Multimodal Flow Measurement & Sensor Diagnostics Agent with Google ADK
Accurate flow rate measurement in process plants requires combining calibration datasheets, piping diagrams, and ultrasonic sensor signals. Develop a multimodal industrial agent using Google ADK to analyze flow meter imagery, detect installation errors, and compute inline calibration offsets.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a multimodal process engineering agent using Google ADK and Gemini models to evaluate visual flow meter installations and diagnostic data.
How work is evaluated
Evaluates agent ability to detect insufficient straight pipe runs upstream of a flow meter from an engineering diagram input.
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.
installation_validation_check
Ensures undersized straight runs are flagged as invalid.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
error_estimation_accuracy
Accuracy of estimated percentage error compared to standard fluid dynamics equations. • target: 0.95 • range: 0.8-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure Google Agent Development Kit (ADK) with Gemini multimodal tools
Process visual P&ID (Piping & Instrumentation Diagram) images and installation photos
Compute Reynolds number and straight pipe run requirements upstream/downstream
Generate automated calibration correction formulas for differential pressure meters
Reference links and supporting material
Images of piping installations, technical P&ID drawings, and flow meter calibration sheets.
Challenge setup
Evaluates agent ability to detect insufficient straight pipe runs upstream of a flow meter from an engineering diagram input.
Challenge input
JSON containing pipe_diameter_inches, upstream_straight_run_diameters, and meter_type.
Google ADK
Provides native integration with Vertex AI and Gemini multimodal models.
Evaluated output
JSON containing installation_valid, estimated_error_pct, and recommendation.
- Ensures undersized straight runs are flagged as invalid.
- Accuracy of estimated percentage error compared to standard fluid dynamics equations. • target: 0.95 • rang...
- Benchmark: Process Engineering Multimodal Benchmark
- Error Estimation Accuracy target: 0.95
- 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
- Google · Required
- Google ADK · Optional
- Google Jules · Optional
- Google Jules · Optional
Evaluation contract
- installation_validation_check · Weight 1
- error_estimation_accuracy · Weight 1
Metrics
- error_estimation_accuracy · Target 0.95
Benchmarks
- Process Engineering Multimodal Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing pipe_diameter_inches, upstream_straight_run_diameters, and meter_type.
Output: JSON containing installation_valid, estimated_error_pct, and recommendation.
Execution harness: Python
Evaluation tools
- Google ADK · Multimodal Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.