About Project

Keven is my personal intelligent conversational AI assistant built with LangGraph that provides multi-modal responses through WhatsApp, Chainlit, and other messaging platforms. The assistant features memory persistence, contextual awareness, and supports text, image, and audio interactions

Features

Multi-Modal Communication: Text, image generation, and voice synthesis

Persistent Memory: Long-term memory with vector storage using Qdrant

Context Awareness: Schedule-based activity injection for personalized responses

Multiple Interfaces: WhatsApp webhook, Chainlit UI, and Messenger support

Conversation Management: Automatic summarization for long conversations

Image Analysis: Upload and analyze images with AI-powered descriptions

Voice Interaction: Speech-to-text and text-to-speech capabilities

Technologies

LangGraph: State-based conversation flow management

Multiple LLM Providers: Groq, OpenAI, Together AI support

Vector Database: Qdrant for semantic memory storage

Image Processing: FLUX.1 for image generation, GPT-4o-mini for image analysis

Speech Processing: Whisper for STT, ElevenLabs for TTS

Designed and developed by Brandon Bwoni