Transcribe PDFs to Markdown using whichever vision-language model is loaded in LM Studio (tested with Qwen3-VL). Each page is rasterized locally and read by the VL model — no OCR install, no Python, no cloud.

Read and write .docx (Word) files locally as markdown — structural conversion via mammoth (read) and the docx library (write). No OCR, no cloud. Pairs with read-pdf for end-to-end document workflows.

Démo pédagogique de la dissertation de philosophie du Bac sous forme de livre dont vous êtes le héros. Narrateur LLM local corpus fermé de citations attestées par ≥2 sources indépendantes vérificateur anti-hallucination en temps réel mécanique ...

PII redaction with two complementary layers. Regex checksums catch structured PII (credit cards via Luhn, IBAN/NIR via mod-97, phones, emails, passport, addresses). An optional ML layer (multilingual XLM-RoBERTa NER) detects names, dates and unstruct...

Sandboxed filesystem access (read, write, list, grep) for your local model. Per-chat allowed root paths, two-step write confirmation with content-bound tokens, automatic backups of overwritten files, and full protection against .. traversal and symli...

Give your local model web search. DuckDuckGo backend works out of the box (no API key); SearXNG backend lets you point at your own privacy-respecting metasearch instance (Google Bing Brave DuckDuckGo combined).

Provides tools to get the current date and time in any IANA timezone, with configurable defaults and output format.