Dataset Viewer
Auto-converted to Parquet Duplicate
attribute_id
int64
1
58.3k
⌀
file_id
int64
1
1.39k
⌀
key_path
large_stringlengths
2
91
⌀
value
large_stringlengths
0
28.5k
⌀
value_type
large_stringclasses
7 values
1
1
name
Backport to release branch
str
2
1
emoji
🔀
str
3
1
description
Cherry-picks commits from master onto a target release branch, resolves any cherry-pick conflicts, and opens a Jira-prefixed PR.
str
4
1
on.workflow_dispatch.inputs.target_branch.description
Branch to backport onto (e.g. release/25.N, release/25.4, release/23.7)
str
5
1
on.workflow_dispatch.inputs.target_branch.type
string
str
6
1
on.workflow_dispatch.inputs.target_branch.required
True
bool
7
1
on.workflow_dispatch.inputs.commits.description
Commit SHAs to cherry-pick, oldest first (space/comma/newline separated, full or short)
str
8
1
on.workflow_dispatch.inputs.commits.type
string
str
9
1
on.workflow_dispatch.inputs.commits.required
True
bool
10
1
on.workflow_dispatch.inputs.jira_ticket.description
Jira ticket ID to prefix the PR title with (e.g. ACS-1234)
str
11
1
on.workflow_dispatch.inputs.jira_ticket.type
string
str
12
1
on.workflow_dispatch.inputs.jira_ticket.required
True
bool
13
1
permissions.contents
read
str
14
1
permissions.pull-requests
read
str
15
1
permissions.issues
read
str
16
1
strict
True
bool
17
1
timeout-minutes
30
int
18
1
tools.bash[0]
git:*
str
19
1
tools.edit
null
20
1
tools.github.mode
gh-proxy
str
21
1
tools.github.toolsets[0]
default
str
22
1
safe-outputs.create-pull-request.max
1
int
23
1
safe-outputs.noop
null
24
2
name
Release Highlights
str
25
2
description
Generate AI-powered release highlights for a release
str
26
2
on.workflow_dispatch.inputs.version.description
Release version tag (e.g. v1.0.0)
str
27
2
on.workflow_dispatch.inputs.version.required
True
bool
28
2
on.workflow_dispatch.inputs.version.type
string
str
29
2
permissions.contents
read
str
30
2
permissions.pull-requests
read
str
31
2
engine
copilot
str
32
2
timeout-minutes
10
int
33
2
network.allowed[0]
defaults
str
34
3
on.issues.types[0]
opened
str
35
3
on.issue_comment.types[0]
created
str
36
3
on.roles
all
str
37
3
if
github.event_name == 'issues' || (github.event_name == 'issue_comment' && github.event.issue.pull_request == null && !endsWith(github.event.comment.user.login, '[bot]') && ( ((github.event.comment.body == '/triage' || startsWith(github.event.comment.body, '/triage ')) && contains(fromJSON('["OWNER","MEMBER","...
str
38
3
engine
copilot
str
39
3
model
auto
str
40
3
environment
issue-triage
str
41
3
permissions.contents
read
str
42
3
permissions.issues
read
str
43
3
permissions.pull-requests
read
str
44
3
tools.bash[0]
cat
str
45
3
tools.bash[1]
find
str
46
3
tools.bash[2]
grep
str
47
3
tools.github.min-integrity
none
str
48
3
safe-outputs.add-comment.max
1
int
49
3
safe-outputs.add-comment.hide-older-comments
True
bool
50
3
safe-outputs.add-labels.allowed[0]
Auto-Triage: Waiting for Author
str
51
3
safe-outputs.add-labels.max
1
int
52
3
safe-outputs.remove-labels.allowed[0]
Auto-Triage: Waiting for Author
str
53
3
safe-outputs.remove-labels.max
1
int
54
3
safe-outputs.noop.report-as-issue
False
bool
55
4
description
A testing-focused repository assistant that runs daily to improve test quality and coverage. Can also be triggered on-demand via '/test-assist <instructions>' to perform specific tasks. - Discovers and validates build, test, and coverage commands for the repository - Identifies testing gaps and high-value test opportun...
str
56
4
on.schedule
daily
str
57
4
on.workflow_dispatch
null
58
4
on.slash_command.name
test-assist
str
59
4
on.reaction
eyes
str
60
4
timeout-minutes
30
int
61
4
permissions
read-all
str
62
4
network.allowed[0]
defaults
str
63
4
network.allowed[1]
dotnet
str
64
4
network.allowed[2]
node
str
65
4
network.allowed[3]
python
str
66
4
network.allowed[4]
rust
str
67
4
network.allowed[5]
java
str
68
4
safe-outputs.add-comment.max
10
int
69
4
safe-outputs.add-comment.target
*
str
70
4
safe-outputs.add-comment.hide-older-comments
True
bool
71
4
safe-outputs.create-pull-request.draft
True
bool
72
4
safe-outputs.create-pull-request.title-prefix
[Test Improver]
str
73
4
safe-outputs.create-pull-request.labels[0]
automation
str
74
4
safe-outputs.create-pull-request.labels[1]
testing
str
75
4
safe-outputs.create-pull-request.max
4
int
76
4
safe-outputs.create-pull-request.protected-files
fallback-to-issue
str
77
4
safe-outputs.push-to-pull-request-branch.target
*
str
78
4
safe-outputs.push-to-pull-request-branch.title-prefix
[Test Improver]
str
79
4
safe-outputs.push-to-pull-request-branch.max
4
int
80
4
safe-outputs.create-issue.title-prefix
[Test Improver]
str
81
4
safe-outputs.create-issue.labels[0]
automation
str
82
4
safe-outputs.create-issue.labels[1]
testing
str
83
4
safe-outputs.create-issue.max
4
int
84
4
safe-outputs.update-issue.target
*
str
85
4
safe-outputs.update-issue.title-prefix
[Test Improver]
str
86
4
safe-outputs.update-issue.max
1
int
87
4
tools.web-fetch
null
88
4
tools.bash
True
bool
89
4
tools.github.toolsets[0]
all
str
90
4
tools.repo-memory
True
bool
91
4
source
githubnext/agentics/workflows/daily-test-improver.md@346204513ecfa08b81566450d7d599556807389f
str
92
4
engine
claude
str
93
5
name
Build Failure Analysis
str
94
5
description
Runs `./build.sh --binaryLog` on every PR; when the build fails, delegates to the `build-failure-analyst` agent (which queries the binlog live via the containerized `binlog-mcp` MCP server) to identify root causes, post a PR comment summarizing them, and attach inline `suggestion` blocks tied to the diff.
str
95
5
on.pull_request.types[0]
opened
str
96
5
on.pull_request.types[1]
synchronize
str
97
5
on.pull_request.types[2]
reopened
str
98
5
on.pull_request.branches[0]
main
str
99
5
on.pull_request.branches[1]
rel/*
str
100
5
on.pull_request.forks
[]
list
End of preview. Expand in Data Studio

GH-AW Workflows Dataset

Dataset relacional con repositorios de GitHub que usan GitHub Agentic Workflows (GH-AW), y el contenido estructurado de sus archivos de workflow: el .md fuente (frontmatter YAML + body Markdown) y su .lock.yml/.lock.yaml compilado.

Generado con Miner, una CLI en Python que identifica repositorios que usan GH-AW y extrae sus pares .md + .lock en un formato analítico.

  • Repositorios incluidos: 338
  • Archivos de workflow incluidos: 1390 (cada uno con su .lock correspondiente)
  • Fecha de generación: 2026-09-12

Estructura del dataset

El dataset está compuesto por 4 tablas Parquet relacionadas:

Tabla Filas ≈ Descripción
repositories.parquet 1 por repo Repositorios de GitHub que usan GH-AW.
workflow_files.parquet 1 por archivo .md Frontmatter completo (como JSON) + body Markdown de cada workflow fuente.
workflow_locks.parquet 1 por archivo .md (relación 1:1) El .lock.yml/.lock.yaml compilado correspondiente, como texto plano (raw_content).
frontmatter_attributes.parquet N por archivo .md Atributos del frontmatter aplanados como pares clave-valor (ej. permissions.contents, on.schedule[0].cron) — el frontmatter de GH-AW varía entre workflows, así que no se fuerzan columnas fijas.

Esquema entidad-relación completo y diccionario de datos (tipos, PKs, FKs) disponibles en el repositorio de Miner: docs/er-diagram.md y docs/data-dictionary.md.

Cómo cargar el dataset

from datasets import load_dataset

files = load_dataset("EstebanCQ/gh-aw-workflows-dataset", data_files="workflow_files.parquet")
attrs = load_dataset("EstebanCQ/gh-aw-workflows-dataset", data_files="frontmatter_attributes.parquet")

O directamente con pandas:

import pandas as pd
from huggingface_hub import hf_hub_download

path = hf_hub_download("EstebanCQ/gh-aw-workflows-dataset", "workflow_files.parquet", repo_type="dataset")
files = pd.read_parquet(path)

Cómo se generó

  1. Se identificaron repositorios candidatos y se filtraron los que usan GH-AW (al menos un par <nombre>.md + <nombre>.lock.yml en .github/workflows/).
  2. Para cada repositorio identificado, se descargó cada par .md + .lock de workflow; el .md se separó en frontmatter YAML y body Markdown, el .lock se conservó como texto plano. Un par solo se incluye si ambos archivos se descargaron con éxito, para garantizar la relación 1:1 entre workflow_files y workflow_locks.
  3. El frontmatter del .md se aplanó en pares clave-valor (tabla frontmatter_attributes) y se conservó completo como JSON en workflow_files.raw_frontmatter, para no perder ningún campo.

Limitaciones

  • Solo incluye repositorios públicos accesibles con un token de GitHub estándar al momento de la generación.
  • Es una instantánea (snapshot): los workflows pueden cambiar o eliminarse después de la fecha de generación indicada arriba.
  • Todo archivo de workflow_files tiene exactamente un .lock en workflow_locks (relación 1:1): solo se incluyen pares .md + .lock.yml/.lock.yaml confirmados (la definición de "usa GH-AW" de este proyecto), y ambos deben haberse descargado con éxito.

Licencia

MIT — ten en cuenta que el contenido de cada workflow (el body Markdown y el frontmatter) es propiedad de sus respectivos repositorios de origen y puede tener su propia licencia; este dataset agrega y estructura metadatos públicos, no releva la licencia del código fuente original.

Downloads last month
46