ChatCellAnno: AI-Powered Single-Cell Annotation Assistant

ChatCellAnno is a privacy-first, desktop application designed to bridge the gap between single-cell analysis data (Scanpy/Seurat) and Large Language Models (LLMs). Utilizing a 'Human-in-the-loop' approach and a 'Zero-API' design philosophy, it ensures your data remains local and secure while leveraging the power of modern AI.

中文文档 (Chinese Documentation)

Core Features

• GUI Interface: Modern, drag-and-drop interface powered by PySide6.
• Model Agnostic: Compatible with any LLM (Copilot, DeepSeek, ChatGPT, Claude, etc.) via clipboard interaction.
• Privacy First: No API keys required. No network requests are made by the application core.
• Intelligent Parsing: Automatically detects Scanpy and Seurat marker file formats.
• Functional Enrichment: Built-in ORA enrichment analysis (Online via Enrichr or Local via GMT files). Uses database evidence (GO, KEGG, CellMarker) to minimize AI hallucinations.
• Automated Code Generation: Parses AI responses to generate executable Python (Scanpy) or R (Seurat) code for cell type annotation.

Architecture

• Core Logic: The \chatcellanno/\ package handles data extraction, prompt engineering, and response parsing.
• GUI: \gui.py\ provides the user interface with an integrated QtWebEngine browser.
• Zero-API: All data transfer between the app and the LLM is handled exclusively via the system clipboard.

Installation & Run

Prerequisites

• Python 3.10+
• Dependencies: Pandas, PySide6, PySide6-WebEngine, Pyperclip

Quick Start

1. Create a virtual environment (recommended):

   conda create -n chatcellanno python=3.10
   conda activate chatcellanno

2. Install dependencies:

   pip install pandas pyperclip PySide6 PySide6-WebEngine

3. Run the application:

   python gui.py

Workflow

1. Load Data: Drag and drop your marker gene file (.csv, .tsv) into Step 1.
2. Enrichment (Optional): Select Online or Local enrichment in Step 2. You can drag a folder with database files into the app to load local libraries.
3. Configure & Generate: Set species and analysis mode in Step 3. Click 'Generate & Copy Prompt'.
4. AI Browser: Paste and send the prompt to an LLM using the integrated browser.
5. Parse & Export: Copy the AI's markdown table response, paste it into Step 4, and click 'Process AI Output' to get the final R/Python code.

License

MIT License