Log Workspace¶
A place to collect logs — imported from a pod, uploaded from a bundle, or dragged in — and then read, filter, compare and hand to the AI, without leaving them in a terminal scrollback.
Tier: Free
What it does¶
A pod's logs die with the pod. The Log Workspace is where you put them before
that happens: import the current tail from any pod in a namespace you can see,
upload a file, or drag a .log/.txt in from a support bundle. Files persist
per user, so the evidence you collected at 02:00 is still there in the morning.
Once a file is in the workspace you can read it with line numbers and wrapping, filter it to matching lines, and switch on an errors-only view that highlights the error keywords in place. Two files can be compared side by side, which is the fastest way to answer "what changed between the working replica and the broken one".
Imported logs are REDACTED on the way in, using the same rules the AI paths use — a log tail routinely contains tokens, connection strings and kubeconfig fragments, and a workspace file is stored rather than glanced at.
Anything in the workspace can be handed to the AI for analysis without
copy-pasting, which is the point of collecting it here rather than in a
terminal. What the model receives is redacted, and it is bounded: the first
50,000 characters of each attached file (AGENT_ATTACHMENT_MAX_CHARS), and at
most 10 files per message (AGENT_ATTACHMENT_MAX_FILES). A long log therefore
reaches the model as a prefix — filter it down to the lines that matter before
attaching it.
Ownership¶
Workspace files belong to the person who created them. Another user cannot list, read, download, overwrite or delete them, and a request for someone else's file returns "not found" rather than "forbidden" — a distinction that matters, because a different answer would let anyone probe for other people's filenames. Admins can reach everything, since they can already read the database and someone has to clean up after a departed colleague.
Setup¶
Reach it from the WORKSPACE group in the sidebar. Import from cluster takes the namespace from the namespace selector at the top of the page and lists the pods in it; pick a pod, choose how many lines of tail you want, and import. Upload accepts a file directly, and the file list area is a drop zone.
Uploads are bounded on purpose: 25 MiB per file, and an archive is limited in both member count and total decompressed size. Without those limits the workspace would be the easiest way for any authenticated user to fill the database volume, which is shared with the audit trail.
Troubleshooting¶
If an imported log looks like one long line full of \n and [32m sequences,
it was captured by an earlier release. Re-import the pod, then delete the old
file — a re-import creates a second file rather than overwriting the first. The
stored copy cannot be repaired in place, because those line breaks were never
line breaks — they are the two characters backslash and n, not newlines.
Current releases decode the log properly and strip terminal colour on import.
An import that fails with a container name error means the pod has more than one container; pick the one you want. A pod that has just restarted returns the CURRENT container's log, which may be nearly empty — the previous instance's log is what you usually want after a crash, and that is what alert evidence captures automatically when a rule fires.
Generated from KubeManta 1.2.0 — this page is rendered from the product's own documentation, so it cannot drift from what ships. Manifest built 2026-09-12T02:14:45Z.