omp packages

omp-archive0.1.0

Export a session's entire history as a single archive: the main transcript, every subagent session verbatim, and background-job outputs

fran2789 Published 1 hour ago MIT ~0 downloads/mo

Built for omp

The author tagged this package omp-plugin.

Did it work for you?

Result

One report per person. A new report replaces your old one.

Reports 0

No reports yet.

Readme

omp-archive

Export the entire history of an omp session as one zip: the main transcript, every subagent session (recursively, verbatim), all untruncated tool / background-job output, live background-job state, and a standalone HTML viewer.

/export embeds subagent entries in a viewer HTML and /dump covers the live main conversation. Neither ships a subagent's own transcript file, its sibling tool-output logs, or background-job state. This plugin archives all of it.

Install

# npm
omp plugin install omp-archive

# local checkout / development (symlinked, edits are live)
omp plugin link /home/asdf/projects/omp-archive

# git
omp plugin install github:alosslessdev/omp-archive-main

# marketplace (this repo ships .omp-plugin/marketplace.json)
/marketplace add <user>/omp-archive
/marketplace install archive@omp-archive

# one-off
omp --extension /home/asdf/projects/omp-archive

Usage

/archive                      # -> ./omp-archive-<session>.zip
/archive /tmp/session.zip
/archive --no-artifacts       # skip the artifact/log tree (smaller zip)
/archive --no-viewer          # skip viewer.html

The same operation is exposed as the export_session_archive tool so the agent can archive a session on request.

What's inside

Path Contents
manifest.json Index of everything below: counts, sizes, subagent list, omissions
session.jsonl The main session transcript, byte-for-byte
agents.json Live per-agent state captured while the session ran (model, system prompt, tools)
jobs.json Background/async jobs: completed (recovered from every transcript) plus live (running/recent at archive time)
artifacts/ Every file under the session artifact directory, byte-for-byte
viewer.html Standalone HTML viewer with all subagent sessions embedded
README.md Human-readable index of the archive

artifacts/ is the complete record: each subagent's own <AgentId>.jsonl (nested children under <AgentId>/), the *.bash.log / *.eval.log / *.read.log files holding the full, untruncated tool and background-job output that the transcript truncates, and *.jsonl.tombstone kill markers.

Limits

  • Background jobs. jobs.json lists two sources: completed — every settled async job recovered from the async-result entries of the main and subagent transcripts (label, status, duration, output handle) — and live, the running/recent jobs observed while the extension was loaded. A job's full output is in its owning transcript plus the matching artifacts/*.log.
  • Subagent system prompts / tool lists are not persisted by the host, so they appear in agents.json only for sessions this process observed. Each subagent's full conversation is always present, verbatim, under artifacts/.
  • The archive is not redacted: it can contain secrets, raw context, and images. it doesn't sanitize or redact any private or personally identifyable information, that's left to the LLM so you might want to tell it to do that.
  • Paths with spaces are not supported in /archive arguments (same parsing as /export).