Spaces:
Sleeping
Sleeping
File size: 2,222 Bytes
cf52e88 e2f9a1d cf52e88 e2f9a1d cf52e88 e2f9a1d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 | ---
title: CSV to Insights Analyst
emoji: 📊
colorFrom: gray
colorTo: red
sdk: docker
app_port: 7860
pinned: false
---
# CSV → Insights Analyst 📊
Upload a CSV → get **auto-generated charts + a plain-English briefing you can trust**.
> Project **#12** of the *"30 AI Projects in 15 Days"* build-in-public challenge.
> Focus: **code execution + a data agent** (vs. #11's long-running service).
## Honest by construction
The AI **never invents a number.** A deterministic pandas pass computes *every* statistic and chart;
the model only turns that computed profile into prose. Units the model tries to guess (mm, °C, …) are
stripped — they aren't in your file. Same grounded discipline as the #10 RAG app.
## What it does
- **Profile** — shape, dtypes, missing %, numeric summaries, top categories, correlations, outliers,
date-range detection, and plain-language data-quality flags.
- **Charts** — trend-over-time, correlation heatmap, distributions, and category bars (GritAI design system).
- **Findings** — a short, plain-English briefing grounded only in the computed numbers, with follow-up
questions the data could answer next.
## How it works
- `profiler.py` — computes the full structured profile (pandas). The source of every number.
- `charts.py` — matplotlib charts from the profile, returned as inline base64 PNGs.
- `narrator.py` — turns the profile into prose. Local Ollama by default (free/private); Claude optional.
A deterministic post-pass strips any units the model added.
- `app.py` — Flask UI: upload or try the sample, 12 MB cap, data handled in-memory (never persisted).
## Two tiers
- **This hosted demo** narrates via Claude-haiku (charts + stats are 100% local pandas).
- **GritAI Studio** (paid) runs the whole thing — narrative included — on your own local GPU fleet:
free per-analysis, fully private, nothing leaves your network.
## Run locally
```bash
pip install -r requirements.txt
python app.py # http://127.0.0.1:7860
# Backend: set INSIGHTS_LLM_BACKEND=ollama (default, local) or =claude (set ANTHROPIC_API_KEY)
```
---
Built by **Robert Lucyk** · [GritAI Solutions](https://gritai.solutions) · part of the 30-in-15 challenge.
|