Recipe: Support Assistant
This recipe builds a searchable support assistant with a markdown knowledge base, vector search, an agent, and a lightweight frontend.
Elements
| Element | Role |
|---|---|
document | Stores support articles as markdown pages. |
vector | Stores embeddings for semantic retrieval. |
triformer | Answers user questions with retrieved context. |
view | Provides an internal support console or customer-facing search UI. |
slack, http, or websocket | Connects the assistant to the outside world. |
Flow
- Create a
documentelement for the knowledge base. - Import markdown pages into the document.
- Vectorize the document.
- Create a
triformeragent. - Attach relevant prompts, tools, and data sources.
- Create a
viewfor search, answer display, and escalation controls. - Add an
ioelement if the assistant needs to receive messages from Slack, HTTP, or realtime clients.
Why Triform Helps
The document stays editable as markdown, the search index is generated from the same content, and the assistant can discover its available actions through the API. You get a working operational app rather than a pile of disconnected services.