Skip to main content
AGR-01 · Agriculture & Food

Farm Agronomy Copilot

Outcome​

Help the agronomist turn per-field agronomic history, soil tests, in-season imagery and weather into a defensible field-level recommendation for nutrition, irrigation or crop protection, with label restrictions and re-entry intervals enforced before any prescription is written.

Portfolio attributeValue
Primary personaFarmer or agronomist
Primary aggregateFieldSeasonRecord
Cosmos DB containers2
Grounding datafield records, weather, soil, imagery, guidance
Development estimateUSD 60-220 monthly dev estimate
Runtime optionsPython FastAPI and C# ASP.NET Core

Productivity hypothesis​

These measures reflect the workflow, evidence, decision boundary, and industry risk for this use case.

Impact typeMeasureConservative internal assumptionCustomer measurement
ProductivityTask cycle time10-20% reduction(baseline median task minutes - assisted median task minutes) / baseline median task minutes * 100
Quality or riskEvidence discovery time15-25% reduction(baseline median evidence-search minutes - assisted median evidence-search minutes) / baseline median evidence-search minutes * 100

Important: Conservative internal pilot assumptions only; not public benchmarks, commitments, or expected customer results. Validate against a comparable baseline or matched control cohort with mandatory human review.

Rationale: Farmers and agronomists can assemble field records and guidance faster, but local conditions require professional judgment.

Pilot guardrail: Keep recommendations advisory and disclose weather, soil, imagery, and local-context freshness limitations.

Customer validation plan​

  1. Baseline the current workflow with representative users and tasks.
  2. Run the accelerator on the same task mix with human review retained.
  3. Measure task cycle time and evidence discovery time with the documented formulas.
  4. Replace internal assumptions with customer-observed results before a business case.

Why Azure Cosmos DB​

This AI solution needs an operational data plane, not only a vector index:

  • Hybrid vector and full-text retrieval combines semantic recall with exact identifiers, codes, names, and domain language.
  • Hierarchical partition keys and metadata predicates keep tenant and domain boundaries in the normal query path.
  • Operational JSON, source chunks, workflow and agent state, feedback, and audit records remain together for low-latency reads and governed writes.
  • Change feed reacts to source and operational updates without polling.
  • Autoscale throughput, configurable consistency, and multi-region distribution support variable demand and enterprise availability.

Domain workflow​

Domain workflow

StageActorInput → outputApproval
1. Field condition sensingImagery, weather and sensor connectorsScoutingObservation → fieldConditionSnapshotPolicy controlled
2. Agronomic diagnosisFarm agronomy copilotfieldConditionSnapshot → diagnosisFindingPolicy controlled
3. Label and threshold groundingRetrieval orchestratordiagnosisFinding → AgronomicGuidanceChunkPolicy controlled
4. Prescription drafting with restriction checksFarm agronomy copilotAgronomicGuidanceChunk → PrescriptionDraftPolicy controlled
5. Agronomist approval and machine releaseCertified agronomist or growerPrescriptionDraft → ApplicationRecordHuman required

Technical architecture​

Technical architecture

Cosmos DB data model​

ContainerHierarchical partition keyDocumentsSearch paths
fieldSeasonRecords/tenantId, /fieldIdField, SoilTestResult, ScoutingObservation, PrescriptionDraft, ApplicationRecord/rationaleVectorEmbedding, /rationaleVector
agronomicLabelLibrary/tenantId, /cropCodeAgronomicGuidanceChunk/chunkVectorEmbedding, /chunkVector

Query contract​

QueryContainerScopePurpose
FieldsAtThresholdThisWeekfieldSeasonRecordscross-partition-boundedList fields whose scouting severity crossed the action threshold for the current growth stage. Intentional cross-partition scouting roll-up: bounded by the tenantId prefix, a one-week observation window and the watch-list pest codes.
MultiYearSoilTrendfieldSeasonRecordssingle-partitionCompare nutrient levels by zone across seasons to justify a variable-rate nutrition plan.
LabelAndThresholdRetrievalagronomicLabelLibrarysingle-partitionRetrieve product label restrictions and extension thresholds for the crop, pest and region in play.
PendingPrescriptionsByWindowfieldSeasonRecordscross-partition-boundedShow drafted prescriptions whose application window opens soonest and still await approval. Intentional cross-partition approval queue: bounded by the tenantId prefix, the pending approval state and the open application window.

Python and C# services route each document type to its declared container and hierarchical partition key. ETag-based writes protect mutable aggregates, while append-only records and bounded cross-partition queries preserve scale.

Azure deployment​

Azure deployment

Clone the source repository, authenticate with Azure, and run:

Set-Location accelerators/agriculture-food/farm-agronomy-copilot
az login
./Deploy.ps1 -Runtime python -Location eastus2

Change python to csharp to deploy the .NET implementation.

Explore the implementation​

Return to all accelerators