Data Analyst

Interactive TUI data analysis tool powered by local/remote LLM models. Built on the Claude Code engine with a streamlined toolset focused on autonomous data exploration, analysis, and visualization.

Features

• Interactive TUI: Rich terminal interface with blue-themed design
• OpenAI-Compatible API: Works with any OpenAI-compatible endpoint — OpenRouter, Ollama, vLLM, LM Studio, SiliconFlow, etc.
• 6 Core Tools: Bash, Read, Write, Edit, Glob, Grep — everything needed for data analysis workflows
• Compact System Prompt: ~2k token data-analyst-focused prompt (vs ~30k generic prompt), optimized for smaller local models
• Python Analysis: Run pandas, numpy, matplotlib, seaborn, scipy, scikit-learn scripts directly
• Auto-Visualization: Models save plots with plt.savefig() to the working directory

Quick Start

Prerequisites

• Bun (v1.1+)
• An OpenAI-compatible API endpoint

Installation

git clone https://github.com/IIIIQIIII/data-analyst.git
cd data-analyst
bun install

Configuration

Set your API endpoint and model via environment variables:

# OpenRouter example
export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_BASE_URL=https://openrouter.ai/api/v1
export OPENAI_API_KEY=sk-or-v1-your-key-here
export OPENAI_MODEL=z-ai/glm-5.1

# Ollama example
export CLAUDE_CODE_USE_OPENAI=1
export OPENAI_BASE_URL=http://localhost:11434/v1
export OPENAI_API_KEY=unused
export OPENAI_MODEL=qwen2.5:14b

Note: CLAUDE_CODE_USE_OPENAI=1 is auto-set by the CLI entry point, but you can also set it explicitly.

Run

bun run start

Or directly:

bun run --preload ./stubs/globals.ts ./src/entrypoints/cli.tsx

Architecture

This project is a focused fork of the Claude Code TUI framework. Key modifications:

Component	Change
System Prompt	Replaced ~30k generic prompt with ~2k data-analyst prompt
Tools	Trimmed to 6 core tools (Bash, Read, Write, Edit, Glob, Grep)
API Layer	Uses built-in OpenAI-compatible shim (openaiShim.ts)
Branding	Custom blue theme, bar chart icon, "Data Analyst" identity
Entry Point	Simplified cli.tsx, auto-enables OpenAI mode

Project Structure

src/
├── entrypoints/cli.tsx     # Simplified entry point
├── constants/prompts.ts    # Data-analyst system prompt
├── services/api/
│   └── openaiShim.ts       # OpenAI-compatible API bridge
├── components/LogoV2/      # TUI components (blue theme)
├── utils/theme.ts          # Blue color scheme
└── tools.ts                # 6-tool configuration

Supported Models

Any model accessible via OpenAI-compatible API, including:

• OpenRouter: GLM-5.1, DeepSeek, Qwen, Llama, etc.
• Ollama: Local models (Qwen2.5, Llama 3, Mistral, etc.)
• vLLM / LM Studio: Self-hosted inference servers
• SiliconFlow: Chinese model hosting platform
• Any OpenAI-compatible endpoint

Example Usage

> Load the CSV file in this directory and show me a summary

> Create a scatter plot of price vs. rating, colored by category

> Run a correlation analysis on all numeric columns

> Find outliers using IQR method and save the cleaned data

License

Based on Claude Code by Anthropic.