Autonomous Factory Telemetry Agent with OpenAI Agents SDK and Devin
Legacy digital transformation projects fail due to fragmented, isolated telemetry streams across heterogeneous factory floors. You will develop an autonomous agent system using OpenAI Agents SDK alongside Cognition Devin integration to aggregate MQTT/SCADA feeds and generate corrective control actions. The agent must successfully process multi-turn facility incidents with under 500ms decision latency.
What you are building
The core problem, expected build, and operating context for this challenge.
Implement an event-driven industrial control agent using OpenAI Agents SDK and Devin to orchestrate SCADA alarm resolution across factory networks.
How work is evaluated
Evaluates response accuracy and latency for real-time SCADA alarm events handled by OpenAI Agents SDK.
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.
safety_override_check
Verifies that control actions do not violate safety interlocks.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
decision_latency_ms
Time taken from telemetry receipt to action dispatch • target: 450 • range: 0-2000
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 multi-turn agentic workflows using OpenAI Agents SDK and function calling
Utilize Cognition Devin for autonomous code patching of PLC translation scripts
Parse and transform real-time MQTT payload messages into standardized ISA-95 asset tags
Measure and optimize decision pipeline execution latency to meet sub-500ms requirements
Reference links and supporting material
High-throughput synthetic MQTT factory messages representing continuous conveyor speed, temperature, and current metrics.
Challenge setup
Evaluates response accuracy and latency for real-time SCADA alarm events handled by OpenAI Agents SDK.
Challenge input
JSON array of raw SCADA MQTT alarm packets
OpenAI Agents SDK
Required framework assignment.
Cognition Devin
Required tool assignment.
Evaluated output
JSON containing control_actions (list of target tag writes) and reasoning_summary
- Verifies that control actions do not violate safety interlocks.
- Time taken from telemetry receipt to action dispatch • target: 450 • range: 0-2000
- Benchmark: SCADA Autonomous Response Benchmark
- Decision Latency Ms target: 450
- 1 public reference case
- Python execution harness
- Python 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
- OpenAI · Required
- Cognition Devin · Optional
- Azure OpenAI Service · Optional
- Cognition Devin · Optional
- Cognition Devin · Optional
Evaluation contract
- safety_override_check · Weight 1
- decision_latency_ms · Weight 1
Metrics
- decision_latency_ms · Target 450
Benchmarks
- SCADA Autonomous Response Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON array of raw SCADA MQTT alarm packets
Output: JSON containing control_actions (list of target tag writes) and reasoning_summary
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Orchestration
- Cognition Devin · Autonomous Code Engineering
Recipe freeze marker
No recipe freeze time is recorded.