Issues

• perf(curation): super-linear cluster pass and storage-pressure attribution in the scale driver

  area:ingestenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#337 · Filter by author davidamacey opened on Oct 17, 2026 · v0.7.0
  1

• docs(screenshots): make the capture script self-seeding (17 states need fixtures)

  area:docsdocumentationepic:maintenance
  davidamacey/OpenProcessor#335 · Filter by author davidamacey opened on Oct 16, 2026 · v0.7.0
  1

• feat(curation): iterative cluster-sampled VLM auto-labeling driver over assist runs

  area:curationenhancementepic:vlm-labeling
  davidamacey/OpenProcessor#286 · Filter by author davidamacey opened on Oct 14, 2026 · v0.7.0
  1

• perf(triton): stage 2 follow-ups - set B photos, minimal/standard profiles, two-card sweeps, CUDA graphs, engine stamps

  area:ingestenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#268 · Filter by author davidamacey opened on Oct 13, 2026 · v0.7.0
  3

• perf(embed): embed reprocess pass stops at 58% GPU (14 img/s vs about 24 ceiling); crop vector mismatch decision

  area:ingestenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#259 · Filter by author davidamacey opened on Oct 12, 2026 · v0.7.0
  5

• perf(serving): ortloom-serve vs Triton head-to-head on the ingest-path models

  area:gpu-servingenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#231 · Filter by author davidamacey opened on Oct 10, 2026 · v0.7.0
  1

• perf(ocr,faces): uint8 OCR input, width-bucketed recognition and face alignment on resident GPU frames

  area:gpu-servingarea:ingestenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#220 · Filter by author davidamacey opened on Oct 10, 2026 · v0.7.0

• perf(curation): work crops at model size for the VLM and region consumers

  area:ingestenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#219 · Filter by author davidamacey opened on Oct 10, 2026 · v0.7.0

• fix(trainer): use MLflow model version aliases instead of deprecated registry stages

  enhancementepic:maintenance
  davidamacey/OpenProcessor#154 · Filter by author davidamacey opened on Oct 9, 2026 · Backlog

• security(vlm): pin the VLM HTTP connection to the validated IP (close the DNS re-resolve window)

  area:curationenhancementepic:vlm-labeling
  davidamacey/OpenProcessor#101 · Filter by author davidamacey opened on Oct 3, 2026 · Backlog

• docs(example): car-wheel teacher-student loop on public COCO (SAM 3 labels, train a wheel detector, compare throughput)

  area:curationdocumentationenhancementepic:any-domainperformance
  davidamacey/OpenProcessor#84 · Filter by author davidamacey opened on Oct 3, 2026 · v0.7.0
  1

• feat(api): request-driven processing plans (run only the requested models and branches)

  area:curationarea:ingestenhancementepic:any-domain
  davidamacey/OpenProcessor#67 · Filter by author davidamacey opened on Oct 3, 2026 · v0.7.0
  3

• roadmap: deferred any-domain features (embedding swap, ingest profile, second segmenter, model-family registry, auth)

  area:curationenhancementepic:any-domain
  davidamacey/OpenProcessor#65 · Filter by author davidamacey opened on Oct 3, 2026 · Backlog

• perf(bench): public benchmark ladders - Open Images category subsets, full COCO, 1M images, 20 MP at 100k

  enhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#45 · Filter by author davidamacey opened on Oct 2, 2026 · v0.7.0
  9

• feat(models): TensorRT engine build target and instance guidance for 12 GB GPUs

  area:gpu-servingenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#42 · Filter by author davidamacey opened on Oct 2, 2026 · v0.7.0

• build(docker): slim the container images (Plan B: runtime image plus export image)

  enhancementepic:maintenance
  davidamacey/OpenProcessor#41 · Filter by author davidamacey opened on Oct 2, 2026 · v0.7.0
  1

• perf(pipeline): epic - decode once and keep frames on the GPU across models, baseline first

  area:gpu-servingenhancementepic:perf-gpuperformance
  davidamacey/OpenProcessor#40 · Filter by author davidamacey opened on Oct 2, 2026 · v0.7.0
  16

• feat(video): video analytics subsystem with DeepStream 9.0 and Triton (RTSP, batch segments, tracking, search)

  enhancementepic:video
  davidamacey/OpenProcessor#8 · Filter by author davidamacey opened on Jul 6, 2026 · Backlog
  4