Guides
How to design production AI systems—architecture, tracing, and model routing—without replacing the API reference.
A question, then the system that answers it.
Who the guides are for
Engineers and platform owners who already have a model in production and need the system around it: sources, tools, and a record of the run. If you are choosing a first chatbot template, these pages will feel early. If you are tired of glue code, start here.
- Production AI architectureGUIDE
Why stitched model calls, vector stores, and tools become an operations burden, and how to keep knowledge, agents, and traces as separate seams you can change.
- Execution tracingGUIDE
What to record on every agent run—retrieval, inference, and tool calls—so a failure can be inspected and replayed instead of reconstructed from three logs.
- Model routing without lock-inGUIDE
How to treat a model as a component: route the same task, compare quality, latency, cost, and groundedness, and leave the agent definition in place.
Questions
What are Obliq guides?
Guides are implementation write-ups for production AI systems: the problem, how the pieces fit, what to build, and what Obliq does not claim. They are not generic AI explainers.
Where should I start?
Start with production AI architecture if you are deciding the seams. Start with execution tracing if you cannot explain a bad run. Start with model routing if a provider change is already on the calendar.
Are the guides a substitute for the API docs?
No. Guides explain the operating model. The developer pages cover the control plane, SDKs, and MCP.