A multi-provider conversational AI interface featuring local and cloud LLM routing, dynamic web search, document
ingestion, and persistent session management.
Client / ContextInfinity Solutions (Local AI Prototype)
Many modern AI workflows require switching between multiple model providers—such as comparing local,
privacy-focused open-weight models with cloud foundation models—while maintaining context, uploading
documents, and accessing real-time web data. The goal was to build a unified, high-performance interface
that grants power users full control over model selection, privacy parameters, and external data grounding
without vendor lock-in. This was a prototype for a client that requested a local AI interface with choice
of AI tools.
The Solution & Technical Highlights
Multi-Provider LLM Switcher
Architected a modular control center allowing instant switching between OpenAI, Anthropic, Gemini, and
local Ollama models mid-conversation.
Document RAG & File Ingestion
Built client-side and server-side document parsers capable of extracting and analyzing text, PDFs, and
data files for document Q&A and summarization.
Real-Time Web Search Augmentation
Integrated live web search toggling to ground assistant responses in current events and citations.
Local-First Privacy & Encryption
Implemented client-side API key storage and persistent chat history stored directly in browser
cache/IndexedDB without mandatory backend database retention.
Low-Latency Token Streaming
Leveraged Server-Sent Events (SSE) with robust markdown, math formula, and code syntax highlighting in
real time.
Key Outcomes & Impact
Created a lightweight, daily-driver AI assistant environment eliminating recurring SaaS platform
subscription overhead.
Enabled hybrid workflows pairing local offline inference with top-tier cloud models in a unified UI.
