HN 표시: Finsight – 개인 정보 보호 우선, AI 신용 카드 및 은행 명세서 분석기

AI-powered personal finance analyzer — runs 100% locally, no cloud, no login. Upload a PDF / CSV / Excel bank or credit card statement → AI extracts and categorizes every transaction → interactive dashboard, spending insights, recurring payment detection & chat with your data. | Feature | Description | |---|---| | AI Parsing | Uses a local LLM (via Ollama or LM Studio) to extract transactions from PDFs with near-perfect accuracy | | Any Model | Works with any model — Gemma, Llama, Mistral, Phi, Qwen, DeepSeek, etc. | | Multiple Providers | Choose between Ollama or LM Studio — whichever fits your workflow | | Auto Currency | AI detects the currency from your statement automatically | | Smart Categorization | AI-powered transaction categorization with confidence scores and review flags | | Credit Card Support | Parses credit card statements, detects international transactions, and tracks card-wise spending | | Recurring Payments | Automatically detects subscriptions and recurring payments — spot forgotten subscriptions | | Financial Insights | Spending trends, category breakdowns, and financial health indicators | | Dashboard | Pie charts, trend lines, income vs expense breakdowns | | Chat | Ask questions about your statement in natural language ("What was my highest expense?") | | Budget | Plan next month's budget based on spending patterns | | Privacy | Everything stays on your machine — no data leaves your browser + local LLM | | Multi-format | PDF, CSV, XLS, XLSX supported | You need to have Node JS and NPM installed on your system. Download or install from the official NodeJS website Download a pre-built binary from [NodeJS Downloads](https://nodejs.org/en/download) OR Install using Docker # Docker has specific installation instructions for each operating system. # Please refer to the official documentation at https://docker.com/get-started/ # Pull the Node.js Docker image: docker pull node:24-alpine # Create a Node.js container and start a Shell session: docker run -it --rm --entrypoint sh node:24-alpine # Verify the Node.js version: node -v # Should print "v24.14.0". # Verify npm version: npm -v # Should print "11.9.0". # Download and install Homebrew curl -o- https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh | bash # Download and install Node.js: brew install node@24 # Download and install Homebrew curl -o- https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh | bash # Download and install Node.js: brew install node@24 Choose your LLM provider: Ollama or LM Studio. Both work equally well. Download from ollama.com or: # macOS brew install ollama # Linux curl -fsSL https://ollama.com/install.sh | sh Pick any model you like. Smaller models are faster; larger models are more accurate. # Small & fast (recommended for parsing) ollama pull gemma3:1b ollama pull llama3.2:1b ollama pull phi4-mini # Medium (good balance) ollama pull gemma3:4b ollama pull llama3.2:3b ollama pull mistral # Large (most accurate) ollama pull llama3.1:8b ollama pull gemma3:12b ollama pull qwen2.5:7b ollama serve Ollama runs on http://localhost:11434 by default. Download from lmstudio.ai - Open LM Studio - Go to the Search tab - Download any model you like (Gemma, Llama, Mistral, Phi, Qwen, etc.) - Go to Developer tab (or press Ctrl+D /Cmd+D ) - Click Start Server (default port: 1234) - Important: Go to Settings → Developer and enable CORS (required for browser requests) LM Studio runs on http://localhost:1234 by default. Clone the repository git clone https://github.com/AJ/FinSight.git # Install dependencies npm install # Build the code npm run build # Start the application server npm start Open http://localhost:3000. - Go to Settings (in the sidebar) - Select your LLM Provider (Ollama or LM Studio) - Click Connect — the default URL should work automatically - Select your preferred model from the dropdown - Done — go back to the dashboard and upload a statement! Browser | +-- Upload PDF/CSV/XLS | | | v | Text Extraction (pdfjs-dist) | | | v | Local LLM (Ollama or LM Studio) | | | +-- Statement Type Detection | +-- Transaction Parsing | +-- Currency Detection | | | v | Review & Confirm | +-- Features | +-- Dashboard (charts, stats) +-- Transactions (filter, categorize) +-- Credit Cards (utilization, due dates) +-- Subscriptions (recurring detection) +-- Chat (ask questions) +-- Budget (planning) - Text extraction — pdfjs-dist (for PDFs) or built-in parsers (for CSV/XLS) extract raw text from the document - Statement type detection — AI analyzes the content to detect if it's a bank statement or credit card statement - LLM parsing — The extracted text is sent to your local LLM. The model returns structured JSON with dates, descriptions, amounts, types, and auto-detected currency - Chunking — Long statements are automatically split into chunks to fit within the model's context window - Validation & deduplication — Every transaction is validated (date, amount, type) and duplicates are removed before being sh