CELESTOS
CELESTOS
Book, lecture recording, contract PDF, folder of scattered notes, audio from a meeting, image of a whiteboard, spreadsheet of sales data — drop any of it into Celestos and it becomes part of your searchable, connected knowledge graph. Extracted, indexed, semantically linked to everything you already know. Not just stored. Understood.
Drop a file, paste a link, or point Celestos at a folder. Everything below is extracted, indexed, and connected to your memory graph — privately, on your machine, with no cloud processing required.
Every book you feed Celestos becomes a searchable node in your knowledge galaxy. Not just the file — the ideas. Chapters link to related memories. Highlights connect to sessions where you discussed the same concept. And when you ask Celestos something, it draws from your books the same way it draws from your conversations.
Drop any audio or video and Celestos transcribes it locally using Whisper — the same model that powers professional transcription, running entirely on your machine. The transcript becomes a searchable node in your knowledge graph. Ask what was said. Find when a topic came up. Extract the key moments. Build flashcards from a lecture. Turn a meeting into tasks grouped by person.
Drop a recording or YouTube link. Celestos transcribes it, then generates study notes, flashcards, and a quiz — all from your actual lecture, not a generic summary.
Drop the recording. Get back a summary, action items, and tasks grouped by who said them. Your meeting is done before you've finished your coffee.
Pull the key moments. Extract quotes. Make short clips from the strongest parts. Find every mention of a topic with timestamp screenshots.
Photo of a whiteboard, scanned page, screenshot of a doc — Celestos OCRs it with Tesseract and makes the text searchable. Nothing stays locked in an image.
Every workflow below has been verified against the live Celestos backend with real files. These aren't demos — they're things you can do right now.
Every word extracted from your files is searchable from Orb, Chat, or the memory galaxy. "Find all my documents that mention X" — it works.
Celestos draws edges between your ingested files and related sessions — even when the words don't overlap. A book about habits connects to your productivity conversations automatically.
Ingested content becomes memory nodes in your galaxy — orbiting the concept planet that fits. Your books, your lectures, your meetings — all visible, flyable, navigable.
Once something is ingested, Celestos uses it to answer your questions — not from its training data, from your actual files. "What did that contract say about termination?" asks your document, not the internet.
Any ingested topic can be turned into a Memory Podcast. "Make a podcast episode from everything I've ingested about machine learning" — script written by your local model, narrated by local TTS voices.
All extraction, OCR, transcription, and indexing runs on your machine. Whisper, Tesseract, qpdf, DuckDB — every tool runs locally. Your files never reach a cloud service.
Celestos ships with 22 built-in CLI tools already wired up and ready. You describe what you want in plain language — Celestos picks the right tool and runs it. You never touch a terminal.
Audio & video transcription. Professional quality. Runs fully on-device.
Image OCR. Extracts text from photos, scans, and screenshots.
PDF text extraction, page splitting, merging, and inspection.
CSV and spreadsheet analysis. Query your data in plain language.
Document conversion — Markdown to DOCX, HTML to Markdown, and more.
E-book ingestion and format conversion. Drop any book format.
Video & audio processing — trim, convert, extract audio, resize.
Metadata inspection and stripping for images and PDFs.