CalvinBackup/donutbrowser
1
0
Fork 0
mirror of https://github.com/zhom/donutbrowser.git synced 2026-04-22 11:56:22 +02:00
Code Issues Packages Projects Releases Wiki Activity

chore: ai inference refresh

Browse Source
This commit is contained in:
zhom
2026-02-22 03:03:23 +04:00
parent ed82f74932
commit 98798b83df
5 changed files with 202 additions and 150 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/prompts/issue-validation.prompt.yml
+76
View File
@@ -0,0 +1,76 @@
messages:
- role: system
content: |-
You are an issue validation assistant for Donut Browser, an anti-detect browser.
Analyze the provided issue content and determine if it contains sufficient information based on these requirements:
For Bug Reports, the issue should include:
1. Clear description of the problem
2. Steps to reproduce the issue (numbered list preferred)
3. Expected vs actual behavior
4. Environment information (OS, browser version, etc.)
5. Error messages, stack traces, or screenshots if applicable
For Feature Requests, the issue should include:
1. Clear description of the requested feature
2. Use case or problem it solves
3. Proposed solution or how it should work
4. Priority level or importance
General Requirements for all issues:
1. Descriptive title
2. Sufficient detail to understand and act upon
3. Professional tone and clear communication
Constraints:
- Maximum 3 items in missing_info array
- Maximum 3 items in suggestions array
- Each array item must be under 80 characters
- overall_assessment must be under 100 characters
- role: user
content: |-
## Issue Content to Analyze:
**Title:** {{issue_title}}
**Body:**
{{issue_body}}
**Labels:** {{issue_labels}}
model: openai/gpt-4.1
responseFormat: json_schema
jsonSchema: |-
{
"name": "issue_validation",
"strict": true,
"schema": {
"type": "object",
"properties": {
"is_valid": {
"type": "boolean",
"description": "Whether the issue contains sufficient information"
},
"issue_type": {
"type": "string",
"enum": ["bug_report", "feature_request", "other"]
},
"missing_info": {
"type": "array",
"items": { "type": "string" },
"description": "Missing information items (max 3, each under 80 characters)"
},
"suggestions": {
"type": "array",
"items": { "type": "string" },
"description": "Suggestions for improvement (max 3, each under 80 characters)"
},
"overall_assessment": {
"type": "string",
"description": "One sentence assessment under 100 characters"
}
},
"required": ["is_valid", "issue_type", "missing_info", "suggestions", "overall_assessment"],
"additionalProperties": false
}
}
.github/prompts/pr-review.prompt.yml
+67
View File
@@ -0,0 +1,67 @@
messages:
- role: system
content: |-
You are a code review assistant for Donut Browser, an open-source anti-detect browser built with Tauri, Next.js, and Rust.
Review the provided pull request and provide constructive feedback. Focus on:
1. Code quality and best practices
2. Potential bugs or issues
3. Security concerns (especially important for an anti-detect browser)
4. Performance implications
5. Consistency with the project's patterns
Constraints:
- Maximum 4 items in feedback array
- Maximum 3 items in suggestions array
- Maximum 2 items in security_notes array
- Each array item must be under 150 characters
- summary must be under 200 characters
- Be constructive and helpful, not harsh
- role: user
content: |-
## Pull Request to Review:
**Title:** {{pr_title}}
**Description:**
{{pr_body}}
**Diff:**
{{pr_diff}}
model: openai/gpt-4.1
responseFormat: json_schema
jsonSchema: |-
{
"name": "pr_review",
"strict": true,
"schema": {
"type": "object",
"properties": {
"summary": {
"type": "string",
"description": "Brief 1-2 sentence summary under 200 characters"
},
"quality_score": {
"type": "string",
"enum": ["good", "needs_work", "critical_issues"]
},
"feedback": {
"type": "array",
"items": { "type": "string" },
"description": "Feedback points (max 4, each under 150 characters)"
},
"suggestions": {
"type": "array",
"items": { "type": "string" },
"description": "Suggestions (max 3, each under 150 characters)"
},
"security_notes": {
"type": "array",
"items": { "type": "string" },
"description": "Security notes if any (max 2, each under 150 characters)"
}
},
"required": ["summary", "quality_score", "feedback", "suggestions", "security_notes"],
"additionalProperties": false
}
}
.github/prompts/release-notes.prompt.yml
+33
View File
@@ -0,0 +1,33 @@
messages:
- role: system
content: |-
You are an expert technical writer tasked with generating comprehensive release notes for Donut Browser, a powerful anti-detect browser desktop app built with Tauri + Next.js that helps users manage multiple browser profiles with proxy support.
Guidelines:
- Use clear, user-friendly language
- Group related commits logically
- Omit minor commits like formatting, typos unless significant
- Focus on user-facing changes
- Use emojis sparingly and consistently
- Keep descriptions concise but informative
- If commits are unclear, infer the purpose from the context
- Only include sections that have relevant changes
- role: user
content: |-
Generate release notes for version {{version}} based on these commits:
{{commits}}
Use this format:
## What's New in {{version}}
[Brief 1-2 sentence overview]
### New Features
### Bug Fixes
### Improvements
### Documentation
### Dependencies
### Developer Experience
model: openai/gpt-4.1
.github/workflows/issue-validation.yml
+20 -111
View File
@@ -26,81 +26,22 @@ jobs:
- name: Checkout repository
uses: actions/checkout@8e8c483db84b4bee98b60c0593521ed34d9990e8 #v6.0.1
- name: Get issue templates
id: get-templates
run: |
if [ -f ".github/ISSUE_TEMPLATE/01-bug-report.md" ]; then
echo "bug-template-exists=true" >> $GITHUB_OUTPUT
fi
if [ -f ".github/ISSUE_TEMPLATE/02-feature-request.md" ]; then
echo "feature-template-exists=true" >> $GITHUB_OUTPUT
fi
- name: Create issue analysis prompt
id: create-prompt
- name: Save issue body to file
env:
ISSUE_TITLE: ${{ github.event.issue.title }}
ISSUE_BODY: ${{ github.event.issue.body }}
ISSUE_LABELS: ${{ join(github.event.issue.labels.*.name, ', ') }}
run: |
cat > issue_analysis.txt << EOF
## Issue Content to Analyze:
**Title:** $ISSUE_TITLE
**Body:**
$ISSUE_BODY
**Labels:** $ISSUE_LABELS
EOF
run: printf '%s' "${ISSUE_BODY:-}" > issue_body.txt
- name: Validate issue with AI
id: validate
uses: actions/ai-inference@a380166897b5408b8fb7dddd148142794cb5624a # v2.0.6
with:
prompt-file: issue_analysis.txt
system-prompt: |
You are an issue validation assistant for Donut Browser, an anti-detect browser.
Analyze the provided issue content and determine if it contains sufficient information based on these requirements:
**For Bug Reports, the issue should include:**
1. Clear description of the problem
2. Steps to reproduce the issue (numbered list preferred)
3. Expected vs actual behavior
4. Environment information (OS, browser version, etc.)
5. Error messages, stack traces, or screenshots if applicable
**For Feature Requests, the issue should include:**
1. Clear description of the requested feature
2. Use case or problem it solves
3. Proposed solution or how it should work
4. Priority level or importance
**General Requirements for all issues:**
1. Descriptive title
2. Sufficient detail to understand and act upon
3. Professional tone and clear communication
Respond ONLY with valid JSON (no markdown fences). Keep responses concise.
JSON structure:
{
"is_valid": true|false,
"issue_type": "bug_report"|"feature_request"|"other",
"missing_info": ["item1", "item2"],
"suggestions": ["suggestion1", "suggestion2"],
"overall_assessment": "One sentence assessment"
}
IMPORTANT CONSTRAINTS:
- Maximum 3 items in missing_info array
- Maximum 3 items in suggestions array
- Each array item must be under 80 characters
- overall_assessment must be under 100 characters
- Output ONLY the JSON object, nothing else
model: gpt-5-mini
prompt-file: .github/prompts/issue-validation.prompt.yml
input: |
issue_title: ${{ github.event.issue.title }}
issue_labels: ${{ join(github.event.issue.labels.*.name, ', ') }}
file_input: |
issue_body: ./issue_body.txt
max-tokens: 1024
- name: Check if first-time contributor
id: check-first-time
@@ -210,7 +151,7 @@ jobs:
model: zai-coding-plan/glm-4.7
- name: Cleanup
run: rm -f issue_analysis.txt comment.md
run: rm -f issue_body.txt comment.md
handle-pr:
if: github.event_name == 'pull_request' && github.actor != 'dependabot[bot]'
@@ -245,54 +186,22 @@ jobs:
gh pr diff ${{ github.event.pull_request.number }} > pr_diff.txt
head -c 10000 pr_diff.txt > pr_diff_truncated.txt
- name: Create PR analysis prompt
- name: Save PR body to file
env:
PR_TITLE: ${{ github.event.pull_request.title }}
PR_BODY: ${{ github.event.pull_request.body }}
run: |
{
printf "## Pull Request to Review:\n\n"
printf "**Title:** %s\n\n" "$PR_TITLE"
printf "**Description:**\n%s\n\n" "$PR_BODY"
printf "**Diff:**\n"
cat pr_diff_truncated.txt
} > pr_analysis.txt
run: printf '%s' "${PR_BODY:-No description provided}" > pr_body.txt
- name: Analyze PR with AI
id: analyze
uses: actions/ai-inference@a380166897b5408b8fb7dddd148142794cb5624a # v2.0.6
with:
prompt-file: pr_analysis.txt
system-prompt: |
You are a code review assistant for Donut Browser, an open-source anti-detect browser built with Tauri, Next.js, and Rust.
Review the provided pull request and provide constructive feedback. Focus on:
1. Code quality and best practices
2. Potential bugs or issues
3. Security concerns (especially important for an anti-detect browser)
4. Performance implications
5. Consistency with the project's patterns
Respond ONLY with valid JSON (no markdown fences).
JSON structure:
{
"summary": "Brief 1-2 sentence summary of what this PR does",
"quality_score": "good"|"needs_work"|"critical_issues",
"feedback": ["feedback point 1", "feedback point 2"],
"suggestions": ["suggestion 1", "suggestion 2"],
"security_notes": ["security note if any"] or []
}
IMPORTANT CONSTRAINTS:
- Maximum 4 items in feedback array
- Maximum 3 items in suggestions array
- Maximum 2 items in security_notes array
- Each array item must be under 150 characters
- summary must be under 200 characters
- Be constructive and helpful, not harsh
- Output ONLY the JSON object, nothing else
model: gpt-5-mini
prompt-file: .github/prompts/pr-review.prompt.yml
input: |
pr_title: ${{ github.event.pull_request.title }}
file_input: |
pr_body: ./pr_body.txt
pr_diff: ./pr_diff_truncated.txt
max-tokens: 1024
- name: Post PR feedback comment
env:
@@ -371,7 +280,7 @@ jobs:
model: zai-coding-plan/glm-4.7
- name: Cleanup
run: rm -f pr_diff.txt pr_diff_truncated.txt pr_analysis.txt comment.md
run: rm -f pr_diff.txt pr_diff_truncated.txt pr_body.txt comment.md
opencode-command:
if: |
.github/workflows/release-notes-generator.yml
+6 -39
View File
@@ -84,45 +84,12 @@ jobs:
if: steps.get-release.outputs.is-prerelease == 'false'
uses: actions/ai-inference@a380166897b5408b8fb7dddd148142794cb5624a # v2.0.6
with:
prompt-file: commits.txt
system-prompt: |
You are an expert technical writer tasked with generating comprehensive release notes for Donut Browser, a powerful anti-detect browser.
Analyze the provided commit messages and generate well-structured release notes following this format:
## What's New in ${{ steps.get-previous-tag.outputs.current-tag }}
[Brief 1-2 sentence overview of the release]
### ✨ New Features
[List new features with brief descriptions]
### 🐛 Bug Fixes
[List bug fixes]
### 🔧 Improvements
[List improvements and enhancements]
### 📚 Documentation
[List documentation updates if any]
### 🔄 Dependencies
[List dependency updates if any]
### 🛠️ Developer Experience
[List development-related changes if any]
Guidelines:
- Use clear, user-friendly language
- Group related commits logically
- Omit minor commits like formatting, typos unless significant
- Focus on user-facing changes
- Use emojis sparingly and consistently
- Keep descriptions concise but informative
- If commits are unclear, infer the purpose from the context
The application is a desktop app built with Tauri + Next.js that helps users manage multiple browser profiles with proxy support.
model: gpt-5-mini
prompt-file: .github/prompts/release-notes.prompt.yml
input: |
version: ${{ steps.get-previous-tag.outputs.current-tag }}
file_input: |
commits: ./commits.txt
max-tokens: 4096
- name: Update release with generated notes
if: steps.get-release.outputs.is-prerelease == 'false'