An AI document Q&A tool that lets you upload PDFs, text files, or website links to build datasets, then ask questions and get answers with source citations (document name and page number). It offers adjustable model/context size/temperature, custom system prompts to shape the AI’s personality, Telegram bot integration, Q&A export, public dataset sharing, API access from the Scale tier, and Docker self-hosting on request. Pricing runs from a free trial (20 queries, 900 KB upload) to Casual at $9/month, Scale at $49/month with API access, and Enterprise at $399/month. Current status is confirmed live (HTTP 200, Cloudflare, Next.js). Note: no permanent free tier, no multimodal support (images, scanned PDFs, or table analysis are not handled), and community validation is thin—no independent benchmarks or deep reviews surfaced.
**Pros & Cons**:
Pros:
Answers include source citations down to document name and page number, making verification straightforward; adjustable model, context size, and temperature give users real control over answer style and depth; custom system prompts let teams shape the AI’s persona for specific workflows; Telegram bot integration extends access beyond the web UI; Docker self-hosting is available for teams that need data control; public dataset sharing supports collaborative knowledge bases; confirmed live and responding as of listing.
Cons: No permanent free tier—only a 20-query trial before payment; no multimodal support, so scanned PDFs, images, and table-heavy documents are not handled; community validation is very thin, with only six Product Hunt mentions and no independent performance benchmarks; API access requires the $49/month Scale tier, leaving the $9 Casual tier without programmatic access; the jump from $49 to $399/month leaves no mid-tier option; no public changelog or detailed docs, which some reviewers flag as a transparency gap.
**Who is it for?**:Teams and individuals who need grounded answers from their own documents with verifiable citations; users who want to tune answer style via system prompts and model parameters; teams that prefer Telegram-based access or self-hosted Docker deployment; organizations building shared public or internal knowledge datasets.