title	AI Data Analyst
emoji	📊
colorFrom	blue
colorTo	indigo
sdk	docker
app_port	7860
pinned	false

AI-Data-Analyst

📦 Source: github.com/Mark007-R/AI-Data-Analyst

Upload any CSV / TSV / Excel dataset and get:

• 📝 A written summary of what the data contains and what matters in it
• 📊 Auto-generated interactive charts (ECharts) picked per column/relationship type
• 💬 A chatbot analyst that answers questions by running real SQL against your data through a custom MCP server — answers are computed, never hallucinated

How it works

1. Ingest — your file is loaded into a private per-upload DuckDB database (multi-sheet Excel: every sheet becomes a table)
2. Multi-stage pipeline — Profiler (column types & quality) → Stats (Pearson/Spearman, Cramér's V, η², time trends) → Viz (ECharts options) → Report (an LLM ranks the insights and writes the summary via structured outputs)
3. Chat — an LLM drives an agentic tool-use loop over a custom FastMCP server exposing get_schema, query (read-only SQL), column_stats, make_chart, search_text

The dashboard, charts and template summary work with no API key; a key enables the LLM-written summary and the chatbot.

Run locally

pip install -r requirements.txt
set ANTHROPIC_API_KEY=...              # optional — app degrades gracefully without it
uvicorn app.main:app --port 7860
# open http://localhost:7860

Without an API key, the dashboard/charts still work with a template summary; only the chatbot and LLM-written summary are disabled.

Environment variables

Var	Default	Purpose
ANTHROPIC_API_KEY	—	Enables the LLM summary + chatbot
DATAAI_MODEL	claude-opus-4-8	LLM model id used
DATAAI_DATA_DIR	./data	Where per-dataset DuckDB files live

Deploy

Ships as a Docker image (Dockerfile, port 7860) — deploys as-is to a Hugging Face Space with sdk: docker. Set ANTHROPIC_API_KEY as a Space secret to enable the LLM features.