CATOCATO

Release Notes

Version history and changelog for CATO.

Current Version:v1.3.0(March 2026)
v1.3.0MINOR
March 2026

Highlights

  • No-code Figure Editor: edit, annotate, and version your visualizations directly in chat
  • 40+ supported file formats: FASTA, VCF, H5AD, PDB, NIfTI, mzML, and more for genomics, single-cell, imaging, and proteomics
  • New analysis libraries: deepTools (epigenomics), CausalML (causal inference), EpiScanpy (scATAC-seq), decoupler (pathway activity), and more
  • Conversation sharing with public read-only links
  • Smarter interview prompts: CATO now asks clarifying questions before generating documents, presentations, and spreadsheets
  • Evidence Matrix adds sortable Relevance column with auto-sort by score
  • Faster code execution: sandbox sessions are reused within a conversation, eliminating startup overhead on follow-up runs

Features

  • +No-code Figure Editor: click the pencil icon on any visualization to open the SVG editor. Select, reposition, and style elements; add text annotations and significance brackets; undo/redo up to 50 steps; export as PNG or SVG. Available on Base plans and above
  • +Figure version history: edits create numbered versions with optional summaries. Browse, compare, and revert to any previous version via the version badge on each figure
  • +Figure properties panel: edit text content, font, color (with Okabe-Ito colorblind-friendly palette), stroke, opacity, and fill for selected elements
  • +3D Molecular Viewer: large PDB structures (>32KB) are now chunked for reliable streaming, with a loading spinner while data assembles. Previously, large structures could fail to render
  • +Conversation sharing: generate a shareable link to any chat session. Recipients see a read-only view of the full conversation including figures and file artifacts. Revoke links anytime
  • +CATO now explicitly states key assumptions after every analysis, including method choice, missing-data handling, grouping, and thresholds
  • +Chat always selects the best available model for your subscription tier (previously remembered manual selection across sessions)
  • +Generated files (AI outputs via save_to_project) are now hidden from the Database view to reduce clutter
  • +CATO recognizes user-edited figures and always uses the edited version when embedding into reports or follow-up analyses
  • +SVG output generated alongside PNG for all matplotlib figures, with role-tagged elements enabling the figure editor
  • +Peer review phases now display as structured containers (Search & Plan, Execute, Verify, Revise) in the chat journal
  • +Chat error recovery: render errors are caught gracefully with a "Try again" button instead of a white screen
  • +Interview guidance improved: CATO proactively asks about structure, content, and style before generating DOCX, PPTX, and XLSX files
  • +Evidence Matrix now includes a Relevance column showing the composite relevance score for each paper, sortable and auto-sorted descending by default
  • +Conversation-scoped sandbox reuse: subsequent code executions in the same chat reuse the same sandbox environment, significantly reducing latency
  • +Visualization improvements: annotations now use data-relative offsets (no more overlapping labels), x-axis tick labels always show meaningful category names
  • +Bullet list rendering in chat: consecutive bullet lists separated by blank lines are now properly merged, matching the existing behavior for numbered lists
  • +Literature search errors (budget limits, usage caps) now display the upgrade prompt dialog instead of raw error messages
  • +Systematic review generation shows user-friendly error messages when insufficient evidence is available
  • +Literature inspector panel toggle now correctly switches between inspector and report modes without requiring a double-click
  • +More accurate cost tracking that accounts for cached prompt tokens, reducing reported costs for repeated interactions
  • +Expanded file format support: 40+ formats including FASTA, FASTQ, VCF, GFF/GTF, BED, H5AD, PDB, CIF, SDF, NIfTI, DICOM, mzML, MGF, Feather, MATLAB, NumPy, GML/GraphML, and more
  • +New pre-installed analysis libraries: pysam (BAM/SAM), deepTools (ChIP-seq), EpiScanpy (scATAC-seq), CausalML and CausalLib (causal inference), decoupler (pathway activity), pmdarima (time series), MSPypeline (proteomics)
  • +Figure version selector: click the version badge on any edited figure to browse and switch between all saved versions, including the original. The selected version is used for downstream tasks like report generation
  • +Usage display auto-refreshes after each chat interaction completes, keeping the billing page current without manual refresh
  • +Figure editor production hardening: unsaved changes warning on close, request timeouts, double-save prevention, and bounded undo history

Bug Fixes

  • ~Fixed scatter plot dots and marker elements disappearing in the figure editor due to SVG sanitization stripping use elements
  • ~Fixed clicking Edit on any figure always opening the first figure instead of the correct one
  • ~Fixed 3D molecular viewer showing infinite loading spinner when no structure data was available
  • ~Fixed dataset and folder deletion using browser confirm() instead of proper in-app confirmation dialog
  • ~Fixed notification toasts potentially sharing duplicate IDs when triggered in rapid succession
  • ~Fixed upgrade prompt buttons remaining disabled for non-loading plans on the pricing page
  • ~Fixed dataset resolution failing when the AI references a bare filename or dataset ID without the full path
  • ~Fixed sandbox file uploads and code execution occasionally timing out due to stdin not receiving EOF signal
  • ~Fixed orphaned sandbox environments accumulating after server restarts — now cleaned up automatically on startup
  • ~Fixed duplicate token usage records being saved on retry/replay
v1.2.2PATCH
February 2026

Highlights

  • AI responses continue generating in the background when you navigate away from chat, and reconnect automatically on return
  • Interview questions now clear from the UI immediately after you answer them, including after reconnect
  • Streaming indicator appears as soon as background generation is detected, before reconnect completes
  • Improved guidance when analysis code produces no output files

Features

  • +Background stream persistence: response generation continues as a background task when you leave the chat. Returning to the conversation reconnects to the live stream and replays any buffered events
  • +Streaming UI (cancel button, loading indicator) activates immediately when background generation is detected, before the reconnect request completes
  • +interview_complete event: interview questions clear from the chat UI the moment answers are submitted, preventing them from reappearing after stream reconnect or page navigation
  • +save_to_project error messages now explain when code produced no output files and guide the model to write files explicitly rather than print to stdout
  • +Graceful server shutdown now cancels deferred stream cleanup tasks, preventing background task leaks on restart

Bug Fixes

  • ~Fixed interview questions reappearing after navigating away from a conversation where the interview was already completed
  • ~Fixed hydration mismatch causing user avatar to flash a fallback initial on first page render
  • ~Fixed conversation guard locks not releasing within the expected window in certain race conditions
  • ~Fixed deferred stream cleanup tasks not being tracked for cancellation on server shutdown
v1.2.1MINOR
February 2026

Highlights

  • Study type filters with accurate word-boundary matching and server-side PubMed filtering
  • ClinicalTrials.gov and Patents connectors hardened with automatic retry and increased results
  • Send-to-Chat workflow improved with prompt templates and context preservation
  • Tour state persistence hardened with retry logic and cross-device sync

Features

  • +Study type filters expanded: added Systematic Review and Case Report to Query Studio
  • +Study type matching upgraded from substring to word-boundary regex — "RCT" no longer false-matches "extraction", "review" no longer matches "reviewed"
  • +PubMed Publication Type injection: selecting study types (RCT, Meta-analysis, etc.) now adds native PubMed [PT] field tags for server-side filtering
  • +Study type context passed to SearchPlanner agent for smarter query construction across all connectors
  • +ClinicalTrials.gov connector: automatic retry on HTTP 429/5xx with backoff, increased default results from 20 to 50
  • +Patents connector (USPTO): automatic retry on transient failures with backoff, increased default results from 20 to 50
  • +arXiv connector: added retry logic and increased timeout for improved reliability
  • +Literature-to-Chat bridge: prompt type selection (Ask, Analyze, Draft, Plan) with template pre-fill in chat input
  • +AI budget system: dollar-based monthly limits with percentage-based usage display and reservation-based concurrency control
  • +Systematic Review feature restricted to Pro tier and above
  • +Terms of Service acceptance tracking with versioned re-acceptance on policy updates
  • +Network graph: visual indicator for pinned nodes in force-directed layouts
  • +Account management: soft deactivation and hard-delete with full data anonymization
  • +System prompt caching for reduced latency on repeated interactions
  • +Geographic term registry for optimized location-based PubMed searches
  • +Percentile-based search quality filtering for evidence-first mode

Bug Fixes

  • ~Fixed Send-to-Chat race condition: chat input disabled until literature packet loads, preventing empty sends
  • ~Fixed literature context cleared before confirming message delivery — context now preserved on send failures
  • ~Fixed stale literature context leaking into New Chat sessions
  • ~Fixed tour re-appearing for returning users: sync now retries 3 times with backoff and stashes in localStorage for next page load
  • ~Fixed tour sync not sending welcome_shown flag, causing welcome animation replay on cross-device login
  • ~Fixed HTTP/2 protocol errors during literature search streaming via SSE result batching
  • ~Fixed Excel .xls upload crashes by adding xlrd dependency
  • ~Fixed token limit validation in API endpoints
  • ~Fixed interview tool race condition in chat system
  • ~Fixed React hydration mismatches in tour and landing page components
  • ~Fixed markdown list rendering in chat responses
v1.2.0MAJOR
February 2026

Highlights

  • Literature Intelligence module with multi-source academic search
  • Citation network analysis with interactive graph visualization
  • Evidence extraction and structured matrix view
  • AI-powered document generation (briefs, reviews, gap analyses)
  • Literature-to-Chat bridge for seamless research workflows

Features

  • +Query Studio: natural-language literature search with evidence-first and novelty-first modes
  • +Multi-source search across PubMed, BioRxiv, arXiv, ClinicalTrials.gov, Patents, and OpenAlex
  • +Results grid with relevance scoring, abstract previews, and bulk selection
  • +Evidence Matrix: structured extraction of PICO elements, outcomes, effect sizes, and bias indicators across studies
  • +Citation Network: interactive force-directed graph with PageRank, betweenness centrality, and Louvain community detection
  • +Contradiction detection via TF-IDF cosine similarity analysis across paper conclusions
  • +Research gap identification from network topology analysis
  • +AI document generation: systematic review briefs, landscape analyses, and gap reports
  • +Library management with collections, annotations, and full-text reader panel
  • +Watch alerts for automated monitoring of new publications matching saved queries
  • +Literature-to-Chat bridge: send curated paper sets to CATO Chat with quick-action buttons (Summarize, Find conflicts, Propose next steps, Export brief)
  • +Run Console for monitoring search agent progress and token usage in real time
  • +5 specialized literature AI agents: search planner, evidence extractor, graph builder, brief writer, QA sentinel
  • +Work Inspector panel for viewing detailed metadata, references, and MeSH terms
  • +Enhanced PubMed connector with abstract fetching and up to 10,000 result support
  • +BioRxiv connector with TF-IDF relevance ranking
  • +Network analysis service with PageRank, community detection, and hub/authority scoring

Bug Fixes

  • ~Fixed large file uploads by raising Next.js body size limit to 2 GB
  • ~Fixed TypeScript narrowing for MessageContent union in queue_status handler
  • ~Fixed const reassignment in queue_status handler
v1.1.1PATCH
February 2026

Highlights

  • Cross-browser tour and welcome state persistence
  • Account management with deactivation and deletion
  • Support ticket email notifications via SMTP
  • AI model tier selection (Lite, Pro, Ultra)

Features

  • +Tour/welcome state synced to backend database — persists across browsers and devices
  • +Account deactivation (soft) and hard-delete with full data anonymization and Stripe cleanup
  • +Support ticket email notifications with configurable SMTP (Gmail, Outlook, etc.)
  • +AI model tiers: Lite, Pro, and Ultra — selectable per subscription plan
  • +Plan mode refactored to standalone toggle button in chat input
  • +Masonry grid layout for project cards with independent column expansion

Bug Fixes

  • ~Fixed follow-up suggestions not appearing after analysis completion (SSE stream loop exiting early)
  • ~Fixed "Writing Response" indicator staying visible during follow-up generation
  • ~Fixed tour/welcome animation replaying on new browsers despite prior completion
  • ~Fixed project card hover expanding all cards in the same grid row
  • ~Fixed select dropdown styling inconsistencies
v1.1.0MINOR
February 2026

Highlights

  • Biomedical tools integration (200+ tools across 20+ categories)
  • Dataset @-mention autocomplete in chat input
  • Document generation: DOCX reports, PPTX presentations, and XLSX spreadsheets
  • Visualization side panel with gallery view and auto-open
  • R language support alongside Python

Features

  • +Biomedical tools: gene/protein lookup, pathway analysis, drug interactions, clinical trial search, sequence analysis, ontology queries, disease associations
  • +Literature search integration with PubMed and BioRxiv
  • +@-mention datasets in chat with real-time autocomplete filtering
  • +Slash commands: /profile, /visualize, /clean, /test, /code, /export, /help
  • +Plan mode (Shift+Tab) for step-by-step analysis planning before execution
  • +DOCX report generation with title pages, headings, tables, embedded figures, and page numbers
  • +PPTX presentation generation with custom slides, tables, and embedded charts
  • +XLSX spreadsheet generation with multi-sheet exports, styling, auto-filters, and freeze panes
  • +File artifact persistence — CSV, Excel, DOCX, PPTX files saved with chat history
  • +Visualization panel with auto-open, gallery view, click-to-navigate, and download
  • +R language support with tidyverse, survival, lme4, DESeq2, Seurat, and more
  • +ZIP archive upload for batch importing multiple datasets
  • +Follow-up suggestions after analysis completion
  • +Interview system for clarifying complex requests
  • +Memory system for retaining context across interactions
  • +TSV and Parquet file format support
  • +Color-coded dataset labels for organization

Bug Fixes

  • ~Fixed SSE event type detection using proper event: line parsing instead of data heuristics
  • ~Fixed visualization panel not displaying generated figures
  • ~Fixed file artifacts not persisting to database on page reload
  • ~Improved viz panel auto-open to trigger on every new figure, not just the first
v1.0.0MAJOR
January 2026

Highlights

  • Initial public release of CATO
  • Database management with CSV, Excel, and JSON support
  • CATO Chat AI assistant with code execution sandbox
  • Project organization with labels and descriptions

Features

  • +Upload and manage research datasets
  • +AI-powered natural language data queries
  • +Python code execution in secure sandbox
  • +Automatic data type inference
  • +Session-based chat history with persistence
  • +Dark/light mode support
  • +Keyboard shortcuts
  • +Admin approval workflow for new accounts
v0.9.0BETA
December 2025

Highlights

  • Beta release for internal testing

Features

  • +Core project management functionality
  • +Basic data upload pipeline
  • +Initial chat interface

Bug Fixes

  • ~Fixed file encoding issues with non-ASCII characters
  • ~Resolved session timeout errors
For older versions, contact support.