SCADA Telemetry Workflow & Bedrock Inference Engine using Mastra AI & Amazon Bedrock
Constructing scalable modern TypeScript industrial microservices requires stateful agent workflows, built-in vector memory for plant manual search, and high-throughput model inference. In this challenge, developers will leverage Mastra AI framework integrated with Amazon Bedrock runtime services to deploy an agentic SCADA event orchestration engine. The Mastra AI workflow will ingest streaming telemetry events, execute automated vector lookups in plant maintenance manuals, and invoke Anthropic Claude 3.5 Sonnet hosted on Amazon Bedrock to validate whether live sensor metrics comply with Rosemount 350A pH/ORP equipment parameters.
What you are building
The core problem, expected build, and operating context for this challenge.
Deploy a Mastra AI workflow with Amazon Bedrock to ingest SCADA streams, query plant manuals, and validate pH/ORP sensor calibration.
How work is evaluated
Evaluates Mastra AI workflow step execution efficiency and calibration verification accuracy for Rosemount sensor inputs.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Recalibration Detection
Sensors uncalibrated > 30 days with pH offset must trigger RECALIBRATE_REQUIRED
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Workflow Completion Time
Execution latency for complete Mastra workflow step evaluation • target: 1200 • range: 0-5000
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Build stateful TypeScript agent workflows using Mastra AI framework
Integrate Amazon Bedrock inference providers within Mastra AI execution trees
Implement vector RAG memory for Emerson Rosemount 350A technical manuals
Execute automated telemetry verification and signal validation steps
Reference links and supporting material
Dataset of technical manuals for Emerson Rosemount 350A pH/ORP sensors converted into markdown and vector embeddings.
Challenge setup
Evaluates Mastra AI workflow step execution efficiency and calibration verification accuracy for Rosemount sensor inputs.
Challenge input
JSON object with pH, ORP_mV, temperature_c, last_calibrated_days
Mastra AI
Assigned primary orchestration framework
Amazon Bedrock
Assigned inference provider tool
Evaluated output
JSON object with status (OK|RECALIBRATE_REQUIRED), confidence, adjustment_offset
- Sensors uncalibrated > 30 days with pH offset must trigger RECALIBRATE_REQUIRED
- Execution latency for complete Mastra workflow step evaluation • target: 1200 • range: 0-5000
- Benchmark: TypeScript Agentic RAG Benchmark
- Workflow Completion Time target: 1200
- 1 public reference case
- JavaScript execution harness
- javascript sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Amazon Bedrock · Required
- Claude Code · Optional
- Anthropic · Optional
- Claude Code · Optional
Evaluation contract
- Recalibration Detection · Weight 1
- Workflow Completion Time · Weight 1
Metrics
- Workflow Completion Time · Target 1200
Benchmarks
- TypeScript Agentic RAG Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with pH, ORP_mV, temperature_c, last_calibrated_days
Output: JSON object with status (OK|RECALIBRATE_REQUIRED), confidence, adjustment_offset
Execution harness: JavaScript
Evaluation tools
- Mastra AI · TypeScript Agent framework
- Amazon Bedrock · Managed model runtime
Recipe freeze marker
No recipe freeze time is recorded.