A timeline of what we have shipped.

Added AI-assisted search generation to help teams build broad evidence-discovery strategies more quickly.

Added hit retrieval and article-detail collection as part of the same guided flow.

Introduced article scoring and ranked selection so teams can focus writing on the most relevant evidence set.

Added title-led report structure generation tailored for scoping-review style outputs.

Added evidence-table generation so extracted study details can feed directly into downstream writing.

Added AI-assisted writing with study-aware citations and preview-friendly output formatting.

Added per-step credit estimates so users can understand the expected cost before running major stages.

Added clearer insufficient-credit messaging to reduce confusion during longer review workflows.

Kept ranked evidence visibility in writing previews so teams can trace selected studies more easily.

Built a focused AQUA interface for upload history, processing visibility, and result retrieval.

Simplified the main AQUA page to center on the current user’s own files and work history.

Added clearer processing states for pending, processing, completed, and failed results.

Added backend-proxied downloads for both original and processed AQUA files.

Enabled AQUA file pairs to appear inside the shared file-management experience.

Added safer file serving and cleaner download behavior for result delivery.

Added AI recommendations for each uploaded file, including suggested assessment type and scale.

Added support for multiple Risk of Bias checklists and frameworks, including:

Allowed users to review and edit AI recommendations before final processing.

Saved user selections immediately so review work is preserved across refreshes and handoffs.

Added grouped processing by assessment scale so similar files can be analyzed together.

Generated final result ZIP packages for download after processing completes.

Added input ZIP and result ZIP downloads to simplify file handoff and auditability.

Added unified ROB status states for uploaded, recommending, ready, processing, processed, and failed stages.

Added project status polling and processing estimates so users can monitor longer-running jobs.

Enabled guarded company-level access for administrators and superadmins.

Added AI-assisted report structure proposal with editable taxonomy, section planning, and reference-aware organization.

Added section-to-study mapping and table planning to support richer scientific report assembly.

Made it easier for teams to review, confirm, and refine the structure before full writing begins.

Added asynchronous writing jobs for longer-running report generation flows.

Created writing payloads that combine project context, search strategy, PRISMA inputs, and included-study evidence.

Added progress handling and report-generation staging for propose, write, and export steps.

Added DOCX export using a report template with headings, lists, table captions, and structured appendix rendering.

Supported appendix search strategy tables and export-friendly formatting for polished delivery outputs.

Added fallback export behavior so teams can still download reports when the preferred path is unavailable.

Added manual extraction-sheet design for columns, sheets, and extraction structures.

Added AI template suggestion so teams can generate draft extraction templates from uploaded study text.

Enabled inline editing, sheet reordering, column management, and auto-save while building templates.

Added background content extraction with session tracking and real-time file status polling.

Added batch processing with parallel workers for scalable extraction across larger projects.

Added retry handling for failed text extraction and better project-level progress visibility.

Saved extraction results as structured JSON for project reuse and review.

Added frontend export generation for Excel and Word outputs.

Added project results views, batch history visibility, and incremental extraction support for new files and new sheets.

Introduced multiple screening modes including AI Screener, AI Assistant, and manual screening.

Added batch AI processing for large study sets with progress visibility and failure handling.

Combined AI support with human review so teams can accelerate screening without losing oversight.

Added support for both FPS (First-Pass Screening) and FTR (Full-Text Review) as part of the screening workflow.

Built separate screener and reviewer workflows with role-aware study views and responsibilities.

Added reviewer conflict resolution for studies where analyst decisions differ.

Added support for pending, done, unsure, conflict, and all-studies review views.

Added project statistics, progress tracking, and exportable screening outputs.

Added color-coded Excel exports so teams can hand off included and excluded studies more easily.

Created the foundation for downstream full-text review and linked workflow use.

Added AI-assisted search string generation to help users create structured biomedical search strategies faster.

Allowed users to edit, delete, and refine generated search steps before running downstream retrieval.

Added manual Excel-based search-step upload so teams can bring in externally prepared strategies.

Added PubMed hit retrieval directly from generated strategies.

Added article-detail retrieval with titles, abstracts, authors, journals, years, DOI, PMC ID, keywords, and MeSH terms.

Added search, hits, and article review tables with pagination and easier large-result navigation.

Added immediate Excel export for search strategies, hit counts, and article details.

Added automatic backend file generation and cloud storage for search output files.

Introduced version-aware search records so evolving search strategies can be tracked over time.

Added a dedicated Change Team flow for active screening projects.

Enabled pending-only reassignment so completed screening decisions remain intact while unfinished work can be reassigned.

Allowed administrators to review active states, lock screening, reassign pending work, and safely reopen the project flow.

Added stronger admin controls around screening freeze and unfreeze handling during active project management.

Reduced the risk of leaving projects in a blocked state after team changes.

Improved visibility for admins returning to active screening projects after coordination changes.

Added timeline shortcuts from workflow activity into linked module histories.

Made cross-module review easier by allowing users to move from workflow milestones into module-specific detail views.

Improved history visibility for screening decisions, file events, extraction runs, and document-processing milestones.

Added article-level and project-level history visibility for AccuScreener screening decisions.

Enabled screening teams to see who changed a decision, what changed, and when it changed.

Added timeline access for full screening-project history alongside broader module history.

Added controlled actions for getting latest hits, retrieving latest articles, and syncing only new studies into screening.

Kept sync additive and safe so newly found studies can be brought into screening without overwriting existing work.

Added provenance tracking so screened studies retain visibility into which search cycle introduced them.

Added search-run history, sync history, and comparison views for living-review cycles.

Added schedule controls for recurring PubMed refresh with optional automatic sync into screening.

Added notifications and actor visibility so teams can tell whether a refresh was manual or scheduled.

Moved file delivery behind backend proxy routes so users can download files without dealing with expiring direct links.

Improved file-serving reliability for module outputs and result documents.

Added cleaner download behavior for shared workflow files and processed module assets.

Added bulk PDF upload for screening projects with upload-session tracking and asynchronous processing.

Added intelligent PDF-to-citation matching using extracted metadata and confidence-based review.

Added manual matching and unmatched-file review so teams can finish linking PDFs even when automatic matching is not enough.

Added workflow stages and sub-stages for search setup, initial search, hits retrieval, article retrieval, screening, extraction, and report generation.

Added automatic progress tracking so work completed inside linked tools updates the workflow timeline and stage completion state.

Added current-stage visibility and overall progress summaries for active review projects.

Added workflow-level research questions, keywords, deadlines, and project settings.

Preserved linked module IDs and workflow metadata so downstream tools can reuse upstream work.

Created the foundation for later features such as shared files, Living SLR, and workflow activity.