Conflux AI is a full-stack

Conflux AI is a full-stack intelligent chatbot built with Python FastAPI backend and a custom vanilla HTML/CSS/JS frontend — designed to feel like a conversation with someone who actually listens. AI & Models: Groq API runs LLaMA 3.1 8B for fast responses and LLaMA 3.3 70B for complex reasoning. Google Gemini 2.0 Flash handles image analysis via REST API with automatic fallback to Gemini 1.5 Flash and 1.5 Pro. Agent Layer: LangChain + LangGraph power ReAct agents that decide whether to search the web, read a document, or answer from memory before responding. Real-Time Data: A custom Apify service delivers live news, crypto prices, and trending topics — bypassing the LLM entirely for fresh, hallucination-free data. Document & Image Intelligence: PDFs are ingested via PyPDFLoader with auto-summarization. Images can be uploaded, attached, or pasted from clipboard and analyzed through Gemini Vision. ReportLab generates downloadable PDF exports of any response. Auth: Firebase Admin SDK verifies JWT tokens server-side with full Google and email sign-in support. Frontend: Three-panel dark UI with a complete CSS animation system — messages fade up on entry, AI responses render character-by-character via a custom typewriter engine, and dynamic follow-up suggestions update after every response with staggered cascade animations. Emotional Intelligence: A fully custom keyword-based engine detects 10+ emotions in real-time — happy, sad, anxious, angry, curious, and more — adapting the AI's tone, language, and empathy level with every single message. Mood history is tracked per authenticated user with trend analytics. Stack: FastAPI · LangChain · LangGraph · Groq · Gemini · Firebase · Apify · ReportLab · highlight.js · Vanilla JS Conflux AI v5.1 — Built by Muhammad Noman Ijaz, Lead Engineer at Conflux. March 2026.