RFP §3.7 — Chatbot Readiness

Chatbot-Ready API

The redesigned site exposes all public content as a queryable API. Whether the County builds an AI chatbot in-house or contracts with a vendor, this is the integration point — already documented, already live.

Single source of truth

Drupal CMS is the editor's tool. Both the website and the chatbot read from the same endpoint — no content drift.

CISA-aligned

Read-only public endpoint. No authentication needed; no PII exposed. CORS-permissive.

Vendor-agnostic

Works with OpenAI Assistants, Anthropic Claude, Voiceflow, Cohere, or any RAG-style integration.

Endpoint

GET https://cheboygan.sparkdemo.cloud/api/chatbot-snapshot.json

In production this is replaced by the live GraphQL endpoint at https://cms.cheboygancounty.gov/graphql with the same item shape.

View the live snapshot

Item shape

{
  "title":     "County Treasurer",
  "path":      "/departments-services/treasurer/",
  "kind":      "department",
  "text":      "The office of County Treasurer was established by …",
  "tags":      ["department", "current"],
  "sourceUrl": "https://www.cheboygancounty.net/departments-services/treasurer/"
}

Sample integration — embed-then-retrieve

A typical RAG (retrieval-augmented generation) integration:

Nightly job fetches /api/chatbot-snapshot.json.
Vendor (e.g. OpenAI, Pinecone, Voyage) embeds each item.
User asks a question → vendor retrieves top-N relevant items.
LLM is given the items as context and asked to answer.
Answer includes the path of each cited source so the chatbot can deep-link to the page on cheboygancounty.gov.

What is not included

Personally identifiable information (PII).
Webform submission data — never exposed via this endpoint.
Authenticated admin content (CMS drafts, internal notes).
The chatbot itself — this proposal scopes the API + integration framework; the conversational UI is a future engagement.

Why this matters for the RFP

Cheboygan County asked for "chatbot readiness" — architectural compatibility for future AI chatbot integration via API. The decoupled drupal-astro architecture answers this requirement natively: the same GraphQL endpoint that drives the website IS the chatbot's data source. No additional infrastructure, no per-document re-export, no content duplication. Adding a chatbot later is a vendor-integration task, not a re-architecture.

Search