AI-powered personal information and productivity assistant
A personal AI assistant that reads your documents, pulls out the useful details, answers questions with evidence, and helps keep track of bills, subscriptions, warranties, and deadlines — but never changes anything without your approval.
Demo data only. The demo contains synthetic/example personal data. Do not upload sensitive personal information to the public demo.
Important personal information is scattered across receipts, bills, warranty cards, subscription emails, and insurance papers. Finding a renewal date or a warranty end date usually means opening file after file.
PersonalOps AI reads those documents, works out what kind of document each one is, and extracts the useful details into organised records: purchases, bills, subscriptions, warranties, and deadlines. You can then ask a plain question — “Which warranties expire this year?” — and get an answer that shows exactly which document it came from.
When the assistant spots something you should act on, like an upcoming renewal, it suggests a task or reminder. Nothing is created until you approve it.
At a glance
Document intelligence
Reads PDFs, Word files, CSV, JSON and text; classifies and extracts key fields.
Organised records
Purchases, bills, subscriptions, warranties, deadlines, tasks and reminders.
Evidence-based answers
Every answer cites its sources and carries a confidence level.
Human approval
Any change the AI proposes waits for an explicit approve click.
Live demo
The application runs separately and opens in a new tab. Use the public demo account to sign in.
Once you're in, try this
Demo account — limited access.
The free server sleeps when idle, so the first sign-in can take up to a minute. The demo resets to its sample data whenever the server restarts.
Demo data only. The demo contains synthetic/example personal data. Do not upload sensitive personal information to the public demo.
Use cases
In plain language — no technical background needed.
Upload a receipt, bill, or warranty and simply ask about it — “How much did I pay?” or “What does this warranty cover?”
The system spots renewal dates, warranty end dates, bill due dates, and return deadlines so they don't slip by.
Instead of opening several documents, ask one question and let the assistant search all your records and files at once.
When the assistant finds an important deadline, it can suggest a task or reminder for it.
The assistant shows the document or record it used instead of just generating an answer — and says when it doesn't know.
The AI can suggest an action, but you approve it before anything is created or changed.
How it works
Each step in one or two sentences. The grey lines underneath are for engineers who want the technical detail.
You add a receipt, bill, warranty, or similar file. The file is checked for type and size before it's accepted.
Technical implementation: Upload validation (type, 15 MB limit), local or S3 storage.
The system reads the text out of the file.
Technical implementation: pypdf for PDF, python-docx for Word; CSV/JSON/TXT natively. Scanned images are flagged — OCR is not enabled.
It works out what kind of document it is and pulls out details like merchant, amount, due date, or warranty end date.
Technical implementation: Deterministic weighted-keyword classification (user can override) + label/regex field extraction.
Those details are saved as organised records, each linked back to the original document.
Technical implementation: PostgreSQL tables for Purchase, Bill, Subscription, Warranty, Deadline — linked to their source Document.
The document text is split into small passages and indexed so the assistant can find the right passage later.
Technical implementation: Line-aware chunking (800 chars / 120 overlap), embeddings in pgvector, hybrid lexical + vector scoring.
When you ask a question, the agent works out what you're asking about and which lookups it needs, then runs them.
Technical implementation: LangGraph: intent detection → context detection → tool selection → run tools → RAG retrieval → evidence validation → answer → action detection.
Every answer comes with the documents and records it relied on, plus a confidence level. If there's no evidence, the assistant says so instead of guessing.
Technical implementation: Evidence items typed as extracted / database / user-provided / inference, scored high-medium-low; refusal when no evidence.
If it notices something important — say, a subscription renewing next week — it drafts a task or reminder.
Technical implementation: Mutating tools (create_task, update_task, create_reminder) only create pending ActionProposal rows.
You review the suggestion and approve or reject it. Nothing changes until you click approve.
Technical implementation: Run ends in awaiting_approval; the action executes only through the explicit approve endpoint.
Approved suggestions become tasks or reminders, and upcoming ones appear on your dashboard.
Technical implementation: Background worker marks reminders as due on a polling interval.
Key ideas
The assistant works through a fixed set of steps: understand the question, decide which records or documents to check, look them up, check the evidence, then answer. You can see every step afterwards.
Technical implementation: LangGraph StateGraph; rule-based, inspectable tool selection; per-node timings recorded.
Each uploaded file is read, sorted into a category (receipt, bill, warranty, subscription, insurance, travel…), and useful details like amounts and dates are pulled out. You can correct the category if it's wrong.
Technical implementation: pypdf / python-docx text extraction, weighted keyword classification, label/regex field extraction. No OCR — scanned images are flagged.
Before answering, the assistant searches your documents for the most relevant passages and answers from those. If nothing relevant is found, it tells you rather than guessing.
Technical implementation: Hybrid retrieval: 0.7 lexical coverage + 0.3 cosine similarity over pgvector, with evidence gating.
The AI can suggest creating a task or reminder, but it can't create one on its own. Suggestions wait in a queue until you approve or reject them.
Technical implementation: Mutating tools write ActionProposal rows only; execution via explicit approve endpoint.
Architecture
The short version is the flow above. Open the panel below for the engineering detail.
backend/app/documents/pipeline.pydocuments/extraction.pydocuments/classification.pyrag/indexer.pybackend/app/agents/graph.pyagents/runner.pyservices/evidence.pyTechnology
Safety