# NousResearch/hermes-agent · VRAM-based resident model cap for local runtime

## Open now

- Nothing open on GitHub.

A bug was identified where the local runtime admitted resident models by count rather than available VRAM, causing silent paging and slowdowns on limited-memory cards. The fixes derive the residency cap from actual VRAM and staged model sizes rather than a simple count ceiling.

## TL;DR

- 3 member(s): 0 open issue(s), 1 closed issue(s), 0 open PR(s), 1 merged PR(s), 1 PR(s) closed without merging.
- A fix was merged to main in PR #117656. Verify it addresses your case.
- **Recommended:** read PR #117656 (and any competing PRs below) before filing. Confirm the merged change covers your symptom.

## Maintainer resolution verdict

- Internal-tracker verdict over this cluster's issue members: `resolved` · deciding PR(s): `pr-117656`. This is distinct from the fix status above. Weigh it yourself.

## How this cluster was grouped

- Link evidence (`STRUCTURAL_SPANNING`): every member is connected by structural evidence: shared changed files, or a closing/superseded reference.
- Pairwise link density `0.67`: the share of member pairs that carry a link, counting a structural link in full and a relevance-only link at 0.36. `1.00` means every pair is linked.

## Closed / duplicate issues

- #116078: \[Bug\]: Managed local runtime admits models by count, not by VRAM — on an 8 GB card the second resident is silently paged (2.6× slower)

## Prior & competing PRs

- **#117656: fix(local-runtime): resident model cap derives from VRAM with models_max as the ceiling (#116078, salvage #116350)**. merged to main · fix tier `verified`
  - maintainer direction: category `merged`
  - files: `hermes_cli/local_runtime/bootstrap.py`, `hermes_cli/local_runtime/presets.py`, `tests/hermes_cli/test_local_residency_admission.py`
- **#116350: fix(local-models): derive the residency cap from the card and the staged models, not the count alone**. closed without merge
  - maintainer direction: category `merged_via`; implication `superseded`; refs `1142a5b3729e`, `#117656`

## Fix status

- Merged to main in PR #117656 (cited above). No release data is tracked. Verify downstream.
- Fix last verified (merged) `2026-09-21`.