The opportunity

• A whole class of software is specific to how CHPL works — our workflows, our systems, our customers.
• Too specific for an outside product to build for us; too big a lift for our in-house capacity to take on.
• That math just changed. Agentic coding makes production-quality, CHPL-specific software practical for one staff person to build — fast.

Suggest-a-Purchase

A customer recommends something for the collection.

Built entirely on tools CHPL already owns — Sierra, existing servers, public APIs.

How it got built

Built in-house, in weeks — and built to last:

• Open-source, non-proprietary tools — no vendor lock-in.
• Fully documented — design-doc first.
• Rigorously tested — 340 automated tests.
• Built to be picked up and extended by whoever comes next.

The capability is the story — not just this one app.

The capability compounds

The real goal: less friction between a customer and the item they want.

• Suggest-a-Purchase: we don’t own it → the customer asks → we can act.
• The same tool could surface more — e.g. subscribe to a favorite author, and a match becomes a one-click hold.
• Never auto-placed — the customer stays in control.

Same Sierra integration. Same Datasette pattern. Same audit trail.

Where this goes next

• Partner with MSA (Materials Selection & Acquisition) to dog-food it internally first.
• Smooth the workflow, work out the bugs — then put it in front of customers.
• Reusable patterns — each next tool of this shape goes faster.

The capability compounds. That’s the story.