Skip to main content
TEC-01 · Technology & Software

Developer Support Copilot

Outcome​

Give software and AI platform engineers one grounded surface that diagnoses service incidents and platform constraints together: correlate deployment manifests, quota and token telemetry, runbooks and prior tickets so a latency spike or a model capacity wall is explained and remediated with an approved, cited change.

Portfolio attributeValue
Primary personaSoftware engineer and AI platform engineer
Primary aggregateServiceOperationalContext
Cosmos DB containers2
Grounding datacode metadata, architecture standards, deployment manifests, quota and usage telemetry, tickets, runbooks, logs
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
ProductivityIssue-to-diagnosis-ready time15-30% reduction(baseline median issue-to-diagnosis minutes - assisted median minutes) / baseline median minutes * 100
Quality or riskDiagnosis and correlation correctness5-10 percentage-point improvementassisted percentage meeting the diagnosis-correctness rubric - baseline percentage

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: Incident, quota, deployment, and migration evidence supports faster diagnosis preparation for engineering teams.

Pilot guardrail: Engineers approve all production changes, migrations, quota changes, and governance actions.

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 issue-to-diagnosis-ready time and diagnosis and correlation correctness 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. Symptom and telemetry correlationDeveloper support copilotPlatformTelemetry → correlatedSymptomSetPolicy controlled
2. Deployment bisectionChange correlation serviceDeploymentManifest → suspectChangeSetPolicy controlled
3. Prior-art and runbook retrievalRetrieval orchestratorsuspectChangeSet → RunbookChunkPolicy controlled
4. Remediation proposal with blast radiusDeveloper support copilotRunbookChunk → RemediationProposalPolicy controlled
5. Change approval and production executionService owner or change advisory approverRemediationProposal → executedChangeHuman required

Technical architecture​

Technical architecture

Cosmos DB data model​

ContainerHierarchical partition keyDocumentsSearch paths
serviceOperationalContexts/tenantId, /serviceIdService, DeploymentManifest, PlatformTelemetry, IncidentTicket, RemediationProposal/symptomVectorEmbedding, /rationaleVectorEmbedding, /symptomVector, /rationaleVector
serviceRunbookLibrary/tenantId, /procedureCategoryRunbookChunk/chunkVectorEmbedding, /chunkVector

Query contract​

QueryContainerScopePurpose
QuotaHeadroomByModelDeploymentserviceOperationalContextssingle-partitionShow which model deployments are approaching their token or provisioned-throughput ceiling in a region.
DeploymentBisectionWindowserviceOperationalContextssingle-partitionList deployments that landed inside the regression window for bisection.
SimilarPriorIncidentsserviceOperationalContextssingle-partitionFind prior tickets with a semantically similar symptom to reuse a proven resolution.
TierAppropriateRunbookRetrievalserviceRunbookLibrarysingle-partitionRetrieve the remediation procedure valid for this service tier along with its change-control level.

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/technology-software/developer-support-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