ollama pdf2markdown.stub

---

id: ollama-pdf2markdown name: OllamaPDF2Markdown repo_path: ~/Documents/sw30labs/repos/OllamaPDF2Markdown remote_url: https://github.com/sw30labs/OllamaPDF2Markdown.git primary_language: Python framework: Ollama + pdf2image + Mistral Small 3.1 24B ingested: 2026-04-13 last_commit_date: 2025-04-15

Purpose

Converts PDF files to Markdown by extracting pages as images and processing them through multimodal vision-language models locally via Ollama. Provides both GUI (Streamlit) and CLI interfaces with support for Mistral Small 3.1 24B and Llama 3.2 models.

Architecture

• Two-interface approach: Streamlit GUI for end-users, Python CLI for automation/batch
• PDF → Image conversion: pdf2image to JPEG with configurable quality
• Multimodal vision-language processing: Ollama integration for local inference
• Streaming and memory management: Explicit garbage collection between images, configurable delays
• Progress tracking: Real-time progress indicators with processing time per image
• Intermediate saving: Preserves progress every 5 images (CLI version)
• Batch processing: Handles multiple PDFs with maintained order

Dependencies

• ollama - local LLM serving
• pdf2image - PDF to image conversion
• Pillow - image manipulation
• streamlit - GUI framework
• Python stdlib (gc, logging, os, subprocess)

Key Features

• Local PDF processing (private, no cloud API)
• Multimodal model support (Mistral Small 3.1 24B, Llama 3.2)
• 128k context window (Mistral) for large documents
• Memory-efficient processing with GC and cleanup
• Progress tracking and logging
• Drag-and-drop GUI interface
• Batch CLI processing
• Configurable model and output directories
• Error handling with graceful recovery

Tags

• python
• ocr
• local-inference
• ollama
• cli
• streamlit
• developer-tools

Entities and Topics

• developer-tools
• local-inference-mlx
• deepseekvl2-pdf-ocr