summary of the issue.
description	string | null	Longer explanation, when present.
assignedToName	string | null	Assigned user’s display name, when assigned to a person.
assignedToRoleName	string | null	Assigned role’s name, when assigned to a role.
raisedByName	string | null	Who raised it (manual issues only).
trigger	string	What raised the issue: AUTO_DETECTED (the system’s own detection), AI_REVIEW (an AI knowledge review), REVIEW (a content review), or MANUAL (a person).
targets	object[]	The entities the issue concerns, each with targetType, targetId, and title.
resolvedAt	string | null	When it was resolved, or null if still open.

Get an issue

GET /v1/issues/{id}

One issue in full: the list fields, plus each target’s affected content and how the issue was resolved. That extra detail is what lets a conflict or duplicate be explained rather than only named.

Request

curl "https://api.mereon.ai/v1/issues/iss_3d81" \
  -H "Authorization: Bearer $MEREON_TOKEN"

Response

{
  "id": "iss_3d81...",
  "type": "CONFLICT",
  "severity": "HIGH",
  "status": "RESOLVED",
  "title": "Two different torque values for the same fastener",
  "description": "One manual says 50 Nm, a later bulletin says 65 Nm.",
  "assignedToName": null,
  "assignedToRoleName": "Maintenance Lead",
  "raisedByName": null,
  "trigger": "AUTO_DETECTED",
  "targets": [
    {
      "id": "tgt_81c4...",
      "targetType": "ITEM",
      "targetId": "itm_5c02...",
      "title": "Torque the wheel fasteners to 50 Nm",
      "blockType": "SOP_STEP",
      "body": "Torque the wheel fasteners to 50 Nm in a star pattern.",
      "handled": true
    }
  ],
  "resolution": { "strategy": "supersede" },
  "resolvedAt": "2026-07-30T14:05:00.000Z"
}

Each target carries the list fields (targetType, targetId, title) plus:

Field	Type	Notes
id	string	This target’s own row id, which is distinct from targetId, the id of the thing it points at.
body	string | null	The item’s own text, for an ITEM target. This is the wording the conflict or duplicate is about, which is what makes the issue explainable rather than merely named. null for other target types.
blockType	string | null	The kind of item, for an ITEM target — for example SOP_STEP or DEFINITION. An open set: new kinds can appear without a new API version. null otherwise.
handled	boolean	Whether this particular target has been dealt with. An issue can be resolved for one target and still open against another.
resolution	object | null	How it was resolved, as a strategy name. null while the issue is still open; resolvedAt carries when.

Note

Like the list, this is content-scoped. An issue you filed is returned even when a topic or item target has no id or names something you can no longer read. Otherwise the issue is returned only when every one of its targets is readable. An issue someone else filed that spans a resource you cannot see answers 404 rather than showing you part of it. An item’s text is included only when you can read that item.

Get an issue’s research log

GET /v1/issues/{id}/research-logs

What the AI knowledge review has already looked at for an issue. Each entry is one pass: its status, the searches it ran, whether it decided the knowledge base already covers the topic, and the summary it wrote. Read it before researching an issue again, so you do not repeat what has been checked. Newest first, not paginated.

Request

curl "https://api.mereon.ai/v1/issues/iss_3d81/research-logs" \
  -H "Authorization: Bearer $MEREON_TOKEN"

Response

{
  "data": [
    {
      "id": "log_52e0...",
      "status": "COMPLETE",
      "selfResolved": false,
      "findings": "The bulletin supersedes the manual, but the manual was never retired.",
      "searchQueries": ["wheel fastener torque", "torque bulletin"],
      "createdAt": "2026-07-29T08:12:00.000Z"
    }
  ]
}

Field	Type	Notes
status	string	PENDING, IN_PROGRESS, COMPLETE, or FAILED.
selfResolved	boolean | null	true when the review decided the knowledge base already covers the topic.
findings	string | null	The review’s summary of what it found. null unless you can read the whole organization.
searchQueries	string[] | null	The searches it ran. null unless you can read the whole organization.

The log is visible to whoever can read the issue. An issue the caller cannot read, for the reasons given above, answers 404 here too.

What the review wrote, findings and searchQueries, is the model’s own text from a run that searched all of the organization’s content, so it can quote material a narrower reader cannot open. The access rules return AI free text on an issue only to admins, and this follows them: anyone who does not read the whole organization sees that each pass ran and how it ended, with those two fields null.

File an issue

POST /v1/issues

Files an issue for your organization’s reviewers, as the form on the issues page does: something wrong, missing, out of date, duplicated, conflicting or in the wrong place. It is recorded as raised by the person behind the token, with trigger set to MANUAL.

This needs a token granted the issue:write scope, which is an OAuth path today, and the person behind it must be allowed to file issues in Mereon, which owners, admins and team admins are. Both are checked on every call, so a demotion takes effect on the next one. A member can read issues but not file them, and gets 403.

Request

curl -X POST "https://api.mereon.ai/v1/issues" \
  -H "Authorization: Bearer $MEREON_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "STALE",
    "severity": "HIGH",
    "title": "The lockout steps show the old breaker panel",
    "description": "Step 3 shows the panel that was replaced in March.",
    "targets": [
      { "targetType": "TOPIC", "targetId": "top_9f2a...", "targetTitle": "Lockout/Tagout" }
    ]
  }'

Field	Type	Notes
type	string	CONFLICT, DUPLICATE, MISCLASSIFICATION, STALE, CONTENT_GAP, STRUCTURE_GAP, or OTHER. UPDATE_HELD and RETIRED_ITEM_REAPPEARED are raised by Mereon and cannot be filed.
severity	string	CRITICAL, HIGH, MEDIUM, or LOW.
title	string	A short statement of the problem, up to 500 characters. It cannot be blank.
description	string | null	Optional. What is wrong and what would fix it, up to 10,000 characters.
targets	object[]	What the issue is about: 1 to 20 entries, each with targetType, targetId and targetTitle.
assignedToRoleId	string | null	Optional. A company role to assign it to, from GET /v1/company-roles.
assignedToId	string | null	Optional. A person to assign it to, from GET /v1/members.

Each target names one thing:

Field	Type	Notes
targetType	string	TOPIC, ITEM, SPACE, or CATEGORY. Every target of one issue has the same type.
targetId	string | null	A topic’s id from GET /v1/topics, an item’s from GET /v1/items/{id}, a space’s from GET /v1/spaces, or a category’s slug. Leave it out to name something by title alone, such as a topic your knowledge base is missing.
targetTitle	string	What the issue calls it, as people will read it.
includeHistory	boolean	Optional. File about this topic or item even though it is archived or superseded. Leave it out unless you mean this record rather than what replaced it.

The body is checked as the app checks it. CONFLICT and DUPLICATE need at least two targets, MISCLASSIFICATION cannot name a space or an item that only exists as a step of another, and an id or assignee that is not in your organization is a 400. A field the body does not take, trigger included, is a 400 too, so a misspelt field is refused rather than ignored. A topic or item you can read that is archived or superseded is a 422 unless that target sets includeHistory to true. The error names its status and, when it was corrected, the id of what replaced it.

Response

201 with the issue, in the shape GET /v1/issues/{id} returns, so you do not have to read it back.

{
  "id": "iss_7a10...",
  "type": "STALE",
  "severity": "HIGH",
  "status": "OPEN",
  "title": "The lockout steps show the old breaker panel",
  "description": "Step 3 shows the panel that was replaced in March.",
  "assignedToName": null,
  "assignedToRoleName": null,
  "raisedByName": "Alex Doe",
  "trigger": "MANUAL",
  "targets": [
    {
      "id": "tgt_c3e9...",
      "targetType": "TOPIC",
      "targetId": "top_9f2a...",
      "title": "Lockout/Tagout",
      "blockType": null,
      "body": null,
      "handled": false
    }
  ],
  "resolution": null,
  "createdAt": "2026-10-13T09:00:00.000Z",
  "resolvedAt": null
}

Note

Every topic and item you name by id has to be one you can read through GET /v1/topics/{id} or GET /v1/items/{id}. One you cannot read is a 404, the same answer as an id that does not exist, so filing cannot be used to learn which ids exist. Spaces and categories are not limited by content access, and a target named by title alone names nothing to check.

That works the other way when you read issues back. An issue you filed stays in GET /v1/issues and GET /v1/issues/{id}, including one that names a topic or item by title alone or names one you can no longer read. Someone else who cannot read that target does not see it, unless they can read the whole organization. An item’s text is included only when you can read that item.

Assigning the issue to a person or a company role notifies them, by email or text as their notification settings allow, as it does in the app. The person who filed it is not notified, and an issue left unassigned waits in the issue list.

Filing is limited to 10 calls a minute and 60 an hour per connection.

Decide an issue

These routes dismiss an issue, reopen one, change who it is assigned to, merge a duplicate, choose the correct version of a conflict, or fill a content gap. Each needs a token granted the issue:manage scope, which is an OAuth path today, and the person behind it must be allowed to update issues in Mereon, which owners, admins and team admins are. Both are checked on every call. Filing stays on issue:write.

Each returns 200 with the issue, in the shape GET /v1/issues/{id} returns. Deciding an issue is limited to 10 calls a minute and 40 an hour per connection, and each route has its own budget.

When a decision can retrain people who were trained on the material it replaces, leave retrainingIntent out to store none, which retrains nobody. Send full_retake or changed_material when those people should train again. retrainingDueDate is YYYY-MM-DD, and it belongs with one of those two choices. A dismiss and a category upload have no retraining fields.

Superseding a claim, declaring one version canonical, reclassifying an issue, clearing every issue that blocks one resource, removing a note, and filing a proposal stay on the issues page.

Dismiss an issue

POST /v1/issues/{id}/dismiss

Closes an open issue with your reason. The reason is kept on the issue.

An update held because it would remove too much (UPDATE_HELD) cannot be dismissed while it is held, and the call returns 400. It is answered in the app, by publishing the update or having it read again. An issue asking whether a retired item should come back (RETIRED_ITEM_REAPPEARED) can be dismissed, or answered by restoring the item, which resolves it.

Field	Type	Notes
reason	string	Why it is being dismissed, up to 10,000 characters. It cannot be blank.

Reopen an issue

POST /v1/issues/{id}/reopen

Reopens an issue that was resolved or dismissed. Send no body. An UPDATE_HELD issue cannot be reopened, because its answer cannot be undone, and the call returns 400.

Assign an issue

PUT /v1/issues/{id}/assignment

Replaces who the issue is assigned to. Both fields are required. An id sets that side, and null clears it. Both null clears the assignment. A person or role newly assigned is notified.

Field	Type	Notes
assignedToId	string | null	A person from GET /v1/members, or null.
assignedToRoleId	string | null	A company role from GET /v1/company-roles, or null.

Merge a duplicate

POST /v1/issues/{id}/merge

Keeps one topic a duplicate issue flags and supersedes the other. winnerId and loserId have to be the two topics this issue flags, and two different ones. A duplicate that flags more than two topics is resolved from the issues page, where each one can be included.

Field	Type	Notes
winnerId	string	The topic that remains.
loserId	string	The topic that is superseded.
retrainingIntent	string	Optional. none, full_retake, or changed_material. Leave it out to store none.
retrainingDueDate	string | null	Optional. YYYY-MM-DD, with full_retake or changed_material.

Choose the correct version

POST /v1/issues/{id}/select

Chooses the correct version of a conflict. selectedTargetId is the id of the issue’s target, from GET /v1/issues/{id}, rather than the topic or item behind it.

Field	Type	Notes
selectedTargetId	string	The issue target that is correct.
retrainingIntent	string	Optional. none, full_retake, or changed_material. Leave it out to store none.
retrainingDueDate	string | null	Optional. YYYY-MM-DD, with full_retake or changed_material.

Fill a content gap

POST /v1/issues/{id}/content-gap

Fills a content gap in one of four ways. Send only that strategy’s fields. A field that belongs to another strategy is a 400.

Field	Type	Notes
strategy	string	contentGapUpload, contentGapCategoryUpload, contentGapLink, or contentGapNote.
contentId	string	A resource already in the library, from GET /v1/content, or one you uploaded. Required for an upload.
via	string	Optional, with an upload. CONTENT_GAP_UPLOAD or CONTENT_GAP_RECORD. Leave it out for a file.
linkedType	string	TOPIC. Required for a link.
linkedId	string	The topic that covers the gap, from GET /v1/topics. Required for a link.
title	string	The note’s title, up to 255 characters. Required for a note.
body	string	The note. Required for a note.
override	boolean	Optional, with a link or a note. Records the resolution after the coverage check declined. Leave it out unless that is what you want.
retrainingIntent	string	Optional, except on a category upload, which has none. Leave it out to store none.
retrainingDueDate	string | null	Optional. YYYY-MM-DD, with full_retake or changed_material.

An upload names a resource you can read, or one you uploaded that has no published version yet. A published resource has to be one you can read. Anything else is a 404, the same answer as an id that does not exist. A link names a topic you can read. A note is the text itself.

An upload whose processing has finished with no knowledge items is a 400. Processing that is still running can fill the gap, and so can a resource that already has knowledge items. A category upload does not wait on those items.