mirror of
https://github.com/harvard-edge/cs249r_book.git
synced 2026-05-01 01:59:10 -05:00
refactor(auto-label): fetch labels dynamically from GitHub API
- Remove all static label definitions - Fetch labels at runtime via GitHub API (listLabelsForRepo) - Group labels by prefix (area:, type:, format:, other) - Pass dynamic label lists to LLM for analysis - Validate LLM response against actual repo labels only - Add workflow_dispatch trigger for manual testing on existing issues/PRs Now adding/removing labels in GitHub automatically updates the workflow without any code changes.
This commit is contained in:
Vijay Janapa Reddi
268
.github/workflows/auto-label.yml
vendored
268
.github/workflows/auto-label.yml
vendored
@@ -4,11 +4,13 @@
|
||||
# Automatically labels issues and PRs using Ollama LLM analysis.
|
||||
#
|
||||
# When an issue or PR is created:
|
||||
# 1. Analyzes the title and body with a small LLM
|
||||
# 2. Selects appropriate labels from our predefined set
|
||||
# 3. Applies the labels automatically
|
||||
# 1. Fetches all labels from the repository dynamically
|
||||
# 2. Groups them by prefix (area:, type:, format:, etc.)
|
||||
# 3. Sends to LLM to analyze and pick appropriate labels
|
||||
# 4. Applies the selected labels
|
||||
#
|
||||
# This enables the all-contributors workflow to auto-detect projects correctly.
|
||||
# NO STATIC LABEL LISTS - everything is fetched from GitHub at runtime.
|
||||
# Just add/remove labels in GitHub and this workflow adapts automatically.
|
||||
# =============================================================================
|
||||
|
||||
name: '🏷️ Auto Label'
|
||||
@@ -18,35 +20,13 @@ on:
|
||||
types: [opened]
|
||||
pull_request:
|
||||
types: [opened]
|
||||
|
||||
env:
|
||||
# ==========================================================================
|
||||
# AVAILABLE LABELS - Update these if you add/change labels in GitHub
|
||||
# ==========================================================================
|
||||
|
||||
# Area labels (pick ONE that best matches)
|
||||
AREA_LABELS: |
|
||||
area: book - Textbook content (Vol I & II), chapters, Quarto files
|
||||
area: tinytorch - TinyTorch framework, tito CLI, modules, tensors, autograd
|
||||
area: kits - TinyTorch Kits/ML Kits, hardware, Arduino, sensors
|
||||
area: collabs - Colab notebooks and collaborations
|
||||
area: tools - Build tools, scripts, CI/CD, GitHub Actions
|
||||
area: website - Website presentation and styling
|
||||
area: socratiq - SocratiQ AI chatbot
|
||||
|
||||
# Type labels (pick ONE that best matches)
|
||||
TYPE_LABELS: |
|
||||
type: bug - Bug in rendering or code
|
||||
type: errata - Error in textual content
|
||||
type: improvement - Improve existing content
|
||||
type: new - New course content
|
||||
type: question - Further info is requested
|
||||
type: code - Programming exercise content
|
||||
type: citation - Citation or reference to include
|
||||
|
||||
# Special labels (optional, add if applicable)
|
||||
SPECIAL_LABELS: |
|
||||
good first issue - Good starter issue for newbies
|
||||
# Manual trigger for testing on existing issues/PRs
|
||||
workflow_dispatch:
|
||||
inputs:
|
||||
issue_number:
|
||||
description: 'Issue or PR number to label'
|
||||
required: true
|
||||
type: number
|
||||
|
||||
jobs:
|
||||
auto-label:
|
||||
@@ -57,6 +37,114 @@ jobs:
|
||||
pull-requests: write
|
||||
|
||||
steps:
|
||||
- name: Get issue/PR details
|
||||
id: get-details
|
||||
uses: actions/github-script@v7
|
||||
with:
|
||||
script: |
|
||||
// Determine which issue/PR to label
|
||||
let number, title, body;
|
||||
|
||||
if (context.eventName === 'workflow_dispatch') {
|
||||
// Manual trigger - fetch the specified issue/PR
|
||||
number = ${{ inputs.issue_number || 0 }};
|
||||
if (!number) {
|
||||
core.setFailed('No issue_number provided for manual trigger');
|
||||
return;
|
||||
}
|
||||
|
||||
try {
|
||||
const { data: issue } = await github.rest.issues.get({
|
||||
owner: context.repo.owner,
|
||||
repo: context.repo.repo,
|
||||
issue_number: number
|
||||
});
|
||||
title = issue.title;
|
||||
body = issue.body || '';
|
||||
console.log(`Manual trigger for #${number}: ${title}`);
|
||||
} catch (e) {
|
||||
core.setFailed(`Could not fetch issue/PR #${number}: ${e.message}`);
|
||||
return;
|
||||
}
|
||||
} else if (context.eventName === 'issues') {
|
||||
number = context.payload.issue.number;
|
||||
title = context.payload.issue.title;
|
||||
body = context.payload.issue.body || '';
|
||||
} else {
|
||||
number = context.payload.pull_request.number;
|
||||
title = context.payload.pull_request.title;
|
||||
body = context.payload.pull_request.body || '';
|
||||
}
|
||||
|
||||
core.setOutput('number', number);
|
||||
core.setOutput('title', title);
|
||||
core.setOutput('body', body);
|
||||
|
||||
- name: Fetch labels from GitHub
|
||||
id: fetch-labels
|
||||
uses: actions/github-script@v7
|
||||
with:
|
||||
script: |
|
||||
// Fetch all labels from the repository
|
||||
const { data: labels } = await github.rest.issues.listLabelsForRepo({
|
||||
owner: context.repo.owner,
|
||||
repo: context.repo.repo,
|
||||
per_page: 100
|
||||
});
|
||||
|
||||
console.log(`Found ${labels.length} labels in repository`);
|
||||
|
||||
// Group labels by prefix
|
||||
const grouped = {
|
||||
area: [],
|
||||
type: [],
|
||||
format: [],
|
||||
other: []
|
||||
};
|
||||
|
||||
for (const label of labels) {
|
||||
const name = label.name;
|
||||
const desc = label.description || '';
|
||||
|
||||
if (name.startsWith('area:')) {
|
||||
grouped.area.push({ name, description: desc });
|
||||
} else if (name.startsWith('type:')) {
|
||||
grouped.type.push({ name, description: desc });
|
||||
} else if (name.startsWith('format:')) {
|
||||
grouped.format.push({ name, description: desc });
|
||||
} else {
|
||||
grouped.other.push({ name, description: desc });
|
||||
}
|
||||
}
|
||||
|
||||
// Format for LLM prompt
|
||||
const formatGroup = (items) => items
|
||||
.map(l => `${l.name} - ${l.description}`)
|
||||
.join('\n');
|
||||
|
||||
const areaLabels = formatGroup(grouped.area);
|
||||
const typeLabels = formatGroup(grouped.type);
|
||||
const otherLabels = formatGroup(grouped.other);
|
||||
|
||||
// Store all valid label names for validation
|
||||
const allLabelNames = labels.map(l => l.name);
|
||||
|
||||
// Set outputs
|
||||
core.setOutput('area_labels', areaLabels);
|
||||
core.setOutput('type_labels', typeLabels);
|
||||
core.setOutput('other_labels', otherLabels);
|
||||
core.setOutput('all_labels_json', JSON.stringify(allLabelNames));
|
||||
|
||||
// Find defaults (first of each type, or fallback)
|
||||
const defaultArea = grouped.area[0]?.name || '';
|
||||
const defaultType = grouped.type[0]?.name || '';
|
||||
core.setOutput('default_area', defaultArea);
|
||||
core.setOutput('default_type', defaultType);
|
||||
|
||||
console.log('Area labels:', grouped.area.length);
|
||||
console.log('Type labels:', grouped.type.length);
|
||||
console.log('Other labels:', grouped.other.length);
|
||||
|
||||
- name: Analyze with LLM
|
||||
uses: ai-action/ollama-action@v2
|
||||
id: llm
|
||||
@@ -65,30 +153,30 @@ jobs:
|
||||
prompt: |
|
||||
You are a GitHub issue labeler. Analyze this issue/PR and select the most appropriate labels.
|
||||
|
||||
TITLE: ${{ github.event.issue.title || github.event.pull_request.title }}
|
||||
TITLE: ${{ steps.get-details.outputs.title }}
|
||||
|
||||
BODY: ${{ github.event.issue.body || github.event.pull_request.body }}
|
||||
BODY: ${{ steps.get-details.outputs.body }}
|
||||
|
||||
AVAILABLE AREA LABELS (pick exactly ONE):
|
||||
${{ env.AREA_LABELS }}
|
||||
AVAILABLE AREA LABELS (pick exactly ONE - these indicate which part of the project):
|
||||
${{ steps.fetch-labels.outputs.area_labels }}
|
||||
|
||||
AVAILABLE TYPE LABELS (pick exactly ONE):
|
||||
${{ env.TYPE_LABELS }}
|
||||
AVAILABLE TYPE LABELS (pick exactly ONE - these indicate what kind of issue):
|
||||
${{ steps.fetch-labels.outputs.type_labels }}
|
||||
|
||||
SPECIAL LABELS (add only if clearly applicable):
|
||||
${{ env.SPECIAL_LABELS }}
|
||||
OTHER LABELS (pick any that clearly apply, or none):
|
||||
${{ steps.fetch-labels.outputs.other_labels }}
|
||||
|
||||
Instructions:
|
||||
- Pick ONE area label based on which project/component this relates to
|
||||
- Pick ONE type label based on what kind of issue this is
|
||||
- Only add "good first issue" if it's clearly a simple, well-defined task
|
||||
- If unsure about area, default to "area: book"
|
||||
- If unsure about type, default to "type: improvement"
|
||||
- Only add other labels if they clearly and obviously apply
|
||||
- If unsure about area, use "${{ steps.fetch-labels.outputs.default_area }}"
|
||||
- If unsure about type, use "${{ steps.fetch-labels.outputs.default_type }}"
|
||||
|
||||
Return ONLY a JSON object with this exact format (no other text):
|
||||
{"area": "area: book", "type": "type: bug", "special": []}
|
||||
{"area": "area: book", "type": "type: bug", "other": []}
|
||||
|
||||
The "special" array should be empty [] or contain "good first issue" if applicable.
|
||||
The "other" array should be empty [] or contain label names that apply.
|
||||
|
||||
- name: Parse and apply labels
|
||||
uses: actions/github-script@v7
|
||||
@@ -97,85 +185,73 @@ jobs:
|
||||
const response = `${{ steps.llm.outputs.response }}`;
|
||||
console.log('LLM response:', response);
|
||||
|
||||
// Valid labels from our GitHub repo
|
||||
const validAreaLabels = [
|
||||
'area: book',
|
||||
'area: tinytorch',
|
||||
'area: kits',
|
||||
'area: collabs',
|
||||
'area: tools',
|
||||
'area: website',
|
||||
'area: socratiq'
|
||||
];
|
||||
// Get all valid labels from the fetch step
|
||||
const allValidLabels = JSON.parse('${{ steps.fetch-labels.outputs.all_labels_json }}');
|
||||
const defaultArea = '${{ steps.fetch-labels.outputs.default_area }}';
|
||||
const defaultType = '${{ steps.fetch-labels.outputs.default_type }}';
|
||||
|
||||
const validTypeLabels = [
|
||||
'type: bug',
|
||||
'type: errata',
|
||||
'type: improvement',
|
||||
'type: new',
|
||||
'type: question',
|
||||
'type: code',
|
||||
'type: citation'
|
||||
];
|
||||
|
||||
const validSpecialLabels = [
|
||||
'good first issue'
|
||||
];
|
||||
console.log(`Valid labels: ${allValidLabels.length}`);
|
||||
console.log(`Default area: ${defaultArea}`);
|
||||
console.log(`Default type: ${defaultType}`);
|
||||
|
||||
// Parse LLM response
|
||||
let result = { area: 'area: book', type: 'type: improvement', special: [] };
|
||||
let result = { area: defaultArea, type: defaultType, other: [] };
|
||||
|
||||
try {
|
||||
const jsonMatch = response.match(/\{[^}]+\}/);
|
||||
// Find JSON in response (LLM might add extra text)
|
||||
const jsonMatch = response.match(/\{[\s\S]*?\}/);
|
||||
if (jsonMatch) {
|
||||
const parsed = JSON.parse(jsonMatch[0]);
|
||||
if (parsed.area) result.area = parsed.area;
|
||||
if (parsed.type) result.type = parsed.type;
|
||||
if (parsed.special) result.special = parsed.special;
|
||||
if (parsed.other) result.other = parsed.other;
|
||||
}
|
||||
} catch (e) {
|
||||
console.log('Failed to parse JSON, using defaults:', e.message);
|
||||
}
|
||||
|
||||
// Validate and collect labels
|
||||
// Validate and collect labels (only allow labels that exist in repo)
|
||||
const labels = [];
|
||||
|
||||
// Validate area label
|
||||
if (validAreaLabels.includes(result.area)) {
|
||||
if (allValidLabels.includes(result.area)) {
|
||||
labels.push(result.area);
|
||||
} else {
|
||||
console.log(`Invalid area label "${result.area}", defaulting to "area: book"`);
|
||||
labels.push('area: book');
|
||||
} else if (defaultArea) {
|
||||
console.log(`Invalid area label "${result.area}", using default "${defaultArea}"`);
|
||||
labels.push(defaultArea);
|
||||
}
|
||||
|
||||
// Validate type label
|
||||
if (validTypeLabels.includes(result.type)) {
|
||||
if (allValidLabels.includes(result.type)) {
|
||||
labels.push(result.type);
|
||||
} else {
|
||||
console.log(`Invalid type label "${result.type}", defaulting to "type: improvement"`);
|
||||
labels.push('type: improvement');
|
||||
} else if (defaultType) {
|
||||
console.log(`Invalid type label "${result.type}", using default "${defaultType}"`);
|
||||
labels.push(defaultType);
|
||||
}
|
||||
|
||||
// Validate special labels
|
||||
if (Array.isArray(result.special)) {
|
||||
for (const label of result.special) {
|
||||
if (validSpecialLabels.includes(label)) {
|
||||
// Validate other labels
|
||||
if (Array.isArray(result.other)) {
|
||||
for (const label of result.other) {
|
||||
if (allValidLabels.includes(label)) {
|
||||
labels.push(label);
|
||||
} else {
|
||||
console.log(`Ignoring invalid label: "${label}"`);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Apply labels
|
||||
const isIssue = '${{ github.event_name }}' === 'issues';
|
||||
const number = isIssue
|
||||
? context.payload.issue.number
|
||||
: context.payload.pull_request.number;
|
||||
const number = ${{ steps.get-details.outputs.number }};
|
||||
|
||||
console.log(`Applying labels to #${number}: ${labels.join(', ')}`);
|
||||
if (labels.length > 0) {
|
||||
console.log(`Applying labels to #${number}: ${labels.join(', ')}`);
|
||||
|
||||
await github.rest.issues.addLabels({
|
||||
owner: context.repo.owner,
|
||||
repo: context.repo.repo,
|
||||
issue_number: number,
|
||||
labels: labels
|
||||
});
|
||||
await github.rest.issues.addLabels({
|
||||
owner: context.repo.owner,
|
||||
repo: context.repo.repo,
|
||||
issue_number: number,
|
||||
labels: labels
|
||||
});
|
||||
} else {
|
||||
console.log('No valid labels to apply');
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user