Skip to content
ederaefePublic

About

πŸ“„βž‘οΈπŸ“Š Open-source AI Document-to-Spreadsheet Studio. Convert PDFs, receipts, and handwritten tables into interactive CSV/Excel files using 100% local, vision models.

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

1 watching

Forks

Latest commit

Β 

History

41 Commits

Folders and files

NameName
Last commit message
Last commit date
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Visiolog

License: AGPL v3 Next.js TypeScript Docker Ollama OpenRouter PWA

Open-Source AI Document-to-Spreadsheet Studio & Tabular OCR Engine

Transform paper logbooks, receipts, financial records, handwritten tables, and PDFs into interactive spreadsheets with 100% privacy sovereignty.

Live Demo β€’ GitHub Repository β€’ Quickstart β€’ Features β€’ Self-Hosting β€’ Guides


What is Visiolog?

Visiolog is an open-source, privacy-first document intelligence platform and interactive spreadsheet studio. It converts physical paper documents, photographs, and PDF files into structured digital spreadsheets (.xlsx, .csv) and clean notes in seconds.

Unlike cloud-dependent SaaS OCR services, Visiolog runs 100% locally and air-gapped using browser-native IndexedDB storage and local Ollama vision models (llama3.2-vision, minicpm-v, qwen2.5-vl), or connects seamlessly to cloud backends (Supabase, OpenRouter, Google Gemini) for multi-device synchronization.


Key Features

1. Multimodal Tabular & Note OCR

  • High-Fidelity Table Extraction: Converts complex tabular matrices, multi-line cells, financial ledgers, and scanned forms into standard CSV and Excel.
  • Handwritten Registry Recognition: Parses handwritten logs, dates, quantities, and currency amounts with spatial alignment.
  • Plain Text & Note Mode: Transcribes memos, letters, and unstructured documents with preserved line breaks and list numbering.

2. Multi-Provider Vision AI Engine

  • Local Air-Gapped Inference: Connects directly to local Ollama instances (llama3.2-vision, minicpm-v, qwen2.5-vl) with zero internet access required.
  • Multi-Model Cloud Routing: Native OpenRouter support (meta-llama/llama-3.2-11b-vision-instruct, qwen/qwen-2.5-vl-72b-instruct, claude-3.5-sonnet).
  • High-Speed Flash Vision: Google Gemini Flash models with an automatic 12-key rotation load balancer and exponential retry backoff.
  • Custom OpenAI Endpoints: Plug into LocalAI, vLLM, LM Studio, or private enterprise vision gateways.

3. Interactive Spreadsheet Studio

  • In-browser reactive spreadsheet engine with formulas (=SUM, =AVERAGE, =PRODUCT, =COUNT).
  • Column management (insert, delete, rename, resize), real-time sorting, search filtering, and instant export to .csv and .xlsx.

4. Fixed Settings & Header Reconciliation

  • Define strict master schemas (e.g. Date, Item, Quantity, Price, Total) for continuous batch scanning.
  • Disparate document column structures are automatically aligned and appended into your project master sheet.

5. In-App Database Studio

  • Complete replacement for external database dashboards.
  • Visual table explorer for projects, documents, spreadsheets, profiles, processing_jobs, and system_logs.
  • Inline JSON drawer editor, column sorting, search filters, and 1-click database backup export/restore (.json).

6. Zero-Telemetry Privacy Sovereignty

  • Zero external tracking, zero analytics packages, zero persistent image retention.
  • Binary files process strictly in volatile memory buffers and are discarded immediately after transcription.

Architecture

+---------------------------------------------------------------------------------+
|                                 VISIOLOG STUDIO                                 |
+----------------------------------------+----------------------------------------+
                                         |
            +----------------------------+----------------------------+
            |                                                         |
            v                                                         v
+-------------------------------------+   +-------------------------------------+
|         LOCAL-FIRST ENGINE          |   |          CLOUD-SYNC ENGINE          |
| - 100% Offline Air-Gapped           |   | - Multi-Device Synchronization      |
| - Browser IndexedDB Storage         |   | - Supabase PostgreSQL Database      |
| - Local Ollama Vision Daemon        |   | - OpenRouter / Gemini Cloud Models  |
| - Direct File Export (.csv / .xlsx) |   | - Edge Serverless Hosting           |
+-------------------------------------+   +-------------------------------------+

Quickstart

Prerequisites

  • Node.js 20+ and npm
  • Git

1. Clone Repository

git clone https://github.com/ederaefe/Visiolog.git
cd Visiolog

2. Install Dependencies

npm install --legacy-peer-deps

3. Run Development Server

npm run dev

Open http://localhost:3000 in your web browser. Visiolog will immediately operate in local-first standalone mode with zero cloud configuration required.


Self-Hosting & Docker

1-Command Air-Gapped Deployment with Ollama

Run Visiolog alongside a local Ollama vision container using Docker Compose:

docker compose -f docker-compose.ollama.yml up -d

Pull your vision model into the Ollama container:

docker exec -it visiolog_ollama ollama pull llama3.2-vision

Access the studio at http://localhost:3000.

Standard Docker Build

docker build -t visiolog:latest .
docker run -d -p 3000:3000 -e LOCAL_FIRST=true visiolog:latest

In-App Database Studio

Accessible at /admin/database or /database:

+---------------------------------------------------------------------------------+
| TABLES            | SEARCH & ACTIONS: [ Search records... ] [Export] [Import] [+] |
| - projects        +-------------------------------------------------------------+
| - documents       | ID          | NAME           | CREATED_AT        | ACTIONS  |
| - spreadsheets    | proj_101    | Inventory Q1   | 2026-08-31T12:00  | Edit Del |
| - profiles        | proj_102    | Lab Scans      | 2026-08-31T13:30  | Edit Del |
| - processing_jobs +-------------------------------------------------------------+
| - system_logs     | Pagination: Page 1 of 4                         [<] [>]     |
+-------------------+-------------------------------------------------------------+
  • Table Navigator: Real-time browsing across all application stores.
  • Dual Mode: Direct inspection of browser IndexedDB or Supabase PostgreSQL.
  • Data Portability: 1-click JSON snapshot backup export and restore.

Setup Guides

For users deploying optional cloud backends, refer to our step-by-step 20-step setup manuals:


Environment Variables Reference

Variable Required Default Description
LOCAL_FIRST No false Enables zero-cloud standalone mode
AI_VISION_PROVIDER No gemini Active inference provider (ollama, openrouter, gemini, custom)
OLLAMA_BASE_URL No http://localhost:11434 Ollama daemon base URL
OLLAMA_MODEL No llama3.2-vision Default Ollama vision model
OPENROUTER_API_KEY Conditional - OpenRouter API authentication token
OPENROUTER_MODEL No meta-llama/llama-3.2-11b-vision-instruct OpenRouter model ID
GEMINI_API_KEY Conditional - Primary Google Gemini API key
GEMINI_API_KEY_2..12 No - Multi-key rotation pool keys
NEXT_PUBLIC_SUPABASE_URL No - Supabase project URL for cloud mode
NEXT_PUBLIC_SUPABASE_ANON_KEY No - Supabase public anonymous key
SUPABASE_SERVICE_ROLE_KEY No - Supabase administrative service role key

Contact & Support

For feature requests, enterprise deployments, bug reports, or partnership inquiries:


License

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0) - see the LICENSE file for details.

About

πŸ“„βž‘οΈπŸ“Š Open-source AI Document-to-Spreadsheet Studio. Convert PDFs, receipts, and handwritten tables into interactive CSV/Excel files using 100% local, vision models.

Topics

Resources

Contributing

Security policy

Stars

1 star

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages