Skip to main content
MED-01 · Media, Entertainment & Gaming

Content Discovery Concierge

Outcome​

Help a viewer or player find something worth their time right now by combining catalogue metadata, transcripts, entitlements, and viewing history into explainable recommendations that respect rights windows and parental controls.

Portfolio attributeValue
Primary personaViewer or player
Primary aggregateViewerDiscoverySession
Cosmos DB containers2
Grounding datametadata, transcripts, reviews, entitlements, history
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
ProductivityDiscovery-to-confirmation time12-22% reduction(baseline median discovery-to-confirmation minutes - assisted median minutes) / baseline median minutes * 100
Quality or riskAccepted recommendations8-14 percentage-point improvementassisted percentage of recommendations accepted - baseline percentage
Business outcomeConfirmed playback or purchase conversion2-6% relative improvement(assisted confirmed actions per qualified discovery session - baseline rate) / baseline rate * 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: Rights-aware discovery can shorten content selection and modestly improve conversion among comparable qualified sessions.

Pilot guardrail: Verify rights, territory, entitlement, availability, and explicit user confirmation before action.

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 discovery-to-confirmation time and accepted recommendations 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. Express what to watch or playViewer or playerNatural-language mood, time budget, or scene memory → Structured discovery intentPolicy controlled
2. Retrieve candidates across metadata and transcriptsRetrieval agentIntent vector and taste profile → Ranked candidate titles and scenesPolicy controlled
3. Enforce entitlements and parental controlsRights and controls serviceCandidate titles, region, and profile settings → Playable, age-appropriate shortlistPolicy controlled
4. Explain why each title fitsDiscovery conciergeShortlist with taste and scene evidence → Recommendations with human-readable reasonsPolicy controlled
5. Confirm playback or purchaseViewerSelected title with any rental or upgrade cost → Playback start or confirmed transactionHuman required

Technical architecture​

Technical architecture

Cosmos DB data model​

ContainerHierarchical partition keyDocumentsSearch paths
viewerDiscoverySessions/tenantId, /viewerProfileIdViewerDiscoverySession, EntitlementGrant, ViewingHistoryEntry, ProfileControlSetting/intentVectorEmbedding, /intentVector
catalogTitleIndex/tenantId, /titleIdTitleMetadata, TranscriptHighlight/synopsisVectorEmbedding, /moodVectorEmbedding, /segmentVectorEmbedding, /synopsisVector, /moodVector, /segmentVector

Query contract​

QueryContainerScopePurpose
Profile taste contextviewerDiscoverySessionssingle-partitionRead viewing history and control settings for the profile in one partition query.
Playable entitlement filterviewerDiscoverySessionssingle-partitionRestrict candidates to titles inside an active rights window for the viewer's region.
Scene-level semantic searchcatalogTitleIndexcross-partition-boundedKnowledge-container query against catalogTitleIndex (bounded fan-out across the /tenantId + /titleId partitions named by the shortlist): find the exact episode or moment a viewer half-remembers, as a bounded fan-out across the title ids returned by the vector shortlist rather than a scan of every viewer partition.
Continue watching resume listviewerDiscoverySessionssingle-partitionSurface partially watched titles with a meaningful resume point.

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/media-entertainment-gaming/content-discovery-concierge
az login
./Deploy.ps1 -Runtime python -Location eastus2

Change python to csharp to deploy the .NET implementation.

Explore the implementation​

Return to all accelerators