AI Capability Reference Architectures
Technical blueprints based on real-world documented implementations. Explore how business capabilities, integration frameworks, memory planes, and security systems coordinate to build production-grade AI ecosystems, mapped against a CMMI-inspired maturity scale.
Browse these diagrams in the Visual GalleryLevel 1: Initial — Streamlit + LangChain Quickstart Chatbot
OpenAI, LangChain, and Streamlit in 18 lines of code — Streamlit's own on-ramp tutorial, and the pattern behind most shadow-AI prototypes.
This architecture traces directly to Streamlit's own official tutorial, "Build an LLM app using LangChain" — the on-ramp Streamlit itself positions for a developer's first LLM-backed app. It's the pattern behind countless internal hackathon tools and shadow-AI prototypes: one Streamlit script wraps LangChain's ChatOpenAI class around the OpenAI API, the end user supplies their own API key through a plaintext sidebar field, and deployment is a three-click push from a GitHub repo to Streamlit's free Community Cloud tier.
Maturity Level Rationale
Matches Level 1 (Initial) because every trait comes from the source tutorial itself: the API key is handled through manual, ad-hoc configuration rather than a vault, there's no logging, evaluation, or guardrail on the raw model output, and the entire "infrastructure" is a single GitHub repo auto-deployed to a free hosting tier with no environment separation or review gate.