The confidential pro — « My files never leave my workstation. »
Lawyer, doctor, HR professional, consultant, CPA, notary—you handle documents covered by professional secrecy or an NDA. A local LLM paired with RAG lets you query, summarize, compare, and cross-check them without ever uploading anything. And without the command line.
By the end, LM Studio will run with a polished graphical interface (no terminal). A 13–24B model will be loaded. A folder of PDFs will be indexed through AnythingLLM or the built-in Documents feature. You’ll be able to ask your archives questions in French, get exact citations with page numbers, and export a summary—without a single byte leaving your machine.
Who it's for this path is for you
We'd rather tell you early than let you waste 30 minutes for nothing.
- ✓Your documents are covered by professional confidentiality (lawyer, doctor, notary, certified public accountant).
- ✓You work under an NDA, and cloud tools are contractually prohibited.
- ✓You handle personal data subject to GDPR—avoiding transfers outside the EU is imperative.
- ✓You want a clean, clear UI—not a command in a terminal.
- ✓You have 100 to 10,000 PDFs to query without re-uploading them every time.
- ·Your data is public and you just want an assistant: a cloud service will be simpler.
- ·You manage terabytes of documents: you need a server deployment, not a workstation.
- ·You want real-time processing on hundreds of documents per second: a simple RAG pipeline won't hold up; see vLLM or TGI.
The path in 5 steps
Each step points to a detailed guide you can read alongside it. The order is optimized: don’t skip a step the first time.
- 1Step 01·5 min·Beginner
Install LM Studio
Standard installer for Windows, macOS, or Linux. Full graphical interface: search, model downloads, chat, settings — without ever opening a terminal.
LM Studio is installed; you can see the model gallery.Read the guide → - 2Step 02·8 min·Intermediate
Choose a 13–24B model
Mistral Small 24B in Q4 if you have 16 GB of VRAM. Qwen2.5 14B in Q4 if you have 12 GB. Direct download in the app, with automatic integrity verification.
A robust model for text analysis is loaded into memory.Read the guide → - 3Step 03·20 min·Beginner
Index your documents
AnythingLLM or Open WebUI for the RAG component. Point it to a folder, choose a French embeddings model (Solon, BGE-M3), and start indexing. Everything stays on disk.
Your PDFs are indexed in a local vector database (ChromaDB).Read the guide → - 4Step 04·8 min·Beginner
Verify network isolation
A checklist: Wireshark to confirm that nothing leaves, disabling any telemetry, encrypting the model directory, and creating an encrypted local backup.
Verified trust: zero outgoing packets during use.Read the guide → - +Step bonus·15 min·Advanced
Bonus — Complete air gap
For the most sensitive files: a firewall rule that blocks all outbound traffic except system auto-updates. Belt and suspenders.
Offline-certified machine for sensitive sessions.Read the guide →
Recommended models
Three choices tested for this path—from lightest to most capable. Click for the detailed specifications (VRAM, context, benchmarks).
The right professional default. Excellent in French, for summarizing long texts, and for legal reasoning.
View details →Lighter (9 GB in Q4), very good at structured analysis. For more modest machines.
For 64 GB Macs or systems with 48 GB of VRAM. Local "2023 GPT-4" quality.
Frequently asked questions
The real questions we get by email and on Mastodon. If yours is missing, open a ticket.
The curious one — « I just want to try it, hassle-free. »
Vous avez entendu parler des LLM locaux et vous voulez voir ce que ça donne sur votre machine. Aucun code, aucune ligne de commande compliquée — en 15 minutes vous chattez avec vot…
Your documents, your AI: a reliable local RAG over your PDFs, notes and mail — nothing leaves your machine.
- Lifetime online access
- PDF + files
- Lifetime updates
A question, a typo, a bug?
This guide evolves with every model release. Your feedback is the raw material.