Most API examples stop at a single curl call. This repo goes the other way: each folder is a deployable mini-app you clone, point at your own data, and ship — the gap between "hello world" and "running in production" is exactly what these quickstarts try to close.
What Sets It Apart
- Each quickstart targets a concrete shape of work, not a feature in isolation: customer support with knowledge-base grounding, a financial analyst that returns charts inside the chat, and computer-use / browser-use agents that drive a desktop or a Playwright browser.
- The computer-use track ships both a containerized demo and a separate best-practices implementation covering prompt caching, server-side compaction, batched tool calls, and trajectory recording — the unglamorous parts that decide whether an agent is affordable to run.
- The autonomous coding agent uses a two-agent pattern (initializer plus coder) with git-based persistence, so it can build an app across multiple sessions instead of one heroic prompt.
Great Fit If
You learn faster by editing working code than reading API docs, and you want a defensible starting point for a support bot, data-analysis chat, or tool-using agent. Look elsewhere if you need a finished product rather than scaffolding — these are intentionally minimal so you replace the example data and prompts with your own. The mix of Python and TypeScript also means you should expect to pick the stack that matches the example you want, not the other way around.
