Case Studies
Languages: English · 中文
Each case study is a short walkthrough of one realistic scenario: the problem, the structure that solves it, the Agently pieces involved, and the trade-offs that drove the decisions.
These are examples, not prescriptions. They highlight common shapes you'll see when building real applications. If a case study mostly fits your scenario, follow it; if not, the Capability Map and Playbooks are better starting points.
Available case studies
| Case study | Combines |
|---|---|
| Daily News Collector | TriggerFlow + tools + structured output + scheduled run |
| Talk to Control | Conversational agent + actions on a domain object + streaming |
| Knowledge Base Dialog | Embeddings + retrieval + session memory + structured answer |
| PRD → Test Cases | Long-input structured output + ensure-marked required fields + per-section streaming |
| Survey Dialog | Multi-turn session + dynamic prompts + branching follow-ups |
What each case study covers
Each page is structured the same way:
- The problem — what someone actually asked for.
- The shape — how the pieces fit together at a glance.
- Walkthrough — the relevant code with explanations.
- Why these choices — the trade-offs and what the alternatives would cost.
- Where it lives — pointer to the runnable example in the repository, when one exists.
Reading order
If you're new, read Daily News Collector first — it covers the most ground in the least space. If you have a specific shape in mind (RAG, conversational with actions, long-input parsing), jump straight to the relevant case.