Changelog

Multi-sheet workbooks now extract significantly faster with reduced timeout risk for large workbooks.

Fixed a bug where some files were sometimes skipped.

Instructions now support an optional context template that prepends document metadata to each extraction call, giving you fine-grained control over what context the model sees.

Entity extraction now includes document name as context for each chunk, producing more accurate results especially in multi-document workflows.

Fixed a reliability issue where Excel files were sometimes misidentified during ingestion, causing them to fail silently. Excel files are now correctly recognized and processed.

Resource limits for a partition are now surfaced in a dedicated Limits tab on the partition details page, making it easier to review at a glance.

Ragie Parse (Agentic OCR) now captures handwritten signatures, detects whether they're signed or unsigned, classifies form fields by type, and extracts key-value pairs from label-dense documents like invoices and contracts.

Introducing Ragie Parse (Early Access), a new Agentic OCR pipeline that uses vision AI to extract structured content with higher fidelity than traditional OCR, supporting 25+ element types including tables, forms, signatures, key-value pairs, barcodes, and stamps.

The web crawler connector can now be created and managed programmatically through the public API without going through the dashboard (Enterprise only).

Spreadsheets now include embedded images and charts alongside table data, with automatic separation of table and freeform regions for cleaner, more complete extraction.

PPTX and PPT files now benefit from the same high-resolution extraction pipeline as native PDFs, resulting in significantly improved output quality.

The file size limit for URL-based document uploads has been increased from 60 MB to 100 MB.

‍

Optimized database indexing to improve speed and reduce CPU usage. Document assembly is now instant (down from seconds to milliseconds), and API key lookups are 100x faster, ensuring the system handles high volumes with minimal latency.

‍

A new endpoint (GET /webhook_endpoints) to retrieve a list of all your configured webhook endpoints.

EU deployments so Ragie workloads and data remain within European regions for GDPR compliance while lowering latency for EU users.

Ragie MCP Bridge (early access), a permissioned bridge that brings RBAC, SSO, SCIM provisioning, and model-agnostic retrieval to internal employees.

Ragie Playground for experimenting with retrieval, generation, top_k, re-rank, recency bias, system prompts, and chunk inspection without writing code.

Large-document ingestion tuned to handle 10,000+ page PDFs up to 2× faster, keeping compliance binders and research troves searchable.

Enterprise security & compliance: SOC 2 Type II, GDPR, and HIPAA certifications plus AES-256 at rest and TLS in transit with strict partition isolation.

Ragie Connect updates featuring whitelabel connectors, Dropbox Teams support, and generalized web crawling controls for richer ingestion options.

Core platform enhancements including RTL language support, faster PDF ingestion, and dependable parsing for 10,000+ page documents.

Streamable HTTP MCP Server with context-aware tool descriptions, multi-tenant isolation, and zero local process management for deploying agents at scale.

Agentic Retrieval GA, enabling automatic query decomposition, multi-hop retrieval, self checks, and grounded answers for complex workflows.

Web Crawler (early access) with domain scoping, link-count limits, and depth controls to convert any website into a Ragie-ready corpus.

Full right-to-left language support (Hebrew, Arabic, Persian, etc.) across ingestion, OCR, indexing, and answer generation so global teams get native-quality responses.

Context-Aware MCP Server that updates tool descriptions dynamically based on the underlying content, keeping agents aligned with the freshest knowledge.

Ragie Connect whitelabeling so connector auth flows can be fully branded and embedded inside customer products.

Agentic Retrieval closed beta, exposing Ragie’s multi-step reasoning pipeline to early adopters.

Partition Limit API endpoints to enforce ingestion quotas per partition programmatically.

Partition Patterns that route webhooks via glob-style keys (for example `dev-*` or `tenant-*`) to simplify multi-tenant automation.

Zendesk and Intercom connectors that pull help-center articles with rich metadata, author info, vote counts, and canonical URLs.

Enterprise file expansion for SharePoint, Backblaze, and granular Google Drive file uploads to cover corporate repositories.

Open-source ragie-mcp-server (v1.0) delivering a full MCP toolset backed by Ragie retrieval.

ragie-cli binaries for macOS, Linux, and Windows that bulk-ingest local files, archives, YouTube content, WordPress docs, and more.

Promptie, a playground for iterating on prompts, retrieval settings, and source inspection.

Audio & Video RAG GA: timestamped citations, playback snippets, visual understanding, and pay-as-you-go pricing with 10 free hours.

Early access to Audio & Video RAG with multilingual transcription, on-screen analysis, and support for media formats such as MP3, WAV, MP4, and AVI.

SharePoint and Backblaze connectors plus Google Drive file-level uploads so enterprises can ingest exactly the assets they need.

Partition Limits for capping pages processed and hosted per tenant, giving operators fine-grained usage controls across multi-tenant environments.

Ragie MCP Server, a turnkey Model Context Protocol endpoint that lets assistants like Cursor or Windsurf securely search Google Drive, Jira, Slack, and more.

Sync Connections API so engineers can programmatically trigger connector refreshes without visiting the UI.

Connector page-limit controls to stop runaway ingestion and manage spend per tenant or integration.

Document Chunks API, exposing the exact context Ragie generated and enabling "chunk fill" optimizations.

Additional webhook statuses announcing when documents reach the indexed state, enabling latency-sensitive applications.

Recency bias controls that weights fresher documents higher during retrieval for time sensitive workloads.

Specialized table chunking that extracts dense tabular data more faithfully for downstream retrieval.

In-dashboard Ragie Connect experience with a launch-week demo showcasing end-to-end connector flows.

Usage analytics and self-serve plan upgrades inside the Ragie dashboard for clearer consumption tracking.

Ragie Connect GA, handling OAuth, permissions, and background syncing for user-owned data sources such as Google Drive, Salesforce, and Notion so teams can ship RAG on customer data quickly.

Table chunking accuracy (+60%) and extraction speed (+25%) plus validated quality on LegalBench (99.4% accuracy) and FinanceBench (137% retrieval uplift) to keep structured data responses reliable.

Webhooks with signed payloads for document lifecycle, entity extraction, and deletion events so apps can react as soon as data changes.

A citation cookbook showing how to generate grounded answers with chunk-level sources.

Gmail Connector that securely ingests conversations and attachments with label-based scoping, per-connection metadata, and API-key attribution for downstream tracing.

Partitions for isolating multi-tenant data sets, preventing cross-tenant leakage and boosting hybrid search relevance through partition-aware keyword weighting.