Version 3.0.0
LatestHighlights
Self-Learning AI
Adaptive learning from user feedback (👍/👎) — your AI gets smarter with every interaction
90%+ Accuracy
Advanced hybrid search with cross-encoder ML reranking achieves enterprise-grade accuracy
Multi-Language Code Generation
Generates cURL, Python, JavaScript, and more in a single answer
Complete Platform Overhaul
Updated all copy, help docs, onboarding, and marketing to reflect advanced capabilities
Cross-Service Guardrails
AI refuses to provide code from wrong services, preventing integration errors
Features
- •Adaptive learning system with per-tenant isolation
- •Feedback submission (thumbs up/down) on all answers
- •Chunk scoring system (+0.1 thumbs up, -0.05 thumbs down)
- •Multi-language code generation (cURL, Python, JS, etc.)
- •Table extraction and markdown formatting
- •Complete procedural flows (merges multi-step guides)
- •Cross-service guardrails (refuses wrong-service code)
- •Endpoint consistency matching
Improvements
- •RAG accuracy improved to 90%+ through advanced techniques
- •Cross-encoder reranking with Sentence-Transformers
- •Enhanced country/service detection with weighted scoring
- •Primary-source biasing for authoritative docs
- •Flow variant detection (other device, same device, async/sync)
- •Conversational intelligence (handles greetings, off-topic)
Content
- •Completely rewrote README.md with architecture details
- •Updated all meta descriptions and SEO
- •Enhanced Help page with adaptive learning section
- •Revised onboarding to teach feedback system
- •New marketing page sections (Problem, Before/After, Bento layout)
- •Updated backend docstrings (main.py, rag_engine.py)
Version 2.1.0
Highlights
New Positioning
Avenai is now positioned as "Integration Support as a Service" for API-first SaaS companies
Redesigned Marketing
New hero section with typewriter animation and clear value proposition: "Cut integration time by 60%"
Generic Terminology
Replaced industry-specific examples with universal API terms (Core API, Data API, Mobile SDK)
RAG Improvements
Increased chunk retrieval to 12 and balanced hybrid search for better answer quality
Simplified Settings
Streamlined Settings page to focus on widget configuration and API key management
Features
- •Typewriter animation on marketing page hero
- •Product-centric dashboard metrics
- •Updated "Test Assistant" functionality
Improvements
- •Enhanced markdown rendering with better typography
- •Improved thinking animation (shimmer effect)
- •Better spacing in admin dashboard
- •Cleaner Analytics page (production data only)
Content
- •Rewrote all Help page guides and FAQs
- •Updated onboarding flows
- •Generic placeholder text across platform
- •New SEO metadata and descriptions
Version 2.0.3
Features
- •Keyboard shortcuts (Cmd/Ctrl+Enter to submit)
- •Confidence chip with visual indicators
- •Empty states with helpful CTAs
- •Version number in footer
Improvements
- •Better low-confidence messaging
- •More encouraging language
- •Professional UI polish
Version 2.0.2
Features
- •Post-extractive summarization
- •Combines fragments into coherent paragraphs
- •Better readability for low-confidence answers
Version 2.0.1
Features
- •Professional code syntax highlighting
- •Copy buttons on all code blocks
- •Collapsible content sections
Version 2.0.0
Core Features
- •Hybrid search (vector + full-text)
- •Confidence scoring system
- •Streaming answers
- •Multi-tenant architecture
- •Background job processing
- •Analytics dashboard
Security
- •API key authentication
- •Tenant isolation
- •Rate limiting
- •Security headers (HSTS, CSP)
Document Processing
- •PDF and TXT support
- •Auto-tagging
- •Metadata extraction
- •Multi-file uploads
Stay Updated: We regularly ship improvements to Avenai. Check this page to see what's new, or watch for the version number in the footer.