Privacy

Updated September 11, 2026. GovernSmyrna is an independent civic information service. You can browse without an account.

Addresses and sharing

As you type an address, this service sends partial address text to Amazon Location Service to suggest nearby street addresses. Selecting a suggestion retrieves its location from Amazon. If suggestions are unavailable, the service searches its captured city address records. We do not log address text or save suggestions or selected Amazon locations to a profile, database or shared cache. These results stay in this page session; reloading opens the shared ward or city view. Amazon processes the lookup under its service terms. City and ward membership is checked against the official boundary records, not postal city names. A building selection can appear in your browser URL and history; anyone with that exact link may resolve that building. The normal share action uses the ward or broader area, not the home address.

Do not put an address or other private information into a discussion search. Search text is sent to Amazon Bedrock, an AWS service, to find related topics, and used to retrieve matching public passages from the site’s database.

Saved views

Only an explicit save action stores a reproducible selection in private Amazon S3 storage. Saved links are readable by anyone who has the link. The saved record includes the selected records and filters; it is not an account or access-controlled personal workspace.

Browser storage

The browser may cache public model assets and retain your motion preference. This tab also remembers the illustrative campus camera position between pages; that position is separate from your selected address. Model caching does not cache address requests, public-record API responses or saved-view requests. Clear this site’s storage through your browser settings to remove these local preferences and cached assets.

Service operation

Requests necessarily reach the server with network information, including an IP address. In-memory request limits use network addresses to protect the service. Idle rate-limit entries expire after about an hour. This in-memory table is separate from the limited security records described below. The application does not write address input, request bodies or search queries to an access log. Unexpected errors are logged as a generic error name and status, without request content.

The application records redacted request logs in Amazon CloudWatch under a seven-day retention policy: time, a newly generated request ID, a broad page or API category, HTTP method and status, duration, response-body byte count and whether the request completed. They also record a broad traffic category based on whether the client identifies itself as a service monitor or automated software, and the type of request limit when a request is throttled. These categories do not identify a person; an unclassified request may come from a browser or automated software. These logs contain no IP addresses, full URLs, query strings, selected building IDs, coordinates, request bodies, cookies, authorization headers, referrers or user-agent strings. Log volume is capped; aggregate counters continue when detailed records are dropped. Aggregated traffic and performance metrics may be retained longer than the request logs. They do not identify visitors or measure unique people. Raw HTTP access logging remains disabled in the web server. Separate service and infrastructure logs record startup, failures and administrative operations in the system journal; there is currently no explicit age-based deletion guarantee for those operational logs. Uptime probes store only service availability, check time and latency. The displayed history covers seven days; older versions may remain in versioned infrastructure storage. We do not use advertising trackers, session replay, analytics cookies or third-party fonts.

Security diagnostics for throttled requests

When the application rejects a request with a rate-limit response (HTTP 429), it may record the source IP address, the user-agent supplied by the client (limited to 1,024 characters), time, a fresh request ID, HTTP method, broad route category, and the limit that rejected the request. A throttled request does not by itself establish abuse. User-agent claims are not verified identities. These records help authorized operators investigate automated load and service abuse; they are kept separately from ordinary traffic logs, audience analytics and the civic records database.

The restricted security log group in Amazon CloudWatch has a one-day retention setting. AWS expires records under that setting, but physical deletion is asynchronous and typically can take up to 72 additional hours, occasionally longer. Access requires authorized AWS log-read permissions; the application can write these records but cannot read them. We do not export them to long-lived reports or run identifying analytics over them. Collection is capped to limit volume. Security records do not copy URL paths or query strings, entered addresses, search text, civic-record IDs, request bodies, cookies, authorization headers or referrers. The client-supplied user-agent is retained only as the bounded diagnostic field described above. This does not create an analytics cookie or visitor profile.

Original public records

Public filings can contain personal details. Residential addresses, phone numbers and signatures are omitted from the disclosure interface. Filing links open the original state portal, whose privacy practices apply there. The data license does not waive anyone’s privacy or other rights.

External services

The service runs on AWS, with Neo4j used for record retrieval, Amazon Location Service for address suggestions and Bedrock for topic embeddings. External source links may open city, state or recording-provider sites. Those sites receive the requests you choose to make.