AI Model Advisor | From Use Case to Model Decision

The idea

Choosing an AI model is becoming a product decision, not just an engineering detail. This prototype explores how a multi-agent workflow can support that decision by combining intake, grounded discovery, comparative scoring, and final recommendation synthesis.

Capture the use case and decision constraints.
Discover current model candidates with grounded search.
Score options across cost, latency, capability, and fit.
Explain the recommendation in language a product or engineering team can act on.

Pipeline flow

1IntakeAgent parses the use case

2SearchAgent performs grounded discovery

3AnalysisAgent scores the candidates

4RecommendationAgent produces the final pick

Cloud Shell terminal showing Spring Boot startup and AI Model Advisor agent logs — **1. Multi-agent orchestration** The system coordinates intake, discovery, analysis, and recommendation agents as separate decision stages.

Cloud Shell terminal showing AI Model Advisor pipeline startup, intake results, search results, and analysis summary — **2. Requirement understanding and discovery** A plain-English RAG chatbot use case is converted into structured requirements and matched against current model candidates.

Cloud Shell terminal showing final AI model recommendation details — **3. Recommendation reasoning** The advisor explains the primary recommendation, runner-up options, cost considerations, and decision trade-offs.

Cloud Shell terminal showing completed recommendation output and build successful message — **4. End-to-end validation** The workflow completes as a working Spring Boot service, showing the concept is already executable rather than just theoretical.