Ask questions about an uploaded document in plain English and get answers grounded in its actual contents. Supports follow-up questions and comparisons across sections.
Pro AI tools send document content to xAI.
Chat with Doc is a Pro feature — included in PDFwarp Pro at $9 USD/month along with every other tool and unlimited conversions.
The file is extracted once on the first turn (~12,000 character cap) and the text is sent to xAI with your question. Later turns reuse that extracted text plus the in-session message history; the file is not re-uploaded. The model is told to answer from the document and to say so when the answer is not there. Pro only.
'What's the termination clause?' gives a specific, verifiable answer. 'Tell me about this document' gets a summary. Specific questions produce the most useful responses.
After asking about a clause, follow-ups like 'What's the notice period?' use your earlier question as context. You don't need to re-explain what document you're asking about.
AI can be confidently wrong on dates, dollar amounts, and legal specifics — especially near document edges or in OCR'd scans. For anything you'll act on, cross-check against the source.
The route is not wrapped in the usual file decorator because the upload is optional after the first message. Auth, Pro plan, and the 30-call daily AI cap are checked by hand. Size still uses the Pro 500 MB ceiling.
The chat helper puts the extracted text in a system prefix and sets prompt_cache_key so follow-ups route to the same xAI server and hit the automatic prefix cache. History is the JSON array the browser sends.
The system prompt says to answer from the document and to admit when the answer is not there. That is a prompt, not a guarantee — still verify numbers before you act on them.
Anything about the extracted text — key facts, dates, figures, clauses, summaries of specific sections. If it is not in those ~12,000 characters, the model should say so.
Yes, within this session. History is posted with each turn from the browser. Close the tab and it is gone.
Text-based PDFs, Word, and .txt. Scanned pages need AI Scan to Text first.
No chat log is written to our database. The first upload counts as an AI call; follow-ups still go to xAI with the cached document text.