Build a searchable knowledge library from multiple documents. Auto-tags, topic clustering, and semantic search across all your files — 100% local.
This tool runs 100% in your browser. All computation happens locally on your device — your input is never uploaded to any server. Results are for reference only.
This tool turns a pile of documents into a searchable library entirely on your device. Uploaded files are parsed with PDF.js and split into text chunks, then each chunk is converted into a TF-IDF vector — a numeric fingerprint of how important each word is to that passage. When you type a query, the same vectorizer scores the query and every chunk with cosine similarity, ranking the most relevant passages on top. In Q&A mode the top-ranked chunks are handed to a local language model, which reads them and writes a synthesized answer with source citations. Because the vector store and model both live in your browser, your documents never leave the tab.
Documents are chunked and TF-IDF vectorized, a query is scored by cosine similarity against every chunk, and the top matches feed either a ranked search or a local LLM for a cited answer.
Nina, a product analyst, uploads three PDFs — a Q2 sales report, a marketing memo, and her own interview notes. The tool parses them into 42 chunks and builds a local TF-IDF index in about a second.
To build a knowledge base: add your notes, documents, and links — the AI automatically tags and categorizes them for smart search and retrieval, all stored locally.
FreeToolHub AI Knowledge Base is a free browser-based tool that organizes notes and documents with AI-powered tagging and search, no signup.
Build your private knowledge base. Upload PDFs and search with TF-IDF semantic similarity. 100% local, no upload, no signup, free.
The Personal Knowledge Base turns a scattered pile of documents into one searchable library, all inside your browser. You upload PDFs, TXT, CSV, or Markdown files; each document is parsed, split into chunks, and indexed using TF-IDF vectorization. When you search, chunks are ranked by cosine similarity, so you get the most relevant passages — with the document name and page number attached — instead of a folder full of files you must open one by one. A dashboard tracks how many documents and text chunks you have indexed, and the library auto-tags your files and groups them into topic clusters. Nothing is uploaded to a server at any point.
Graduate students drowning in papers use it to find the one methodology paragraph they remember reading months ago. Consultants working under NDA appreciate that confidential client PDFs can never touch a third-party cloud, which most hosted search services require. Paralegals and small-firm lawyers index case files and contracts, then pull every clause mentioning a specific term in seconds. Technical writers and bloggers keep a library of source material and verify quotes against the original page. Even casual users benefit: upload a handful of manuals, insurance policies, and tax documents once, and every future question becomes a one-line search instead of an afternoon of scrolling.
(1) Upload one or many files — PDF, TXT, CSV, or Markdown. Each is parsed and chunked locally, and the dashboard's document and chunk counters confirm what has been indexed. (2) Choose a mode: Search returns the ten best-matching passages ranked by relevance score, or Q&A has a local language model read the top excerpts and write a synthesized answer. (3) Type your query and press search. In Search mode you see passages with document name, page number, and similarity score; in Q&A mode you get a concise answer with numbered citations like [1] and [2] pointing back to the sources. Copy any result for your notes, and remove documents individually whenever you like.
Search mode works instantly, with no download, and behaves like a precise keyword-plus-relevance engine: perfect for fact-finding, quote hunting, or comparing how several documents treat the same term. Q&A mode goes further — it takes the top five matching excerpts and has an on-device language model compose a direct answer, citing each source by number. That model requires a one-time download of about 200MB, after which questions are answered without any network call. A practical rhythm: start in Q&A mode for questions like 'what termination notice does this lease require,' and drop into Search mode when you need to read the full surrounding paragraph yourself. Both modes read only from your device.
Everything — parsing, chunking, embedding, search, and answer generation — runs in your browser or on your machine. No document ever leaves your device, which makes it suitable for contracts, medical notes, internal docs, and anything covered by confidentiality obligations. Cloud RAG offers bigger models and shared access, but requires uploading your corpus to someone else's servers; local wins whenever privacy or offline availability is the priority.
PDFs and plain-text formats (txt, md, csv) are supported — documents are parsed, chunked, and indexed locally on import. Scanned PDFs without a text layer need OCR first. Very large corpora are limited by browser memory: hundreds of pages work comfortably, entire document archives may require splitting into separate sessions by topic.
Keyword search matches literal strings — searching 'refund' finds only documents containing that exact word. Semantic search embeds chunks into vectors and matches meaning, so 'refund' also surfaces passages about 'money back', 'cancellation credit', or 'return policy'. The QA mode goes further: it retrieves the most relevant chunks and generates a synthesized answer grounded in them, with the source chunks traceable.
Yes — this is one of its strongest cases. Import a set of papers, then ask cross-document questions ('which of these studies report effect sizes?', 'what methodologies do they share?') and get answers with the relevant passages identified. The tag and cluster views help organize a literature review: related papers group together, making gaps and themes visible across the collection.
ChatGPT and NotebookLM upload your documents to cloud servers for processing. This tool indexes and queries everything locally in your browser—no data leaves your device. It functions as a private, offline alternative: upload multiple PDFs and text files, then ask cross-document questions with answers cited to specific sources.
You can add multiple PDFs and text files to a single knowledge base. All indexing (chunking, embedding, vector storage) happens in-browser using WebGPU. Practical limits depend on device RAM: 10-20 documents totaling 500+ pages work well on a 16GB machine. The index persists in IndexedDB for instant reloading in future sessions.
Why did the cow cross the road?
No signups, no data sold. The core of every tool is free forever — the optional Pro plan adds batch processing, unlimited downloads, white-label exports and an ad-free experience.
☕Support me on Ko-fi— keep tools free100% of proceeds go towards hosting & building more free tools.