AWS Builder Center

Agents for Humans: Building Joruri Shohay — An Emergency Helper Agent for Bangladesh

জরুরি সহায় (Joruri Shohay) is a bilingual Strands Agents emergency-response assistant for Bangladesh.

Series: #AgentsofFootball (3 articles)

  1. 2
    Agents for Humans: Building Joruri Shohay — An Emergency Helper Agent for Bangladesh This article
When an earthquake or major disaster hits a country — like the recent ones in Nepal, Venezuela, and Colombia — the first hours are chaos. People don't know which number to call. They don't know what first aid to do while waiting. They don't know which hospital is closest, which one is still standing, or which blood bank has what they need. Every minute spent searching for information is a minute not spent helping someone.
Bangladesh is one of the most disaster-prone countries in the world — dense population, active fault lines, and a monsoon season that regularly floods entire districts. When something serious happens here, most people are on their own for the first 10–30 minutes before any official help arrives. That gap is where lives are lost.
I built জরুরি সহায় (Joruri Shohay) — a bilingual emergency helper agent — to close that gap. It answers three questions in one calm, fast reply:
Who do I call right now? — real, verified emergency numbers (999, 1090, 109)
What do I do while waiting? — standard first-aid steps, matched to the situation
Where do I go? — real nearby hospitals filtered by emergency type and budget, plus blood banks
The idea is simple: no matter what happens — a road accident in Dhaka, a heart attack at home, an earthquake in Sylhet, or a disaster somewhere else entirely — anyone should be able to open this agent and get a straight answer without wading through a hundred Google results.
How it works The agent is built with the Strands Agents SDK, runs on Amazon Bedrock with Claude Sonnet 4.6, and is deployed on Amazon Bedrock AgentCore Runtime.
Six tools do the real work:
get_emergency_number — returns 999 / 1090 / 109 in the user's language
first_aid_steps — situation-matched first-aid steps, in one language only
geocode_location — converts an area name like "Dhanmondi" into real coordinates
find_hospitals_via_agent — delegates to a Hospital Finder specialist agent
find_blood_banks — returns nearby donation centers with a reminder to call first
submit_user_feedback — logs corrections for human review
Four guardrails keep it honest:
RateLimiterHook — caps tool calls per turn, prevents loops
LocationWorkflowHandler — guarantees geocoding runs before hospital search
ToneGuardrailHandler — a second LLM pass that reviews the reply for safety violations
ScopeGuardrailHandler — politely refuses unrelated requests
The whole thing streams via Server-Sent Events, keeps per-chat memory with Strands' FileSessionManager, and has a sliding conversation window to keep context predictable.
Why this matters This is a small project. But the idea behind it — that anyone, anywhere, with a phone and an internet connection should be able to get a straight, honest answer in the first minutes of an emergency — is worth building for.

Series: #AgentsofFootball (3 articles)

  1. 2
    Agents for Humans: Building Joruri Shohay — An Emergency Helper Agent for Bangladesh This article
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article