Bedrock Agent Weather | Architecture

Click any component to explore it in the full case study.

Natural Language Weather Query

User asks: "What's the weather?", "Will it rain?", "Are conditions favorable for storms?". Bedrock Agent routes to the appropriate action group tool based on semantic understanding.

Natural Language · Bedrock routing

Path A: Lambda Direct (AWS-native)

Bedrock → Lambda integration. Agent calls Lambda function directly with actionGroup + apiPath + parameters. Lambda routes to GCP Cloud Run weather API. Response envelope echoes actionGroup, apiPath, messageVersion.

Bedrock → Lambda → GCP Cloud Run

bedrock-agent-weather.html

Path B: Logic Apps Hybrid (Azure bridge)

Bedrock → Azure Logic App HTTP trigger. Logic App relays to GCP Cloud Run endpoint. Tests cross-cloud integration path: AWS intelligence → Azure orchestration → GCP data.

Bedrock → Logic Apps → GCP Cloud Run

bedrock-agent-weather.html

3 OpenAPI-Defined Weather Tools

get_weather_conditions

Current METAR: temperature, humidity, precipitation, wind, pressure, visibility, cloud cover, wx_string

get_weather_forecast

5-period hourly NWS forecast. Short description, temp, wind speed, precipitation probability.

get_atmospheric_conditions

Upper-air radiosonde data: CAPE, CIN, Lifted Index, K-Index, Total Totals for convective analysis.

OpenAPI 3.0 Schema Contract

All 3 tools defined via OpenAPI 3.0 schemas. Bedrock routes on semantic understanding, not keyword matching. Proper response envelope (actionGroup + apiPath + messageVersion echo) is required — malformed responses cause silent failures.

OpenAPI 3.0 tool definitionsSemantic routing (not keyword)Response envelope contract3 clouds: AWS + Azure + GCP2 integration paths testedSilent failure if malformed

3

Action group tools

2

Integration paths

3

Clouds (AWS + Azure + GCP)

2

Weather stations

StackAmazon BedrockClaudeAWS LambdaAzure Logic AppsGCP Cloud RunOpenAPI 3.0NWS APIUWY SoundingsMETAR