Changelog

Meeting Intelligence

• New AI-Extracted Action Items — Automatically identifies action items with assignees, priorities, and due dates
• New Key Decision Capture — AI extracts decisions made during meeting transcripts
• New Smart Re-Send Detection — Prevents duplicate processing when the same transcript is sent again

Daily Briefings

• New AI Morning Briefings — AI-generated morning briefings delivered at 7am UTC
• New Cross-Meeting Themes — Cross-references themes across yesterday's meetings
• New Follow-Up Tracking — Highlights pending action items that need attention

Team Sharing

• New Meeting Sharing — Share meetings with teammates via email invitation
• New AI Privacy Detection — Warns about sensitive content before sharing
• New Ownership Model — Meeting owner controls sharing permissions
• New Inverted Index Table — Efficient shared transcript queries via new DynamoDB index

Document Library

• New File Uploads — Upload PDFs, DOCX, TXT files directly to your library
• New Document Linking — Link documents to meetings and speakers
• New PDF Text Extraction — Automatic text extraction via AWS Textract

Companies & Knowledge Graph

• New Company Auto-Detection — Automatically detect companies mentioned in meetings
• New Company Profiles — Meeting history and related speakers per company
• New Entity Relationship Browser — Visual browser for knowledge graph connections

Infrastructure

• New CloudWatch Monitoring — Dashboard and alarms for system health
• New SQS Dead Letter Queue — Pipeline for failed processing recovery
• New Data Protection Controls — Enhanced data protection across the platform
• New Krispy Live OAuth — OAuth connection flow for Krispy Live integration
• Change MCP Server Migration — Migrated from local stdio to Lambda-hosted HTTP

Bug Fixes

• Fix Auth Route 401 Errors — API routes now use getToken() instead of auth()
• Fix Build Version Logging — Build version and deploy timestamp now logged to browser console

Infrastructure as Code

• New AWS CDK Migration — Full infrastructure now managed via CDK with type-safe TypeScript
• New GitHub Actions CI/CD — Auto-deploy on push to main, daily drift detection at 6am UTC
• New OIDC Authentication — Secure GitHub Actions IAM role without static credentials

AI Transcript Parser

• New Multi-Format Support — Automatically parse Zoom, Teams, Slack, podcast, and custom formats
• New AI Fallback — When rule-based parsing fails, Bedrock Nova Lite intelligently extracts structure
• New Confidence Scoring — Parser reports confidence level for each transcript interpretation

Knowledge Graph

• New Entity System — Universal entity store for speakers, companies, topics, documents
• New Relationships — Graph edges connecting entities (works_at, participant, mentioned)
• New MCP Tools — New list_speakers, list_companies, get_entity_relationships tools

Multi-Tenant Authentication

• New User Isolation — All queries scoped by user ID for data privacy
• New API Key Support — Generate API keys from dashboard for programmatic access

Scheduled Jobs

• New Morning Briefing — Daily briefings generated at 7am UTC
• New Speaker Enrichment — Nightly bio enrichment at 2am UTC

Bug Fixes

• Fix Speaker Name Persistence — Edits now save correctly
• Fix URL Encoding — Double-encoded speaker URLs handled properly

Speaker Intelligence

• New AI Bio Enrichment — One-click web search finds LinkedIn profiles, AI validates matches against your meeting context
• New Human Verification — Confirm or reject bio matches. "Wrong Person" button lets you add hints and re-search
• New Full Name Discovery — When you verify a "Babak", the system updates to show "Babak Hosseinzadeh"
• New Topic Analysis — AI extracts topics from conversations. Browse speakers by topics they discuss
• New Speaker Profiles — Rich profile pages with meeting history, AI insights, verified bios, and topic tags

Transcript Experience

• New iMessage-Style Chat View — View transcripts as chat bubbles. Your messages on right, others on left
• New Speaker Talk Time — Visual breakdown of who talked most in each meeting
• New Clickable Speaker Badges — Click name to view profile, click pencil to edit speaker name

Search & Organization

• New Advanced Search — Filter by date range, speaker, duration. Full-text search across transcripts
• New Company Tracking — Relationship intelligence tracks companies mentioned in calls
• New Topics Browser — Browse all extracted topics, see related meetings and speakers

Privacy & Data Management

• New Delete Transcripts — Remove meetings with full cascade (S3, DynamoDB, vectors)
• New Private Transcripts — Mark sensitive meetings as private to hide from MCP queries

Content Import

• New Document Upload — Upload PDF, DOCX, TXT files directly
• New Link Import — Import content from URLs with web scraping

Speaker Identification

• New Speaker Correction Tool — Map unknown speakers like "Speaker 2" to real names via the update_speakers MCP tool
• New LinkedIn Integration — Optionally store LinkedIn URLs when identifying speakers
• New Auto-Apply Corrections — Speaker corrections automatically applied when fetching transcripts
• New Transcript Text Replacement — Corrected names replace original names throughout the transcript text

Admin Dashboard

• New Optional Web Dashboard — Browse transcripts, semantic search, and manage speakers via web UI
• New Docker Deployment — Dashboard deploys to AWS App Runner with automatic SSL
• New Password Protection — Simple password auth protects your dashboard

Website

• New Favicon — Added site favicon in all standard sizes
• New Mobile Responsive — Full mobile-friendly design for all pages
• New Changelog Page — You're looking at it!

Security & Performance

• New Webhook Authentication — API key validation ensures only your Krisp can send data
• New Instant Indexing — Transcripts indexed to DynamoDB immediately on webhook receipt
• New Auto MCP Configuration — Installer auto-configures Claude Desktop and Claude Code
• New Debug Logging — Comprehensive logging in MCP server for troubleshooting

Bug Fixes

• Fix listRecent() Query — Now uses date GSI instead of arbitrary scan results

Initial Release

• New Webhook Lambda — Receives Krisp webhooks and stores transcripts to S3
• New Processor Lambda — Indexes transcripts and generates vector embeddings
• New DynamoDB Index — Fast metadata queries with date and speaker GSIs
• New S3 Vectors — Semantic search using Amazon Bedrock Titan embeddings
• New MCP Server — Three tools: list_transcripts, search_transcripts, get_transcripts
• New One-Line Installer — curl | bash deployment to your AWS account
• New Documentation Website — krispy.alpha-pm.dev