- Upgraded the CI pipeline for better automation, faster feedback, and error visibility.
- Introduced and expanded
pytest
framework for ETL testing. - Added test coverage across shared libraries and core modules.
- Refactored shared libraries and internal tools for better modularity.
- Upgraded ETL Dockerfile to Python 3.11.11-slim for improved performance.
- Enhanced agent backend with versioning, naming conventions, and templated auto-instantiation.
- Improved auto agent pipeline and prefix handling logic.
- Added support for chat deletion and improved chat endpoint stability.
- Minor typo corrections and internal cleanup of chat code.
- Enabled Jira ticket generation via API and Google Forms integration for workflow triggers.
- Introduced Cypher query tools for graph-based retrieval.
- Added domain-specific query support for customized search experiences.
- Enabled microphone and voice input within embedded iframes.
- Production-ready transcription integrated for voice-to-text.
- Updated Dockerfiles and dependencies across modules for security and performance.
- Migrated chat system to DynamoDB, supporting full conversation tracking and querying.
- Improved
conversation_id
handling for user-level context tracking.
- Integrated WhatsApp as a communication channel.
- Improved MongoDB connection and DeepSeek model support.
- Added image reading and basic PDF metadata extraction.
- Improved Jira delta sync logic with enhanced JQL parsing and debugging tools.
- Built robust HTML content extractor and metadata sync for Confluence.
- Enhanced ingestion and cleanup tools.
- Reduced excessive logs, improved log clarity, and optimized pipeline stringification.
- Added support for hypothetical questions in RAG.
- Introduced batch semantic retriever and advanced Cypher query handling.
- Upgraded connectors for Google Cloud, Azure, MongoDB, and Box.
- Improved ETL logic for multimodal and multifile document support.
- Integrated semantic retrievers, hypothetical rephrasing, and
rag_tool
foundations.
- Enabled graph traversal using Cypher and TigerGraph integration.
- Deployed graph-based query pattern tools.
- Improved website crawling using FireCrawl and bot-avoidance mechanisms.
- Enforced subscription-based limits for file uploads and applications.
- Enhanced billing UI with upgrade prompts and usage tracking.
- Launched evaluation interface with metrics like ROUGE and token overlap.
- Users can configure, run, and view evaluation results via dashboard.
- Released chatbot workflow config UI with prompt templating and zero-knowledge search.
- Redesigned chat config with new logic layers and routing.
- Enabled chatbot preview before publishing, including summary and sample question customization.
- UI-configurable hypothetical questions for improved RAG.
- Runtime knowledge base switching support.
- Full support for Google Drive and Google Cloud Storage connectors.
- Added OAuth-based Notion and Outlook integrations.
- Step-by-step app creation flow with chat, search, and evaluation modes.
- Enhanced drag-and-drop upload and real-time previews.
- Enabled project-level access controls, tiering, and form validations.
- Added Azure OpenAI and Gemini support.
- Internal tooling for rerankers and graph traversal logic.
- Released
epsilla.js
for website embedding. - Custom logos, markdown, LaTeX, and theme controls supported.
- Upgraded default LLM to GPT-4o for speed and performance.
- Improved web content syncing and chunking logic.
- Added dynamic knowledge refresh options.
- Introduced debug mode for apps with real-time render visualization.
- Stored debug sessions for replay and analysis.
- Per-file sync tracking and metadata display.
- Special character support in file names and paths.
- Opacity control, improved iframe embedding, table rendering in markdown.
- Enabled public sharing of chatbot messages and apps via URL.
- Added financial-specific prompt templates and timeout settings.
- Integrated CambioML for custom vertical workflows.
- Better file chunking, preview rendering, and filename validation.
- Added dynamic stream of knowledge base during query resolution.
- Real-time updates to rephrasing logic and query adaptation.
- Experimental voice transcription and multi-modal UX introduced.
- Markdown enhancements, input resizing, responsive layout tuning.