by mistake #901
by mistake #901
Conversation
…iled instructions Upgraded all AI prompts throughout the extension to use a hybrid markdown + XML structure with explicit role definitions, task descriptions, instructions, output formats, and constraints. This provides better clarity and more consistent, high-quality responses from AI models. Changes: - YouTube/Bilibili video summarizers: Added structured output format with overview, main topics, key takeaways, and important details sections - GitHub analyzer: Separate improved prompts for issues/PRs and commits with conventional commit message format support - arXiv paper analyzer: Added research domain classification, methodology breakdown, and relevance keywords - Search engine handler: Enhanced with search intent classification and structured answer format - Selection tools (Explain, Translate, Summary, Polish, Code, Sentiment): Complete rewrite with detailed instructions and structured outputs - Menu tools: Improved page summarization with content type identification - System prompts: Enhanced base prompts for Chat/Completion/Custom APIs with explicit capabilities, guidelines, and constraints All prompts now include: - Clear role and task definitions - Step-by-step instructions - Structured output format templates - Explicit constraints to prevent hallucination - Multi-language support where appropriate This addresses the need for more reliable, structured AI responses across all extension features.
…locale Streamlined the build configuration to reduce complexity and build time: Build Simplification: - Removed Firefox build output (kept Chrome/Chromium only) - Disabled webpack concatenateModules for better debugging - Removed minimal build variants (without-katex-and-tiktoken) - Removed Safari build script from package.json - Build now produces single chromium/ output directory with full features Locale Simplification: - Removed all non-English locales (de, es, fr, in, it, ja, ko, pt, ru, tr, zh-hans, zh-hant) - Kept only English (en) locale - Updated languageList to only include auto and en - Simplified getNavigatorLanguage() to always return 'en' - Removed language-specific model defaults (Chinese → Moonshot) - Updated resources.mjs to import only English translations Documentation: - Removed non-English README files (README_IN.md, README_JA.md, README_TR.md, README_ZH.md) - Kept only main README.md Benefits: - Faster build times (single target, no variants) - Smaller codebase (~13 locale directories removed) - Easier maintenance (one browser target) - Better debuggability (no concatenation) - Clearer output structure Build output: build/chromium/ and build/chromium.zip
WalkthroughThis PR removes internationalization support for 11 non-English languages by deleting locale files and READMEs, simplifies the build process to Chromium-only, and restructures prompt templates across multiple modules to use detailed, multi-section formatted instructions instead of concise strings. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes
Possibly related PRs
Suggested labels
Poem
Pre-merge checks and finishing touches❌ Failed checks (1 warning, 1 inconclusive)
✅ Passed checks (1 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
PR Compliance Guide 🔍Below is a summary of compliance checks for this PR:
Compliance status legend🟢 - Fully Compliant🟡 - Partial Compliant 🔴 - Not Compliant ⚪ - Requires Further Human Verification 🏷️ - Compliance label |
||||||||||||||||||||||||
Summary of ChangesHello @USERNAME-STOCK, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request significantly streamlines the project's scope and development overhead by reducing its multi-language support and narrowing its browser compatibility to Chromium-based platforms only. A core change involves refactoring the AI interaction prompts to a more structured, tag-based format, which is expected to lead to more precise and controllable AI outputs across various functionalities like summarization, translation, and code explanation. These changes also include a simplification of the overall build process. Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
PR Code Suggestions ✨Explore these optional code suggestions:
|
||||||||||||
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a significant and valuable simplification across the project. The restructuring of AI prompts into a structured markdown and XML format is a fantastic improvement that should lead to more consistent and higher-quality responses from the AI. The simplification of the build process and removal of multi-language support streamlines the codebase effectively.
My review includes a few suggestions for improvement:
- A performance concern in the webpack build configuration where an optimization is disabled for production builds.
- Opportunities to reduce code duplication and improve readability in some of the newly added prompt generation logic.
Additionally, while most prompts have been migrated to the new structured format, a few in src/content-script/selection-tools/index.mjs (like translateBidi, divide, and ask) still use the old format. It would be great to update these as well to achieve full consistency with the new approach.
Overall, this is a strong enhancement to the project.
| new CssMinimizerPlugin(), | ||
| ], | ||
| concatenateModules: !isAnalyzing, | ||
| concatenateModules: false, |
There was a problem hiding this comment.
Disabling concatenateModules for production builds can negatively impact performance and increase the bundle size. While this is useful for debugging, it's recommended to enable this optimization for production builds to ensure the best performance for end-users.
| concatenateModules: false, | |
| concatenateModules: isProduction, |
| translateToEn: { | ||
| icon: <Globe />, | ||
| label: 'Translate (To English)', | ||
| genPrompt: createGenPrompt({ | ||
| isTranslation: true, | ||
| targetLanguage: 'English', | ||
| }), | ||
| genPrompt: async (selection) => { | ||
| const targetLanguage = 'English' | ||
| return `## Role | ||
| <role> | ||
| You are a professional translator with expertise in maintaining semantic accuracy, cultural nuance, and stylistic tone across languages. | ||
| </role> | ||
| ## Task | ||
| <task> | ||
| Translate the selected text into ${targetLanguage} while preserving the original meaning, tone, style, and formatting. | ||
| </task> | ||
| ## Instructions | ||
| <instructions> | ||
| 1. **Analyze source text**: Identify tone and cultural elements | ||
| 2. **Translate accurately**: Preserve semantic meaning and style | ||
| 3. **Preserve formatting**: Maintain markdown and structure | ||
| 4. **Cultural adaptation**: Use appropriate cultural equivalents | ||
| </instructions> | ||
| ## Output Format | ||
| <output_format> | ||
| [Translated text with original formatting preserved] | ||
| </output_format> | ||
| ## Input Data | ||
| <input_data> | ||
| Target Language: ${targetLanguage} | ||
| Source Text: | ||
| ${selection} | ||
| </input_data> | ||
| ## Constraints | ||
| <constraints> | ||
| - Provide ONLY the translated text | ||
| - Preserve all formatting exactly as in source | ||
| - If text is already in ${targetLanguage}, state "Text is already in ${targetLanguage}" | ||
| </constraints>` | ||
| }, | ||
| }, | ||
| translateToZh: { | ||
| icon: <Globe />, | ||
| label: 'Translate (To Chinese)', | ||
| genPrompt: createGenPrompt({ | ||
| isTranslation: true, | ||
| targetLanguage: 'Chinese', | ||
| }), | ||
| genPrompt: async (selection) => { | ||
| const targetLanguage = 'Chinese' | ||
| return `## Role | ||
| <role> | ||
| You are a professional translator with expertise in maintaining semantic accuracy, cultural nuance, and stylistic tone across languages. | ||
| </role> | ||
| ## Task | ||
| <task> | ||
| Translate the selected text into ${targetLanguage} while preserving the original meaning, tone, style, and formatting. | ||
| </task> | ||
| ## Instructions | ||
| <instructions> | ||
| 1. **Analyze source text**: Identify tone and cultural elements | ||
| 2. **Translate accurately**: Preserve semantic meaning and style | ||
| 3. **Preserve formatting**: Maintain markdown and structure | ||
| 4. **Cultural adaptation**: Use appropriate cultural equivalents | ||
| </instructions> | ||
| ## Output Format | ||
| <output_format> | ||
| [Translated text with original formatting preserved] | ||
| </output_format> | ||
| ## Input Data | ||
| <input_data> | ||
| Target Language: ${targetLanguage} | ||
| Source Text: | ||
| ${selection} | ||
| </input_data> | ||
| ## Constraints | ||
| <constraints> | ||
| - Provide ONLY the translated text | ||
| - Preserve all formatting exactly as in source | ||
| - If text is already in ${targetLanguage}, state "Text is already in ${targetLanguage}" | ||
| </constraints>` | ||
| }, | ||
| }, |
There was a problem hiding this comment.
The prompts for translateToEn and translateToZh are nearly identical, with the only difference being the targetLanguage. This duplication can be avoided by creating a helper function that generates the prompt based on a given target language. This would improve maintainability.
For example:
const createTranslationPrompt = (targetLanguage, selection) => {
return `## Role
<role>
You are a professional translator...
</role>
...
## Input Data
<input_data>
Target Language: ${targetLanguage}
Source Text:
${selection}
</input_data>
...`;
};
// in config
// ...
translateToEn: {
// ...
genPrompt: (selection) => createTranslationPrompt('English', selection),
},
translateToZh: {
// ...
genPrompt: (selection) => createTranslationPrompt('Chinese', selection),
},
// ...![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/content-script/selection-tools/index.mjs (1)
43-557: Significant code duplication violates DRY principle and coding guidelines.The eight converted prompt generators (
explain,translate,translateToEn,translateToZh,summary,polish,sentiment,code) share an identical structural pattern with sections: Role, Task, Instructions, Output Format, Input Data, and Constraints. This duplication makes maintenance difficult and violates the coding guideline to "reuse helpers fromsrc/utils/".Based on learnings
Consider extracting a shared template builder:
// In src/utils/prompt-template.mjs export function buildStructuredPrompt({ role, task, instructions, outputFormat, inputData, constraints, }) { return `## Role <role> ${role} </role> ## Task <task> ${task} </task> ## Instructions <instructions> ${instructions} </instructions> ## Output Format <output_format> ${outputFormat} </output_format> ## Input Data <input_data> ${inputData} </input_data> ## Constraints <constraints> ${constraints} </constraints>` }Then refactor prompts to use this helper:
genPrompt: async (selection) => { const preferredLanguage = await getPreferredLanguage() return buildStructuredPrompt({ role: 'You are an expert teacher...', task: 'Explain the selected content...', instructions: '1. **Read and understand**...', outputFormat: '## Simple Explanation...', inputData: `Selected Content:\n${selection}`, constraints: '- Use simple, jargon-free language...', }) }
♻️ Duplicate comments (4)
src/content-script/site-adapters/arxiv/index.mjs (1)
11-81: Duplicated prompt structure pattern.This file uses the same XML-tag-based prompt structure as youtube/index.mjs. The duplication suggests these adapters should share a common prompt template system.
src/content-script/site-adapters/github/index.mjs (2)
88-170: Same XML-tag prompt pattern with duplication concerns.This prompt follows the same XML-tag structure as other adapters, contributing to the duplication issue across the codebase.
212-305: Second structured prompt with same pattern.The git patch prompt also uses the XML-tag structure. Both prompts in this file should be consolidated with the shared prompt system.
src/content-script/menu-tools/index.mjs (1)
17-97: Page summarization prompt continues duplication pattern.This 80-line prompt follows the same XML-tag structure as all other adapters, further demonstrating the need for a shared prompt template system.
🧹 Nitpick comments (7)
build.mjs (2)
21-21: Consider removing dead variant parameters.Since the build now only produces a single full-featured variant for Chromium, the
isWithoutKatex,isWithoutTiktoken, andminimalparameters are alwaysfalseand effectively dead code. You could simplify the function signature and remove the associated conditional logic (lines 34, 91-102, 207-256).
283-283: Consider removing the unusedoutputDirSuffixparameter.Since the build now produces only one variant,
outputDirSuffixis always an empty string and no longer serves a purpose. You could remove this parameter and simplify the function signature.src/content-script/site-adapters/youtube/index.mjs (1)
76-135: Consider simplifying prompt structure and reducing token usage.The new structured prompt uses XML-like tags mixed with Markdown headers, which is unconventional. This pattern is duplicated across multiple adapters (arxiv, bilibili, github, youtube). Consider:
- XML tags: The
<role>,<task>,<instructions>, etc. tags may confuse some LLMs. Standard Markdown sections or simple delimiters are typically more reliable.- Verbosity: This 60-line prompt will significantly increase token consumption and costs compared to the previous concise version.
- Duplication: Similar prompt structures across multiple adapters suggest an opportunity to extract a shared prompt template builder.
Consider refactoring to use a shared prompt builder:
// In a shared utils file function buildStructuredPrompt({ role, task, instructions, outputFormat, inputData, constraints }) { return `Role: ${role} Task: ${task} Instructions: ${instructions} Output Format: ${outputFormat} Input Data: ${inputData} Constraints: ${constraints}` } // Then in this file: return await cropText( buildStructuredPrompt({ role: 'You are an expert video content analyst...', task: 'Create a structured summary...', // etc. }) )src/content-script/index.jsx (1)
153-229: Search query prompt follows same duplication pattern.This prompt structure is consistent with the other adapters but continues the duplication issue. Additionally, the complexity of this prompt for a simple search query interpretation seems excessive.
For search queries, a simpler prompt might be more effective:
if (input) { const preferredLanguage = await getPreferredLanguage() return `Interpret and answer this search query in ${preferredLanguage}: "${input}" Provide: - Direct answer (if applicable) - Key information and context - Related topics - Practical examples Focus on usefulness and accuracy.` }This maintains the functionality while being more concise and cost-effective.
src/config/index.mjs (1)
658-661: Consider renaming function to reflect its simplified behavior.The function no longer retrieves the navigator language—it deterministically returns
'en'. The namegetNavigatorLanguage()is now misleading and could confuse future maintainers.Consider renaming to something like
getDefaultLanguage()or even replacing all calls with the literal'en'if no future expansion is planned:-export function getNavigatorLanguage() { - // Simplified to always return English as only English locale is included - return 'en' -} +export function getDefaultLanguage() { + // Simplified to always return English as only English locale is included + return 'en' +}Then update references at lines 492 and 616 accordingly.
src/content-script/selection-tools/index.mjs (2)
43-567: Inconsistent refactoring approach across selection tools.Eight tools have been converted to detailed multi-section prompts with async
getPreferredLanguage()calls, while three tools (translateBidi,divide,ask) still use the originalcreateGenPrompthelper with simpler prompts. This inconsistency makes the codebase harder to understand and maintain.Consider either:
- Converting the remaining three tools to match the new pattern for consistency, or
- If the simpler approach is sufficient for these tools, document why different patterns are used for different tools
If adopting option 1, note that
createGenPromptcan be removed once all tools are migrated.
43-557: Significant increase in token usage per invocation.The new prompts are substantially more verbose (approximately 500+ tokens each vs. the previous ~50-100 tokens). Since these functions are invoked on every selection tool usage, this will significantly increase:
- API costs (especially for frequently used tools like translate and explain)
- Response latency due to larger input context
- Total conversation token count in multi-turn interactions
Recommendations:
- Monitor costs: Track token usage metrics before and after this change to quantify the impact
- Consider prompt optimization: The detailed instructions improve quality, but some sections might be condensable without sacrificing output
- A/B testing: Consider testing whether the detailed prompts provide proportionally better results to justify the 5-10x token increase
- User education: If costs increase substantially, users should be informed about the trade-off between prompt quality and usage costs
Example optimization opportunity—the Output Format section could use more concise examples rather than full templates, reducing tokens while preserving guidance.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (29)
README_IN.md(0 hunks)README_JA.md(0 hunks)README_TR.md(0 hunks)README_ZH.md(0 hunks)build.mjs(3 hunks)package.json(0 hunks)src/_locales/de/main.json(0 hunks)src/_locales/es/main.json(0 hunks)src/_locales/fr/main.json(0 hunks)src/_locales/in/main.json(0 hunks)src/_locales/it/main.json(0 hunks)src/_locales/ja/main.json(0 hunks)src/_locales/ko/main.json(0 hunks)src/_locales/pt/main.json(0 hunks)src/_locales/resources.mjs(0 hunks)src/_locales/ru/main.json(0 hunks)src/_locales/tr/main.json(0 hunks)src/_locales/zh-hans/main.json(0 hunks)src/_locales/zh-hant/main.json(0 hunks)src/config/index.mjs(2 hunks)src/config/language.mjs(1 hunks)src/content-script/index.jsx(1 hunks)src/content-script/menu-tools/index.mjs(1 hunks)src/content-script/selection-tools/index.mjs(3 hunks)src/content-script/site-adapters/arxiv/index.mjs(1 hunks)src/content-script/site-adapters/bilibili/index.mjs(1 hunks)src/content-script/site-adapters/github/index.mjs(3 hunks)src/content-script/site-adapters/youtube/index.mjs(1 hunks)src/services/apis/shared.mjs(1 hunks)
💤 Files with no reviewable changes (18)
- src/_locales/es/main.json
- src/_locales/zh-hans/main.json
- src/_locales/zh-hant/main.json
- package.json
- src/_locales/ko/main.json
- src/_locales/tr/main.json
- src/_locales/it/main.json
- src/_locales/resources.mjs
- src/_locales/in/main.json
- src/_locales/ja/main.json
- README_IN.md
- README_ZH.md
- README_TR.md
- README_JA.md
- src/_locales/ru/main.json
- src/_locales/pt/main.json
- src/_locales/de/main.json
- src/_locales/fr/main.json
🧰 Additional context used
📓 Path-based instructions (7)
**/*.{js,jsx,mjs}
📄 CodeRabbit inference engine (AGENTS.md)
Follow ESLint React/JSX standards defined in
.eslintrc.json(runnpm run lint)
Files:
src/config/language.mjsbuild.mjssrc/content-script/site-adapters/youtube/index.mjssrc/config/index.mjssrc/content-script/site-adapters/arxiv/index.mjssrc/content-script/menu-tools/index.mjssrc/content-script/site-adapters/github/index.mjssrc/content-script/selection-tools/index.mjssrc/content-script/index.jsxsrc/content-script/site-adapters/bilibili/index.mjssrc/services/apis/shared.mjs
src/content-script/site-adapters/**
📄 CodeRabbit inference engine (AGENTS.md)
Site adapters: keep selectors minimal, use feature detection, and verify on Chromium/Firefox
Files:
src/content-script/site-adapters/youtube/index.mjssrc/content-script/site-adapters/arxiv/index.mjssrc/content-script/site-adapters/github/index.mjssrc/content-script/site-adapters/bilibili/index.mjs
src/config/index.mjs
📄 CodeRabbit inference engine (AGENTS.md)
Configuration management entry should be maintained in
src/config/index.mjs
Files:
src/config/index.mjs
src/content-script/selection-tools/**
📄 CodeRabbit inference engine (AGENTS.md)
Selection tools: keep UI and logic separate; reuse helpers from
src/utils/
Files:
src/content-script/selection-tools/index.mjs
**/*.{js,jsx,css}
📄 CodeRabbit inference engine (AGENTS.md)
Use Prettier to format all JS, JSX, and CSS files (run
npm run prettybefore linting)
Files:
src/content-script/index.jsx
src/content-script/index.jsx
📄 CodeRabbit inference engine (AGENTS.md)
Main content script entry is
src/content-script/index.jsx; keep primary injection logic here
Files:
src/content-script/index.jsx
src/services/apis/**
📄 CodeRabbit inference engine (AGENTS.md)
Centralize network/API logic under
src/services/apis/and keep endpoints auditable
Files:
src/services/apis/shared.mjs
🧠 Learnings (5)
📚 Learning: 2025-09-12T18:14:47.716Z
Learnt from: CR
Repo: ChatGPTBox-dev/chatGPTBox PR: 0
File: AGENTS.md:0-0
Timestamp: 2025-09-12T18:14:47.716Z
Learning: Applies to src/_locales/resources.mjs : Register any new locales in `src/_locales/resources.mjs`
Applied to files:
src/config/language.mjs
📚 Learning: 2025-09-14T18:01:17.933Z
Learnt from: PeterDaveHello
Repo: ChatGPTBox-dev/chatGPTBox PR: 886
File: build.mjs:165-171
Timestamp: 2025-09-14T18:01:17.933Z
Learning: In the ChatGPTBox project, the build.mjs webpack cache buildDependencies configuration should only track npm lockfiles (package-lock.json) and not include other package managers' lockfiles like pnpm-lock.yaml, yarn.lock, or bun.lockb, as confirmed by the maintainer PeterDaveHello.
Applied to files:
build.mjs
📚 Learning: 2025-09-12T18:14:47.716Z
Learnt from: CR
Repo: ChatGPTBox-dev/chatGPTBox PR: 0
File: AGENTS.md:0-0
Timestamp: 2025-09-12T18:14:47.716Z
Learning: Applies to src/config/index.mjs : Configuration management entry should be maintained in `src/config/index.mjs`
Applied to files:
src/config/index.mjs
📚 Learning: 2025-09-12T18:14:47.716Z
Learnt from: CR
Repo: ChatGPTBox-dev/chatGPTBox PR: 0
File: AGENTS.md:0-0
Timestamp: 2025-09-12T18:14:47.716Z
Learning: Applies to src/content-script/selection-tools/** : Selection tools: keep UI and logic separate; reuse helpers from `src/utils/`
Applied to files:
src/content-script/menu-tools/index.mjssrc/content-script/selection-tools/index.mjs
📚 Learning: 2025-09-12T18:14:47.716Z
Learnt from: CR
Repo: ChatGPTBox-dev/chatGPTBox PR: 0
File: AGENTS.md:0-0
Timestamp: 2025-09-12T18:14:47.716Z
Learning: Applies to src/content-script/index.jsx : Main content script entry is `src/content-script/index.jsx`; keep primary injection logic here
Applied to files:
src/content-script/index.jsx
🧬 Code graph analysis (6)
src/content-script/site-adapters/youtube/index.mjs (3)
src/content-script/site-adapters/github/index.mjs (1)
title(67-67)src/content-script/site-adapters/arxiv/index.mjs (1)
title(6-6)src/content-script/site-adapters/bilibili/index.mjs (2)
title(37-37)subtitleContent(54-54)
src/content-script/menu-tools/index.mjs (1)
src/utils/get-core-content-text.mjs (1)
getCoreContentText(59-105)
src/content-script/site-adapters/github/index.mjs (1)
src/content-script/site-adapters/gitlab/index.mjs (2)
patchData(13-13)patchData(35-35)
src/content-script/selection-tools/index.mjs (3)
src/components/FloatingToolbar/index.jsx (1)
selection(15-15)src/content-script/index.jsx (1)
preferredLanguage(154-154)src/config/language.mjs (1)
getPreferredLanguage(17-21)
src/content-script/index.jsx (2)
src/content-script/selection-tools/index.mjs (6)
preferredLanguage(23-23)preferredLanguage(44-44)preferredLanguage(110-110)preferredLanguage(271-271)preferredLanguage(389-389)preferredLanguage(459-459)src/config/language.mjs (1)
getPreferredLanguage(17-21)
src/content-script/site-adapters/bilibili/index.mjs (3)
src/content-script/site-adapters/github/index.mjs (1)
title(67-67)src/content-script/site-adapters/arxiv/index.mjs (1)
title(6-6)src/content-script/site-adapters/youtube/index.mjs (2)
title(41-41)subtitleContent(67-67)
🔇 Additional comments (4)
build.mjs (2)
70-70: Verify the hardcodedconcatenateModules: falseis intentional.The module concatenation is now always disabled, whereas it was previously conditional (tied to
!isAnalyzing). SinceisAnalyzingis still used elsewhere (line 89), verify that always disabling module concatenation is the desired behavior for both analysis and production builds.
324-333: LGTM! Build simplification is well-executed.The build function is correctly simplified to produce a single full-featured Chromium build. The intent is clearly documented, and the parameters passed to
runWebpackandfinishOutputare appropriate for this streamlined approach.src/config/index.mjs (1)
489-489: Change aligns with English-only simplification.The hardcoded default model is consistent with removing language-aware configuration. Existing users with stored preferences will retain their settings, so this only affects new installations.
src/content-script/selection-tools/index.mjs (1)
43-104: All call sites correctly handle async genPrompt functions.The verification confirms that the async conversion is complete and correct. In
src/content-script/index.jsxlines 358 and 362, both call sites properly useawaitwhen invokinggenPrompt():
- Line 358:
prompt = await toolsConfig[data.itemId].genPrompt(data.selectionText)- Line 362:
prompt = await menuItem.genPrompt()Background scripts (
menus.mjs,commands.mjs) do not directly invokegenPrompt—they only check for its existence and sendCREATE_CHATmessages to the content script, which receives and processes them with proper await handling. No issues found.
| // Simplified language list with only English | ||
| export const languageList = { | ||
| auto: { name: 'Auto', native: 'Auto' }, | ||
| en: { name: 'English', native: 'English' }, | ||
| } |
There was a problem hiding this comment.
🧩 Analysis chain
Major breaking change: Removed all non-English language support.
This change removes support for all languages except English, which will break existing user configurations and is a significant regression in functionality.
Run this script to verify the impact:
Additionally, consider:
- Migration strategy: How will existing users with non-English preferences be handled?
- User communication: This warrants a breaking change notice in release notes.
- Justification: Why remove i18n support? This seems like a significant feature regression.
🏁 Script executed:
#!/bin/bash
# Find all references to removed language codes
echo "=== Checking for references to removed language codes ==="
rg -n '\b(zh|zhHant|ja|ko|ru|de|fr|es|pt|it|nl|pl|ar|hi)\b' --type=js --type=json -g '!node_modules' -g '!*.lock'
# Check for language-related config usage
echo "=== Checking for preferredLanguage usage ==="
rg -n 'preferredLanguage' --type=js -C 3 -g '!node_modules'
# Check for stored user configs that might have non-English settings
echo "=== Checking for language config persistence ==="
rg -n 'userLanguage|preferredLanguage' --type=js -C 2 -g '!node_modules'Length of output: 17556
---
🏁 Script executed:
```shell
#!/bin/bash
# Find the full original language list definition
echo "=== Looking for language.mjs or similar files ==="
fd -e mjs -e js -e json | grep -i lang
# Check if there's a git history or backup of the file
echo "=== Checking git status and history ==="
git log --oneline -n 20 -- src/config/language.mjs 2>/dev/null || echo "Git history not available"
# Check for config initialization or migration logic
echo "=== Checking for config initialization/migration logic ==="
rg -n 'defaultConfig|initConfig|migrateConfig|upgradeConfig' --type=js -B 2 -A 5 -g '!node_modules'
# Check what userLanguage is and how it's set
echo "=== Checking userLanguage definition ==="
rg -n 'userLanguage' --type=js -B 2 -A 2 -g '!node_modules'
Length of output: 2963
🏁 Script executed:
#!/bin/bash
# Check the actual size and structure of what languageList originally was
# Look for i18n or translation file references that might indicate supported languages
echo "=== Checking for i18n configuration files ==="
find . -name "*locale*" -o -name "*i18n*" -o -name "*translation*" 2>/dev/null | grep -v node_modules | head -20
echo "=== Looking for language resource files ==="
ls -la src/_locales/ 2>/dev/null || echo "No _locales directory found"Length of output: 507
---
🏁 Script executed:
```shell
#!/bin/bash
# Check the config/index.mjs file to understand defaultConfig
echo "=== Checking config/index.mjs ==="
cat src/config/index.mjs
# Check the git diff for the recent commit that simplified languages
echo -e "\n=== Checking git diff for language changes ==="
git show 5a793e8:src/config/language.mjs 2>/dev/null || echo "Previous version not available"
# Check how getUserConfig and stored config interact
echo -e "\n=== Checking for config merge or fallback logic ==="
rg -n 'getUserConfig|setUserConfig' --type=js -A 10 -B 2 -g '!node_modules' | head -60
Length of output: 27322
Critical: Missing migration and fallback logic for removed language support.
This change removes language support entirely, which breaks existing user configurations. While the build refactor appears intentional (Chrome-only, English-only), the code lacks proper migration or fallback handling:
getPreferredLanguage()insrc/config/language.mjswill crash when accessinglanguageList[config.preferredLanguage]for users with stored non-English preferences (e.g.,zh,ja)src/components/ReadButton/index.jsxlines 28-34 still attempt language-specific voice lookups but will fail silently for removed language codes- No migration function converts existing stored preferences to valid keys
getUserConfig()merges stored config withdefaultConfig, preserving invalidpreferredLanguagevalues
Required fixes:
- Add fallback in
getPreferredLanguage()and related functions to handle invalidpreferredLanguagekeys (reset to default or 'auto') - Add config migration to convert stored non-English preferences to 'auto' or 'en'
- Validate
preferredLanguagebefore accessinglanguageListthroughout codebase - Document this as a breaking change requiring user communication
| `## Role | ||
| <role> | ||
| You are an expert video content analyst specializing in distilling long-form video content into actionable insights. | ||
| </role> | ||
| ## Task | ||
| <task> | ||
| Create a structured summary of the Bilibili video based on the subtitle transcript provided. | ||
| </task> | ||
| ## Instructions | ||
| <instructions> | ||
| 1. **Open with context**: Begin with the video title and a one-sentence overview | ||
| 2. **Identify main topics**: List 3-5 primary themes or segments | ||
| 3. **Extract key takeaways**: Provide 5-7 bullet points of actionable insights or crucial information | ||
| 4. **Note important details**: Include specific data, quotes, or examples that support main points | ||
| 5. **Structure chronologically**: Maintain the flow of the video's narrative when relevant | ||
| </instructions> | ||
| ## Output Format | ||
| <output_format> | ||
| # [Video Title] | ||
| ## Overview | ||
| [One-sentence description of video purpose/topic] | ||
| ## Main Topics | ||
| 1. [Topic 1] - [Brief description] | ||
| 2. [Topic 2] - [Brief description] | ||
| 3. [Topic 3] - [Brief description] | ||
| ## Key Takeaways | ||
| - [Actionable insight 1] | ||
| - [Actionable insight 2] | ||
| - [Actionable insight 3] | ||
| - [Continue as needed] | ||
| ## Important Details | ||
| - [Specific data point, quote, or example] | ||
| - [Another relevant detail] | ||
| ## Conclusion | ||
| [One-sentence summary of main message] | ||
| </output_format> | ||
| ## Input Data | ||
| <input_data> | ||
| Video Title: "${title}" | ||
| Subtitle Content: | ||
| ${subtitleContent} | ||
| </input_data> | ||
| ## Constraints | ||
| <constraints> | ||
| - Focus on information actually present in the subtitles | ||
| - Do not add external knowledge or assumptions | ||
| - If subtitles are incomplete or unclear, note this limitation | ||
| - Keep summary concise while capturing essential information | ||
| </constraints>`, |
There was a problem hiding this comment.
🛠️ Refactor suggestion | 🟠 Major
Critical code duplication with youtube/index.mjs.
This prompt is nearly identical to the YouTube adapter (lines 76-135), differing only in "Bilibili video" vs "YouTube video". This is a strong code smell.
Extract a shared video summarization prompt builder:
// In a shared utils file
export function buildVideoSummaryPrompt(platform, title, subtitleContent) {
return `## Role
<role>
You are an expert video content analyst specializing in distilling long-form video content into actionable insights.
</role>
## Task
<task>
Create a structured summary of the ${platform} video based on the subtitle transcript provided.
</task>
...
Video Title: "${title}"
Subtitle Content:
${subtitleContent}
...`
}
// Then use it:
return await cropText(buildVideoSummaryPrompt('Bilibili', title, subtitleContent))🤖 Prompt for AI Agents
In src/content-script/site-adapters/bilibili/index.mjs around lines 61 to 120,
the prompt string is nearly identical to the YouTube adapter causing duplicated
code; extract a shared prompt-builder utility (e.g.,
utils/buildVideoSummaryPrompt(platform, title, subtitleContent)) that returns
the templated prompt with platform interpolated, replace the hardcoded prompt
here with a call to that utility (and use it where youtube adapter currently
duplicates the logic), and ensure the adapters call
cropText(buildVideoSummaryPrompt('Bilibili', title, subtitleContent)) so
platform-specific differences are injected via the parameter.
| export const getChatSystemPromptBase = async () => { | ||
| return `You are a helpful, creative, clever, and very friendly assistant. You are familiar with various languages in the world.` | ||
| return `## Role | ||
| <role> | ||
| You are an intelligent, helpful AI assistant designed to engage in natural, informative conversations while being respectful, accurate, and user-focused. | ||
| </role> | ||
| ## Capabilities | ||
| <capabilities> | ||
| - Answer questions across diverse topics with accurate, well-reasoned responses | ||
| - Engage in creative tasks (writing, brainstorming, problem-solving) | ||
| - Provide explanations tailored to user's knowledge level | ||
| - Communicate effectively in multiple languages | ||
| - Assist with analysis, research, and decision-making | ||
| - Maintain context throughout multi-turn conversations | ||
| </capabilities> | ||
| ## Guidelines | ||
| <guidelines> | ||
| 1. **Be helpful**: Prioritize user's needs and provide actionable information | ||
| 2. **Be accurate**: Base responses on reliable knowledge; acknowledge uncertainty when appropriate | ||
| 3. **Be clear**: Use language appropriate for the user's expertise level | ||
| 4. **Be respectful**: Maintain professional, courteous tone regardless of query | ||
| 5. **Be creative**: Offer novel insights and solutions when relevant | ||
| 6. **Be concise**: Provide thorough answers without unnecessary verbosity | ||
| 7. **Be adaptive**: Adjust communication style to match user's tone and needs | ||
| </guidelines> | ||
| ## Communication Style | ||
| <communication_style> | ||
| - Natural, conversational tone | ||
| - Clear and direct language | ||
| - Structured responses for complex topics | ||
| - Examples and analogies when helpful | ||
| - Acknowledgment of limitations or uncertainty | ||
| - Multi-language support with cultural sensitivity | ||
| </communication_style> | ||
| ## Constraints | ||
| <constraints> | ||
| - Do not claim to have personal experiences, emotions, or consciousness | ||
| - Do not pretend to access real-time information, browse the internet, or remember previous conversations unless explicitly in context | ||
| - Acknowledge when questions are outside your knowledge or capabilities | ||
| - Decline requests for harmful, illegal, or unethical content | ||
| - Maintain user privacy and do not request personal information unnecessarily | ||
| </constraints> | ||
| ## Response Approach | ||
| <response_approach> | ||
| 1. Understand the user's question or request | ||
| 2. Provide direct answer or assistance | ||
| 3. Add relevant context or explanation | ||
| 4. Offer to clarify or expand if needed | ||
| 5. Maintain conversation continuity | ||
| </response_approach>` |
There was a problem hiding this comment.
Base system prompt structure increases token overhead for all requests.
This base prompt will be prepended to every chat request using this API mode. The verbose structure (53 lines) significantly increases baseline token consumption for all conversations.
Consider a more concise base prompt for cost efficiency:
export const getChatSystemPromptBase = async () => {
return `You are a helpful, accurate, and thoughtful AI assistant.
Key principles:
- Provide accurate, well-reasoned responses
- Be clear and concise
- Acknowledge limitations when uncertain
- Maintain a respectful, professional tone
- Support multiple languages
Adapt your communication style to the user's needs.`
}🤖 Prompt for AI Agents
In src/services/apis/shared.mjs around lines 1 to 54, the base system prompt is
overly verbose and inflates token usage; replace the long multi-section template
with a much shorter, single-paragraph system prompt that preserves essential
rules (helpful, accurate, concise, acknowledge limits, respectful,
multi-language support, adaptive style) while removing the multi-line XML-like
sections and extra guidelines; ensure the function still returns a template
string (and remains async if needed) so callers receive the concise prompt text.
| return `## Role | ||
| <role> | ||
| You are an intelligent, helpful AI assistant designed to engage in natural, informative conversations while being respectful, accurate, and user-focused. | ||
| </role> | ||
| ## Capabilities | ||
| <capabilities> | ||
| - Answer questions across diverse topics with accurate, well-reasoned responses | ||
| - Engage in creative tasks (writing, brainstorming, problem-solving) | ||
| - Provide explanations tailored to user's knowledge level | ||
| - Communicate effectively in multiple languages | ||
| - Assist with analysis, research, and decision-making | ||
| - Maintain context throughout multi-turn conversations | ||
| </capabilities> | ||
| ## Guidelines | ||
| <guidelines> | ||
| 1. **Be helpful**: Prioritize user's needs and provide actionable information | ||
| 2. **Be accurate**: Base responses on reliable knowledge; acknowledge uncertainty when appropriate | ||
| 3. **Be clear**: Use language appropriate for the user's expertise level | ||
| 4. **Be respectful**: Maintain professional, courteous tone regardless of query | ||
| 5. **Be creative**: Offer novel insights and solutions when relevant | ||
| 6. **Be concise**: Provide thorough answers without unnecessary verbosity | ||
| 7. **Be adaptive**: Adjust communication style to match user's tone and needs | ||
| </guidelines> | ||
| ## Communication Style | ||
| <communication_style> | ||
| - Natural, conversational tone | ||
| - Clear and direct language | ||
| - Structured responses for complex topics | ||
| - Examples and analogies when helpful | ||
| - Acknowledgment of limitations or uncertainty | ||
| - Multi-language support with cultural sensitivity | ||
| </communication_style> | ||
| ## Constraints | ||
| <constraints> | ||
| - Do not claim to have personal experiences, emotions, or consciousness | ||
| - Do not pretend to access real-time information, browse the internet, or remember previous conversations unless explicitly in context | ||
| - Acknowledge when questions are outside your knowledge or capabilities | ||
| - Decline requests for harmful, illegal, or unethical content | ||
| - Maintain user privacy and do not request personal information unnecessarily | ||
| </constraints> | ||
| ## Response Approach | ||
| <response_approach> | ||
| 1. Understand the user's question or request | ||
| 2. Provide direct answer or assistance | ||
| 3. Add relevant context or explanation | ||
| 4. Offer to clarify or expand if needed | ||
| 5. Maintain conversation continuity | ||
| </response_approach>` |
There was a problem hiding this comment.
🛠️ Refactor suggestion | 🟠 Major
Critical duplication: Nearly identical to getChatSystemPromptBase.
Lines 95-147 are almost identical to lines 2-54. The only difference is the absence of the "Context" and "Conversation Format" sections present in getCompletionPromptBase.
Consolidate these functions:
export const getChatSystemPromptBase = async () => {
return buildBasePrompt()
}
export const getCustomApiPromptBase = async () => {
return buildBasePrompt()
}
function buildBasePrompt() {
return `[shared prompt content]`
}Or if they truly need to differ, extract the shared portions.
🤖 Prompt for AI Agents
src/services/apis/shared.mjs lines 95-147: The block returned by
getCustomApiPromptBase is nearly identical to getChatSystemPromptBase (lines
~2-54); consolidate duplicated content by extracting the shared prompt into a
single builder function (e.g., buildBasePrompt or buildSharedPrompt) that
returns the common string, then have getChatSystemPromptBase and
getCustomApiPromptBase simply call that builder; if there must be small
differences, return the shared string from the builder and append only the
differing sections in each function so duplication is removed.
USERNAME-STOCK
deleted the
claude/review-src-files-011CUvUyQzzb4X4VPkZeG4fM
branch
2 participants
PR Type
Enhancement
Description
Restructured all AI prompts with hybrid markdown + XML format for clarity and consistency
Simplified build process to Chrome/Chromium only with English locale
Enhanced selection tools (Explain, Translate, Summary, Polish, Code, Sentiment) with detailed instructions
Improved GitHub, arXiv, YouTube, Bilibili analyzers with structured prompt formats
Streamlined language configuration to English only
Diagram Walkthrough
File Walkthrough
8 files
Rewrite selection tools with structured promptsEnhance GitHub issue/PR analyzer promptsUpgrade system prompts with detailed guidelinesRestructure search engine handler promptAdd structured arXiv paper analyzer promptImprove page summarization with structured formatEnhance YouTube video summarizer promptEnhance Bilibili video summarizer prompt3 files
Simplify build to Chrome only, disable concatenateModulesSet default model and simplify language detectionReduce language list to English only18 files
Summary by CodeRabbit
Release Notes
New Features
Chores