LLM-based Chatbot
In Progress
Experimenting with OpenRouter API to build an internal chatbot that understands my research papers, creative work, and personal knowledge systems. This is a playground for understanding how AI can augment creative research and make sense of social data.
What is this?
This project explores the integration of OpenRouter API—a unified gateway to multiple Large Language Models (LLMs). By using selective free-tier models, I'm building AI-driven tools without the cost overhead, making experimentation accessible and sustainable.
Why OpenRouter?
- Model diversity – Access to 30+ free models from Meta, Google, and others
- Free-first routing – Intelligent fallback system prioritizes free models
- No vendor lock-in – Switch between models seamlessly
- Cost transparency – Built-in quota management and usage tracking
Technical Architecture
| Component | Purpose |
|---|---|
OpenRouter Client | HTTP wrapper for API communication |
Model Catalog | Discovers and caches available models |
Quota Manager | Enforces rate limits (20/min, 1000/day) |
Policy Engine | Routes requests to optimal free models |
Executor | Handles retries and intelligent failover |
Current Capabilities
- Real-time conversational AI with context memory
- Response latency tracking (typically 1.4–1.9s)
- Transparent model identification in every response
- Automatic quota management and rate limiting
- Graceful error handling with model failover
Future Vision
The long-term goal is to train this chatbot on my research papers, photography notes, and creative writings. Imagine asking: "What are the recurring themes in my grief photography?" or "How does my work on memory connect to my emotional research?"
This is still early-stage exploration. Right now, it's a general-purpose chat interface—think of it as a testing ground for understanding how LLMs can become thought partners for creative research.