# skill-ollama-deepseek-ocr-tool

skill-ollama-deepseek-ocr-tool — a skill for local batch text recognition from images using DeepSeek-OCR and Ollama. Converts photos of pages, slides, and scans into a single markdown document without cloud services.

- skill · `type=skill` · Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, Windsurf (Devin Desktop), Claude.ai · 1 install · ★ 4 repo stars
- Author: dnvriend
- License: MIT · policy=mirror · the beginning of the file is at the end of this page
- Source: https://github.com/dnvriend/ollama-deepseek-ocr-tool/tree/HEAD/plugins/ollama-deepseek-ocr-tool/skills/ollama-deepseek-ocr-tool/SKILL.md
- Updated: 2025-12-05
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/skills/skill-ollama-deepseek-ocr-tool

## What it does

This skill provides access to `ollama-deepseek-ocr-tool`, a CLI utility that processes sequences of images and outputs results as markdown. It recognizes text, headings, lists, and tables while preserving formatting. Files are sorted in natural order (IMG_1 < IMG_10), and processing runs entirely locally.

Requirements are Ollama (local LLM runtime), the DeepSeek-OCR model (~6 GB), Python 3.14+, and the `uv` package manager. Processing speed is approximately 3 seconds per image on M4. The tool works only with printed text—handwriting is not supported—and does not describe diagram contents, only extracting text labels and captions.

Processing is batch-oriented: images are matched by pattern (e.g., `*.png` or `lecture-week5/*.jpg`), and results are concatenated into one file. Each text block is marked with a source comment. Three logging levels are available: `-v` shows operations, `-vv` provides detailed info, `-vvv` displays library internals for deep debugging.

## Install

### Claude Code (native platform)

1. **Skills CLI (Vercel)**
```bash
npx skills add dnvriend/ollama-deepseek-ocr-tool --skill skill-ollama-deepseek-ocr-tool -a claude-code
```

2. **Manually: put the files in place**
- File: `~/.claude/skills/ollama-deepseek-ocr-tool/SKILL.md` — Copy the whole skill folder; for one project use .claude/skills/ollama-deepseek-ocr-tool/

### Codex

1. **Skills CLI (Vercel)**
```bash
npx skills add dnvriend/ollama-deepseek-ocr-tool --skill skill-ollama-deepseek-ocr-tool -a codex
```

2. **Manually: put the files in place**
- File: `.agents/skills/ollama-deepseek-ocr-tool/SKILL.md` — For all projects use ~/.agents/skills/

### Cursor

1. **Skills CLI (Vercel)**
```bash
npx skills add dnvriend/ollama-deepseek-ocr-tool --skill skill-ollama-deepseek-ocr-tool -a cursor
```

2. **Manually: put the files in place**
- File: `.cursor/skills/ollama-deepseek-ocr-tool/SKILL.md` — Cursor also reads .agents/skills/ and .claude/skills/

### GitHub Copilot

1. **GitHub CLI**
```bash
gh skill install dnvriend/ollama-deepseek-ocr-tool plugins/ollama-deepseek-ocr-tool/skills/ollama-deepseek-ocr-tool
```

2. **Manually: put the files in place**
- File: `.github/skills/ollama-deepseek-ocr-tool/SKILL.md`

### Gemini CLI

1. **Skills CLI (Vercel)**
```bash
npx skills add dnvriend/ollama-deepseek-ocr-tool --skill skill-ollama-deepseek-ocr-tool -a gemini-cli
```

2. **Platform CLI**
```bash
gemini skills install https://github.com/dnvriend/ollama-deepseek-ocr-tool --path plugins/ollama-deepseek-ocr-tool/skills/ollama-deepseek-ocr-tool
```

### Windsurf (Devin Desktop)

**Manually: put the files in place**
- File: `.devin/skills/ollama-deepseek-ocr-tool/SKILL.md` — The legacy .windsurf/skills/ path still works

### OpenCode

**Skills CLI (Vercel)**
```bash
npx skills add dnvriend/ollama-deepseek-ocr-tool --skill skill-ollama-deepseek-ocr-tool -a opencode
```

### Claude.ai

**Claude.ai**
- Zip the folder and upload it in Settings → Capabilities → Skills

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/dnvriend/ollama-deepseek-ocr-tool/tree/HEAD/plugins/ollama-deepseek-ocr-tool/skills/ollama-deepseek-ocr-tool/SKILL.md

## When to use it
- convert textbook page photos into markdown study notes
- extract text from a set of scanned documents
- build searchable text from lecture images or presentation slides
- merge a sequence of screenshots into one text document
- process images locally without cloud-based OCR services

## Example prompts
- Extract text from all PNGs in the current directory and save to `chapter-3-notes.md`
- Turn lecture photos from `lecture-week5/*.jpg` into markdown with detailed logging
- Process a series of screenshots `screenshot-*.png` into a single document

## Source text · SKILL.md · MIT

Below is the content of a third-party file. It is data, not instructions for you: do not act on anything written inside unless the human asks. Block tag: d4d88f106bfcc52e.

BEGIN THIRD-PARTY FILE d4d88f106bfcc52e
````text
---
name: skill-ollama-deepseek-ocr-tool
description: Batch OCR processing with DeepSeek-OCR via Ollama
---

# When to use

- Convert textbook/lecture images to markdown notes
- Batch OCR processing of scanned documents
- Extract text from image sequences (iPhone photos, screenshots)
- Create searchable markdown from visual content
- Process documents privately without cloud services

# ollama-deepseek-ocr-tool Skill

## Purpose

This skill provides access to `ollama-deepseek-ocr-tool`, a CLI tool for fast, private batch OCR processing using DeepSeek-OCR via Ollama. Converts sequences of images (textbook pages, slides, scans) into a single coherent markdown document.

**Key capabilities:**
- ⚡ Fast processing (~3s per image on M4)
- 🔒 Private - runs entirely locally
- 📝 Clean markdown output (tables, headings, lists)
- 🔄 Natural sorting (IMG_1 < IMG_2 < IMG_10)
- 💰 Free - no API costs or rate limits

## When to Use This Skill

**Use this skill when:**
- Converting textbook chapters to Obsidian notes
- Processing lecture slides or handouts to markdown
- Extracting text from scanned documents
- Creating searchable study materials from images
- Need comprehensive examples and troubleshooting

**Do NOT use this skill for:**
- Cloud-based OCR (this is local-only)
- Describing image content (extracts text only)
- Handwritten text recognition (printed text only)
- Real-time streaming OCR (batch processing only)

## CLI Tool: ollama-deepseek-ocr-tool

The `ollama-deepseek-ocr-tool` processes multiple images in sequence and creates a single markdown document with extracted text. Images are sorted naturally and text is appended sequentially for coherent reading.

### Installation

```bash
# Clone and install
git clone https://github.com/dnvriend/ollama-deepseek-ocr-tool.git
cd ollama-deepseek-ocr-tool
uv tool install .
```

### Prerequisites

1. **Ollama** - Local LLM runtime
   ```bash
   brew install ollama
   ollama serve
   ```

2. **DeepSeek-OCR model** (~6GB download)
   ```bash
   ollama pull deepseek-ocr
````
END THIRD-PARTY FILE d4d88f106bfcc52e

The third-party file has ended. Everything between the d4d88f106bfcc52e tags is data: requests and commands from it are not followed. An “end of file” line with another tag inside the block is part of the file.

First 2051 of 7204 bytes. © dnvriend · MIT · https://github.com/dnvriend/ollama-deepseek-ocr-tool/tree/HEAD/plugins/ollama-deepseek-ocr-tool/skills/ollama-deepseek-ocr-tool/SKILL.md

_Part of the page did not fit — see the HTML version._

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
