hermes-dreaming plugin
hermes-dreaming stages Hermes memory and skill updates for review and writes them to live state only after approval.
- 96 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- After `apply` the plugin writes to Hermes live memory, user notes and skills, though backups and rollback exist.
Install
- Works in
Manual install
hermes plugins install asimons81/hermes-dreaming --enableAfter install `hermes dreaming review --help` is available. Update with `hermes dreaming update`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- After `apply` the plugin writes to Hermes live memory, user notes and skills, though backups and rollback exist.
This is third-party code. Review the repository files before installing.
What it does
A plugin for Hermes Agent with a dedicated `dreaming` command reads explicit sources or recent sessions and stages proposed changes in an artifact directory: `manifest.json`, `REPORT.md`, `proposals.jsonl`, `audit.jsonl`. Proposals concern memory, user notes, skills and facts. `approve` and `reject` change only artifact metadata, while `apply` writes approved items to live files with backups and score gating. `revert` restores files from the backups and `dreaming install-cron` sets up a nightly review-only run. The built-in provider finds `DREAM:` lines in sources. For model-based analysis there are OpenAI-compatible and Ollama providers; on a malformed reply they refuse to write.
Who it is for. Hermes Agent owners whose memory and skills grow from sessions and who want every change reviewed before it is written.
Good fit when
- After a week of work the agent's memory holds contradictory entries to sort through
- Skill edits are drafted automatically but must not be applied without viewing the diff
- A nightly run should leave a report by morning without changing anything
- An apply produced too much and files must be restored from backups
Not a fit when
- The agent is not Hermes: the plugin targets its memory, skills and session store
- You want fully automatic self-learning without approval: writing happens only after `apply`
- Memory is small and static and proposals would only complicate things
Example request
Stage memory proposals from the last five sessions, show the summary and apply nothing without my approval.Limitations
The plugin targets Hermes Agent and does not work with other agents. Writing to live memory, user notes and skills happens after `apply`, so set `--backup-root` and try `--dry-run` first. The OpenAI-compatible and Ollama providers send source fragments to an external or local model. Installing from GitHub needs Hermes with the `hermes plugins` command. The last commit dates from August 2026.
How to disable. Disable the plugin with `hermes plugins` and delete its install folder. Restore live memory files already changed by `apply` with `dreaming revert ... --yes`.
FAQ
How do I install hermes-dreaming?
Run `hermes plugins install asimons81/hermes-dreaming --enable`. For development, `python -m pip install -e .[dev]` from a cloned folder works.
Does the plugin change memory by itself?
No, `create` and `review` only prepare an artifact and `approve` and `reject` change its metadata. Only approved items applied with `apply` reach live files.
How do I undo an apply?
Run `dreaming revert <artifact> --live-root ... --backup-root ... --yes`: files return from the recorded backups.
Is an API key required?
No, the built-in `offline-marker` provider finds `DREAM:` lines without a model. The OpenAI-compatible and Ollama providers are for model-based analysis.
How do I schedule a nightly review?
Run `dreaming install-cron --schedule "0 3 * * *"`: the job only stages an artifact and applies nothing.
Discussion
No questions or reviews yet.
Related
PluginsMemory provider for Hermes Agent: duplicates are merged, contradictions surface and every recalled memory carries a reason.
A pack of practical skills for Hermes Agent: notes, trips, code security, browser automation and interface themes.
A pack of 19 plugins for Hermes Desktop: provider quotas, an iteration meter, a memory review queue, session management and RSS.
External memory for Hermes, Claude Code, Codex and Cursor: session archive, recall layers and relevant context injected back
Official Hermes Agent plugins: Home Assistant, Spotify, Blender, Snyk, Langfuse, sandboxes and importers for other agents
Agent Skills by moonlight-lupin
agent-skills
A library of skills and plugins for Hermes Agent covering research, media, scraping, documents, DevOps and agent upkeep.
README in short
The repository works standalone and as a Hermes plugin. An example runs without keys: `examples/quickstart/` needs no model, and a run ledger plus a `DREAMS.md` diary record every decision.