project-nomad/admin
Chris Sherwood ba661a9da1 fix(RAG): pace continuation batches when embedding is CPU-only
Stacks on top of the multi-batch ZIM ingestion fix. After that fix,
multi-batch ZIM ingestion completes correctly — but on installs where
Ollama runs the embedding model on CPU (currently every AMD ROCm
install, since Ollama's ROCm build doesn't accelerate nomic-bert),
the now-correct sustained 100% CPU saturation across all cores can
starve other services hard enough to take the box down. Confirmed
on a Threadripper 3960X + RX 6800 NOMAD: a wikipedia-class ZIM
ingestion pegged 48 threads cleanly enough that sshd lost
banner-exchange responsiveness and the box ultimately required a
power-cycle.

NVIDIA installs aren't affected — nomic-embed-text:v1.5 runs at
100% GPU on RTX 5060 (verified via `ollama ps`).

Detect placement at runtime, pace only when needed:

1. OllamaService.isEmbeddingGpuAccelerated() — queries /api/ps and
   returns true if any loaded embedding model reports size_vram > 0.
   Fails closed (returns false) if /api/ps is unreachable or no embed
   model is loaded yet — over-pacing is safer than crashing.

2. EmbedFileJob.handle() — between batches (hasMoreBatches: true
   branch), check placement and `await setTimeout(CPU_BATCH_DELAY_MS)`
   when CPU-only. CPU_BATCH_DELAY_MS = 1000 (1s) — enough to give the
   OS scheduler a window for sshd/disk-collector/etc., small enough
   that total ingestion time isn't meaningfully affected (each batch
   is ~60-90s of work).

GPU-accelerated installs see zero behavior change.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-13 13:52:02 -07:00
..
app fix(RAG): pace continuation batches when embedding is CPU-only 2026-05-13 13:52:02 -07:00
bin feat: curated content system overhaul 2026-02-11 15:44:46 -08:00
commands feat(Maps): regional map downloads via go-pmtiles extract (#780) 2026-05-03 13:47:53 -07:00
config fix: cache docker list requests, aiAssistantName fetching, and ensure inertia used properly 2026-04-03 14:26:50 -07:00
constants feat(Maps): regional map downloads via go-pmtiles extract (#780) 2026-05-03 13:47:53 -07:00
database feat(Content): custom ZIM library sources with pre-seeded mirrors (#593) 2026-05-04 11:30:59 -07:00
docs docs: update release notes 2026-05-04 19:30:06 +00:00
inertia fix(Maps): render notes in marker popup when populated 2026-05-05 10:38:56 -07:00
providers fix(System): self-heal stale updateAvailable flag after sidecar-driven update (#825) 2026-05-04 11:54:56 -07:00
public feat: switch all PNG images to WEBP (#575) 2026-04-03 14:26:50 -07:00
resources feat(Maps): regional map downloads via go-pmtiles extract (#780) 2026-05-03 13:47:53 -07:00
start feat(Content): custom ZIM library sources with pre-seeded mirrors (#593) 2026-05-04 11:30:59 -07:00
tests fix(UI): improve global map banner display logic (#702) 2026-04-28 20:55:25 -07:00
types fix(AI): vendor-aware AMD HSA override + benchmark discrete-GPU detection 2026-05-05 12:11:56 -07:00
util feat: display model download progress 2026-02-06 16:22:23 -08:00
views feat: initial commit 2025-06-29 15:51:08 -07:00
.editorconfig feat: initial commit 2025-06-29 15:51:08 -07:00
.env.example feat: Add Windows Docker Desktop support for local development 2026-01-19 10:29:24 -08:00
ace.js feat: initial commit 2025-06-29 15:51:08 -07:00
adonisrc.ts fix(System): self-heal stale updateAvailable flag after sidecar-driven update (#825) 2026-05-04 11:54:56 -07:00
eslint.config.js feat: openwebui+ollama and zim management 2025-07-09 09:08:21 -07:00
package-lock.json build(deps): bump picomatch in /admin 2026-05-04 10:50:13 -07:00
package.json chore(deps): pin all deps to exact versions 2026-05-04 17:45:18 +00:00
tailwind.config.ts feat: initial commit 2025-06-29 15:51:08 -07:00
tsconfig.json feat: initial commit 2025-06-29 15:51:08 -07:00
vite.config.ts fix(Maps): ensure proper parsing of hostnames (#640) 2026-04-03 14:26:50 -07:00