Rerank existing tables with LLMs, custom workflow chains, and continuous optimization campaigns
Keywords: Python, FastAPI, LLM, Entity Resolution, Excel Automation, Vector Mapping, Database Normalization.
This project demonstrates a reusable architecture for building Excel add-ins with Python backend servers. This implementation specifically tackles database identifier assignment: Matching free-form text entries to standardized terminology using web research, LLM reasoning, and intelligent ranking algorithms.
The workflow diagram above shows the Python/FastAPI backend pipeline that powers real-time terminology normalization directly within Excel.
For Users:
- Installation Guide - Complete setup instructions
- Setup Guide - How to use the add-in
- Configuration Guide - Config file examples and multi-user setup
- Troubleshooting - Common issues & solutions
For Developers:
- Developer Guide - Full development setup, VS Code Office Add-ins Kit, modifying UI/backend
- CLAUDE.md - Architecture principles and internal documentation
- Database Identifier Assignment: Assign free-form names to standardized database identifiers
- Classification & Terminology Management: Assign categories to unstructured text and maintain consistent terminology across documents
- Data Normalization & Entity Linking: Standardize product names, material codes, process terms, and match free-text entities to standardized knowledge bases
- Seamless Excel Integration: - Allows for direct integration with legacy workflows and user-facing spreadsheets.
- Instant AI-Powered Matching - Type in Excel, get standardized results automatically—web research and LLM reasoning happen in real-time
- Three-Tier Intelligence - Exact cache for speed → Fuzzy matching for variants → AI research for unknowns
- Transparent Decision Making - See ranked candidates with confidence scores, entity profiles, and web sources—understand why each match was suggested
- Zero Database Setup - Session-based architecture stores everything in memory—no database installation, no migration scripts
- Complete Audit Trail - Every match logged with timestamps and metadata—perfect for compliance and training data collection
- Drag-and-Drop Configuration - Single JSON file defines everything—share configs via email or network drive
- Visual Confidence Feedback - Color-coded cells show match quality at a glance—green for high confidence, yellow for uncertain
- Multi-User Ready - IP-based authentication with hot-reload—add users without server restart
User Input (Excel Cell)
↓
1. Quick Lookup (cached exact matches)
↓
2. Fuzzy Matching (similarity algorithms)
↓
3. LLM Research (web + entity profiling)
↓
Ranked Candidates with Confidence Scores
↓
Auto-apply or Manual Selection
↓
Logging & State Update
- Microsoft Excel (Desktop or Microsoft 365)
- Python 3.9+ (for backend server)
- LLM API key (Groq recommended, OpenAI supported)
For users who just want to deploy and use the add-in:
1. Download the deployment package
Visit the releases page and download termnorm-deploy-vX.X.X.zip:
👉 Download from GitHub Releases
This package contains only the pre-built files needed for deployment:
- ✅ Pre-built frontend (
dist/) - ✅ Python backend (
backend-api/) - ✅ Manifest files
- ✅ Configuration templates
Extract the zip file to your desired location (e.g., C:\TermNorm-excel\)
2. Start backend server
Run start-server-py-LLMs.bat in the project directory.
Note: Changes to backend-api/config/users.json are hot-reloaded automatically (no server restart needed).
What does the script do?
The script automatically:
- ✅ Sets up virtual environment
- ✅ Installs all dependencies
- ✅ Chooses deployment type (Local or Network)
- ✅ Runs diagnostics and starts server
Manual setup (alternative)
cd backend-api
python -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt
# Set API key
setx GROQ_API_KEY "your_api_key_here"
# Start server
python -m uvicorn main:app --reload # Local
python -m uvicorn main:app --host 0.0.0.0 --port 8000 --reload # Network3. Install Excel add-in
- Microsoft 365 (Simple): Upload
manifest-cloud.xml(from extracted folder) via Home → Add-ins → Upload My Add-in - Desktop Excel (Complex): Requires network deployment. See Desktop Excel Deployment below for full setup.
📖 Full Installation Guide | Client Setup Guide (German)
Desktop Excel cannot use the simple cloud upload method. Instead, it requires hosting the add-in on an internal web server (IIS) and distributing the manifest via network shared folder. Users then configure their Trust Center to access the shared catalog.
This process involves:
- Extracting the release package (
termnorm-deploy-v1.xx.xx.zip) - Deploying to IIS (Windows Server)
- Setting up network share for manifest distribution
- Configuring Trust Center on each user's Excel
📖 Complete deployment guide with step-by-step instructions
Want to modify this codebase? Check out the comprehensive developer guide:
Covers:
- ✅ VS Code Office Add-ins Developer Kit setup
- ✅ Full installation from source (git clone or download)
- ✅ Frontend development (UI customization)
- ✅ Backend development (Python server)
- ✅ Building and deployment
- ✅ Debugging tips and best practices
Learn Excel Add-ins:
For developers who want to modify the code:
git clone https://github.com/runfish5/TermNorm-excel.git
cd TermNorm-excel
npm install # Downloads ~900MB to node_modules/ (required for development)
npm run dev-server # Frontend dev server
# Open another terminal
cd backend-api
python -m venv .venv && .\.venv\Scripts\activate
pip install -r requirements.txt
python -m uvicorn main:app --reload # Backend serverPress F5 in VS Code with Office Add-ins Developer Kit to start debugging in Excel!
Note: The ~1GB local directory size (mostly node_modules/) is normal for development. End users who download the deployment package don't need any of this.