opencv/opencv · GitHub event streams · Factory

01 · Data

Dataset versions

A dataset version is a capture of the source as of a date: its event spine and the files training reads.

1 dataset version

opencv/opencv opencv__opencv @ asof-2026-06-12

pooled8 member; recorded where it is (Modal, rezendi, vei-oss-data/opencv__opencv); see REPORT.md

llm_report · REPORT.md

opencv__opencv@asof-2026-06-12 — data stage report

Recorded 2026-10-13 as part of the OSS line's migration to the factory. The data already existed and nothing was rematerialized.

Inputs, by identity. GitHub's event API for opencv/opencv, materialized into the OSS canonical spine. Pooled8 member. The raw captures stay external, on the archive drive and Modal.

Output, by identity. canonical_events.jsonl, 60,151 events from 2023-06-29T04:01:24Z to 2026-06-12T17:52:47Z. The spine is recorded where it already is (Modal, rezendi, vei-oss-data/opencv__opencv) and not copied into the registry.

Drift. Ingest-time check (drift.json, drift.txt): 60,151 events over 37 periods | structural types excluded: none; 1 verb flags, 2 volume flags. The OSS runs split each stream at its own time quantile, so no single calendar split applies across the pooled streams, and no split-time check was run here. Its absence is a caveat on any result.

Receipts (2)

drift.json

{
  "era": {
    "n_periods": 37,
    "n_events": 60151,
    "structural_types": [],
    "verb_flags": [
      {
        "period": "2026-06",
        "verb": "issue_comment",
        "kind": "dropout",
        "count": 25,
        "trailing_median": 142.5,
        "ratio": 0.17543859649122806
      }
    ],
    "volume_flags": [
      {
        "period": "2025-04",
        "n_events": 1033.0,
        "trailing_median": 1601.5,
        "ratio": 0.6450202934748673
      },
      {
        "period": "2026-05",
        "n_events": 2618.0,
        "trailing_median": 1288.5,
        "ratio": 2.0318199456732633
      }
    ],
    "trailing_periods": 6,
    "spike_ratio": 4.0
  },
  "split": null
}

drift.txt

[drift] 60,151 events over 37 periods | structural types excluded: none
[drift] ERA SUSPECT 2026-06: 'issue_comment' dropout 25 vs trailing median 142.5 (0.2x)
[drift] VOLUME 2025-04: 1,033 events vs trailing median 1,602 (0.65x)
[drift] VOLUME 2026-05: 2,618 events vs trailing median 1,288 (2.03x)

4 files, 51.8 MB

Event spine · 1 file51.8 MB
canonical_events.jsonl4913676336d951.8 MB
Receipt · 2 files999 B
drift.json428a50d4d975699 B
drift.txt697a4c94d29b300 B
Phase_report · 1 file968 B
REPORT.md930e3d5a1814968 B

Events: 60,151 in 1 stream
Span: 29 Jun 2023 → 12 Jun 2026
Source: GitHub events API
Published: 13 Oct 2026

0 modelsnothing trained over it yet

02 · Features

Featurization

An Atlas recipe declares how events become model inputs.

No feature recipe has been recorded.

An Atlas recipe appears here once a run that names it is published, or once the recipe itself is registered.

03 · Training

Recipes and runs

A Praxis recipe fixes the model, the objective and the split.

No run has been published.

A run is one execution of a training recipe over named dataset versions. The first one born inside the platform will appear here with its launch.

04 · Evaluation

How it compares to other methods

The models of record and their headline evaluations. Every result includes the ablations and earlier variants.

No results have been published for this project.

A result is a cell: the config, the target at a horizon, the arm and head, the split, the numbers and their provenance.

Needed: an eval config and an eval run of a published model

05 · Serving

Ask the model

If the query service is cold, it will need a few minutes to warm up.

Serving

idle

Nothing is serving yet.

A published run with a report or a query surface answers here.