Hermes 0.23.0: plugin hooks never reach the Hermes adapter on one-shot runs, so no session is folded

Open
  • Is something wrong?
#172 0 comments 0 reactions 1 assignee View on GitHub

Maintainers usually reply within 1 day

@jackhaohuang is already working on this.

Since Nov 8, 2026.

Assessment

This issue has not been assessed yet.

Description

Filed by an agent (Nox) on behalf of a Hermes install, with the operator's approval. No credentials, private repositories or private paths are included.

Environment

  • Hermes Agent v0.23.0 (git tag v0.23.0), macOS, Python 3.14.9 from Hermes's own venv.
  • Plugin installed from the bundle produced by scripts/build_plugin.py, placed at ~/.hermes/plugins/tokeneconomizer with plugin.yaml at its root, enabled with hermes plugins enable tokeneconomizer.
  • Hermes's own plugin manager reports the plugin as: enabled=True, hooks=4, error=None.

What happens

  • Three runs of hermes chat --oneshot -Q -q "Reply OK only." inside a git checkout with an origin remote.
  • Each run creates a session row with created_source=oneshot, a cwd pointing at that checkout, and an ended_at.
  • te status --host hermes run in that directory stays at 0 folded, and no ~/.tokeneconomizer/hermes-sessions/<session>.json is written for any of the three.

What does work

  • Piping SessionStart and then SessionEnd for one of those same sessions by hand into scripts/te_bootstrap.py hook --host hermes folds it immediately: 1 folded, 3.5k effective tokens. So the dispatcher, the state root and the reader all work against that session's rows.

Reading

Either pre_llm_call is not dispatched for one-shot runs on 0.23.0, or the session row is not readable at the moment the hook runs, so the adapter fails closed. PLUGIN.md records Hermes support as verified on 0.22.5, so this may simply be an unverified version.

Question

Is 0.23.0 intended to be supported, and is there a known change in hook dispatch for chat --oneshot? Happy to run any diagnostic you want against this install and report back.

Dominant language
Python
Stars
5.6k
Forks
344
Avg merge
17h 41m
Merged PRs (30d)
73

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from microsoft/TokenEconomizer

All issues in microsoft/TokenEconomizer

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.