Try the demo · This page runs the decision rules locally. For the model-driven workflow, see the separate Strands agent below. View recorded run ↓
Your next collection
Closest confirmed option
School borehole
18-minute estimated walk · caretaker report 10 minutes old
Reported operating, not guaranteed. Check again if departure is delayed.
Reported queue: 12 minutes. Not included in walking estimates.
Opt-in only. Duplicate requests suppressed. Maximum three requests per session.
Check closer options, then add a caretaker reply in the inbox.
Market tap
8 min
Report expired
All reports have expired. Ask again.
caretaker · 95 min ago · evidence r1
East standpipe
12 min
Reports disagree
Fresh reports disagree. A newer caretaker confirmation is needed.
resident · 3 min ago · evidence r4
Garden standpipe
16 min
Unconfirmed
No fresh, unambiguous caretaker confirmation.
resident · 5 min ago · evidence r8
School borehole
18 min
Reported operating
Recent caretaker report; still within freshness window at estimated arrival.
caretaker · 10 min ago · evidence r2
Bus-stop tap
23 min
Reported closed
Caretaker recently reported not operating.
caretaker · 8 min ago · evidence r5
Community tank
26 min
Reported operating
Report will expire before estimated arrival. Confirm again.
caretaker · 48 min ago · evidence r6
Workshop tap
28 min
Report expired
All reports have expired. Ask again.
resident · 100 min ago · evidence r9
Hill tank
31 min
Reported operating
Recent caretaker report; still within freshness window at estimated arrival.
caretaker · 7 min ago · evidence r10
North handpump
35 min
Report expired
All reports have expired. Ask again.
caretaker · 120 min ago · evidence r7
South handpump
40 min
Unconfirmed
No report received.
No evidence yet · not opted in
Strands agent run
The source includes five Strands tools: inspect evidence, request confirmation, read a reply, record a grounded interpretation, and return a policy-checked answer. A local model orchestrates them; policy enforces consent, freshness, and message limits.
This is a recording of a real model run, not a live model call. Local inference uses Ollama. Amazon Bedrock is an optional configuration, not claimed as tested.