A universal AI assistant for analyzing and working with codebases, featuring visual diff comparison.
CodePilotAI allows you to converse with the Google Gemini language model, using your project files (from GitHub or a local folder) as context. This version focuses on improving the convenience of working with code changes and exporting your work.
In addition to all the capabilities of previous versions, v3.0.0 introduces new tools to boost productivity:
- Visual Diff Viewer: No more mental code comparison. If the AI suggests changes, the "Show Changes" button opens a clear window where original and proposed code are compared side-by-side, highlighting all additions, deletions, and modifications.
- Export Dialogs to Markdown and HTML: Save your entire conversation history with a single click. Ideal for creating documentation, writing articles, or sharing solutions with colleagues. HTML export preserves styling and code highlighting from the application.
- Enhanced Session Management:
- Backward Compatibility: The application now correctly opens session files created in older program versions, automatically migrating them on the fly.
- Context Menu: Right-clicking on the list of recent projects now opens a menu for quickly opening, deleting, or exporting the current conversation.
- Interface Improvements:
- Network Indicator: The status bar now includes an indicator showing internet availability (green for online, red for offline).
- Warning Cleanup: Obsolete API calls have been removed, resulting in a cleaner startup log.
This is a key new feature in version 2.2.0. We taught the assistant not just to read files, but to understand their structure and interdependencies.
How it works: When analyzing a project, CodePilotAI now builds a "map" of your code (Code-Graph). Using Tree-sitter, it extracts key information from each file:
- Which modules it imports.
- Which functions and classes it defines.
- From which classes others inherit.
This map is passed to the AI along with the code. As a result, the AI sees not just a collection of files, but a holistic architecture.
Code-Graph Example:
--- Обзор структуры проекта (Code-Graph) ---
File: main.py
- Imports:
- from utils import greet_user, add_numbers
- Defines Functions:
- main()
File: utils.py
- Defines Functions:
- greet_user(name: str)
- add_numbers(a: int, b: int)
Advantages:
- Deep Understanding: The AI can answer complex architectural questions, such as: "Which parts of the project will be affected by changing this function?"
- Accuracy: Reduces the likelihood of "hallucinations" because the AI precisely knows where each function is defined.
- Works Across All Modes: Code-Graph complements and improves each of the three analysis modes.
| Mode | How it works | Best suited for |
|---|---|---|
| No RAG (Full Files) | Project Map (Code-Graph) + full code of all files. | Very small projects. |
| RAG (Standard) | Project Map + file summaries + all code in chunks. | Small and medium projects. |
| RAG + Semantic Search | Project Map + summaries + only the most relevant chunks to your query. | Large projects. Maximum efficiency. |
🛠️ Interactive Workflow* Update from Git: If you're working with a local Git repository, the "Update" button quickly re-analyzes only the modified files.
- Saving Generated Files: If the AI suggests code for a new file, the "Save As..." button will automatically prompt you to save it with the appropriate name in the project folder.
- Viewing Changes (Diff Viewer): If the AI suggests changes for an existing file, the "Show Changes" button will open a window for visual version comparison.
- Session Management: All work (chat history, settings, context) is saved into a single
.cpaifile.
- GUI: Python + PySide6 (Qt for Python)
- Language Model: Google Gemini API (
google-generativeai) - GitHub Integration: PyGithub
- Git Integration: GitPython
- Code Parsing (AST & Chunks): Tree-sitter
- Semantic Search:
numpy - Response Rendering: Markdown, Pygments
- Version Comparison:
difflib - Session Storage: SQLite (within the
.cpaifile) - Environment Management:
python-dotenv
To set up the project on your local machine, follow these steps.
- Python 3.10 or higher.
- Git.
- C/C++ build tools for your OS (required for Tree-sitter compilation).
- Windows: Install "Build Tools for Visual Studio" (with the "Desktop development with C++" workload).
- Linux (Debian/Ubuntu):
sudo apt-get install build-essential - macOS:
xcode-select --install
git clone https://github.com/kobaltgit/CodePilotAI
cd CodePilotAIIt is highly recommended to use a virtual environment.
- Windows (Command Prompt):
python -m venv .venv
.venv\Scripts\activate- Linux / macOS:
python3 -m venv .venv
source .venv/bin/activatepip install -r requirements.txtImportant: QtWebEngine is used for the chat functionality. If it doesn't install automatically with PySide6, install it separately:
pip install PySide6-WebEngineFor intelligent code analysis, the application requires compiled language grammars.1. In the project's root folder, create a new folder named grammars.
2. Clone the repositories with the necessary grammars into this folder. The recommended set includes:
cd grammars
git clone https://github.com/tree-sitter/tree-sitter-python
git clone https://github.com/tree-sitter/tree-sitter-javascript
git clone https://github.com/tree-sitter/tree-sitter-html
git clone https://github.com/tree-sitter/tree-sitter-css
git clone https://github.com/tree-sitter/tree-sitter-json
git clone https://github.com/tree-sitter/tree-sitter-java
git clone https://github.com/tree-sitter/tree-sitter-c-sharp
git clone https://github.com/tree-sitter/tree-sitter-cpp
git clone https://github.com/tree-sitter/tree-sitter-go
git clone https://github.com/tree-sitter/tree-sitter-ruby
git clone https://github.com/tree-sitter/tree-sitter-rust
git clone https://github.com/tree-sitter/tree-sitter-bash
git clone https://github.com/ikatyang/tree-sitter-yaml
cd ..- Run the compilation script from the project's root directory:
python build_grammars.py- This will create a single library file (e.g.,
languages.dllfor Windows) in theresources/grammarsfolder.
The application requires at least a Google Gemini key. A GitHub key is necessary for working with repositories on GitHub.
- In the project's root folder, create a file named
.env. - Open it in a text editor and add the lines:
GEMINI_API_KEY="ВАШ_GEMINI_API_КЛЮЧ"
GITHUB_TOKEN="ВАШ_GITHUB_PAT"
- GEMINI_API_KEY: Get your key from Google AI Studio.
- GITHUB_TOKEN: Create a Personal Access Token (classic) in GitHub settings. When creating the token, make sure to check the box in the
reposection.
Alternatively, you can run the program and enter the keys via the interface. They will be automatically saved to the .env file.
python main.py- Setup (first run): If you haven't created the
.envfile, enter and save your API keys in the "Settings" section. - Select source: Choose the "GitHub Repository" or "Local Folder" tab and specify the source.
- Select analysis mode: In the "Settings" section, choose the most suitable mode (Full, RAG, or RAG + Semantic Search).
- Analyze: Click the "Analyze" button. Progress will be displayed on a new progress bar.
- Dialogue: After the analysis is complete, you can ask questions.
- If you've modified files in the local Git repository, click "Refresh" for a quick re-analysis.
- If the AI suggests saving a new file, use the "Save as..." button next to the code block.
- Saving: To avoid analyzing the project every time, save the session via the menu File -> Save Session.
Standard actions for managing sessions (New, Open, Save) and exiting the application.
Show Logs (Ctrl+L): Opens a separate window displaying detailed application logs in real-time.
Allows you to select the interface language (Russian/English). Changes will take effect after restarting the program.
Opens the help window and program information.
The project is distributed under the MIT license. See the LICENSE file for details.