|
Download README.md from FurqanAli12345/finance-analyst-concept: direct link, hf CLI and curl.
- Browser
- Download file 2.43 kB
-
https://huggingface.co/spaces/FurqanAli12345/finance-analyst-concept/resolve/main/README.md
- Command line
-
hf download hf://spaces/FurqanAli12345/finance-analyst-concept/README.md
-
curl -L -o README.md https://huggingface.co/spaces/FurqanAli12345/finance-analyst-concept/resolve/main/README.md
2.43 kB
metadata
title: Finance Analyst (Concept)
emoji: 💰
colorFrom: green
colorTo: gray
sdk: static
pinned: false
license: other
Finance Analyst — Document Review Platform (Working manager + concept analytics)
A browser app that pairs a functional financial-document manager with a concept analytics/AI layer for monthly financial review.
Live demo: https://finance-analyst-concept.vercel.app
⚠️ Honest scope: the document-management side is genuinely functional; the analytics/AI side is a concept (illustrative figures). See below.
What actually works today
A real, self-contained client-side document manager:
- Upload financial documents (PDF / XLSX / CSV / images) by click or drag-drop.
- Persist them in the browser via IndexedDB — the actual file bytes are stored and survive a refresh.
- Organize by period/month, with per-month counts and storage totals.
- Download or delete any stored file; clear a month or everything.
- A stats bar (active periods, total files, storage used, last upload) computed from the real stored files.
What is a concept (not yet wired up)
- The KPIs, charts, "AI-generated findings", and extracted line items are illustrative sample data, not computed from your uploads.
- The "AI analyst crew" is a visual workflow demo; there is no LLM call yet, and uploaded files are stored but not yet parsed.
What would make the analytics real (roadmap)
- Parse uploads client-side — SheetJS (
xlsx) for Excel/CSV and pdf.js (+ OCR for scans) to extract tables/text. - A normalization layer mapping parsed rows to accounts/categories.
- Compute the metrics (revenue, expenses, margin, cash flow, variances) from the parsed data and drive the KPIs/charts/table from it.
- For genuine AI findings: a backend proxy calling an LLM (e.g. Anthropic Claude) on the extracted data, so no key is exposed client-side.
- Real report export (PDF/XLSX) instead of the placeholder action.
Tech
Single self-contained HTML file. Vanilla JS, Chart.js, IndexedDB. No build step. Deploys as-is to any static host (Vercel, Netlify, GitHub Pages).
Privacy
No API keys, no backend, no data leaves your browser — uploaded files live only in your browser's IndexedDB.
License
See LICENSE. Provided for review and evaluation.