✨ New
Free Tier
DocuChat now has a Free tier with no credit card required. You can build a chatbot, upload documents, and try the product end-to-end without committing to a plan. Signup requires email verification, with a dedicated landing page and a reliable post-verify redirect.
AI Credits Billing
The opaque "questions per month" metric has been replaced by a transparent AI Credits unit that maps directly to model usage. Per-token metering means light queries cost less and heavy queries cost more, with a clear ceiling on every paid tier and an opt-in toggle for allowing extra credits beyond your plan limit.
Smarter Document Ingestion
Document parsing has been rebuilt around a new ingestion engine that handles formats DocuChat previously struggled with:
- Scanned PDFs: Image-only PDFs are now indexed via OCR instead of failing silently or importing empty content.
- Tables: Tables in PDFs and Office files are extracted with structure intact, so cell-level questions retrieve the right row.
- Multi-Column Layouts: Reports and academic papers with multi-column layouts are parsed in correct reading order instead of jumbling columns.
- PowerPoint and Excel: PPTX and XLSX files are now ingested with the same quality as PDFs, not just text-stripped.
- Figures and Captions: Figure and table captions remain searchable even when image bytes are skipped.
Heading-Aware Chunking
Document headings are preserved as chunk metadata, so retrieval can surface the right section instead of fragmenting on arbitrary character boundaries. Combined with the new ingestion engine, this produces dramatically better answers on financial reports, manuals, regulatory documents, and other structured content.
Document Detail Page
Each document now has its own detail page with a file preview and a chunk-by-chunk inspector. Useful for understanding exactly what the retriever sees, and for debugging answer quality on a specific source.
Refreshed Design
A refreshed look and feel is rolling out across the app, starting with the chatbots list. Lighter, cleaner, and rebuilt on a consistent design system. The marketing site at docuchat.io has been fully rebuilt to match.
🛠️ Improved
- Trial Downgrade: Trial-expired users can now switch to the Free tier directly from the billing modal without contacting support.
- Documents List Refresh: The documents list refreshes immediately after uploads, deletes, and edits, instead of requiring a page reload.
- Resources Pages: Consistent page padding across Resources pages, and wider settings forms for better readability.
- Google Drive Folder Status: Folder sync status now reports correctly instead of falsely stamping folders as "Synced" when they were still in progress.
- Documents Pagination: A subtle note on the chunk list makes clear it isn't in reading order.
🐛 Fixed
- Mistral Models: Mistral chat models are working again after a regression caused by another library hijacking native fetch.
- Long Conversation Sessions: More reliable session subject generation on long conversations, with fewer transient failures.
- Verify Flow Edge Cases: A post-verify userData refresh failure no longer breaks the verification flow.