A multi-agent investigation system built with LangGraph and SAP Cloud SDK for AI to solve an art theft case. This TypeScript implementation provides a sophisticated agent orchestration system that mirrors the functionality of the Python CrewAI solution.

The solution uses three specialized agents working in sequence:

1. Insurance Appraiser Agent 🎨

   • Uses the RPT-1 model to predict missing insurance values
   • Analyzes stolen artwork data and completes appraisals
   • Tool: call_rpt1

2. Evidence Analyst Agent 🔍

   • Searches through document repositories using the grounding service
   • Analyzes evidence for each suspect
   • Tool: call_grounding_service

3. Lead Detective Agent 🕵️

   • Synthesizes findings from both agents
   • Identifies the most likely culprit
   • Provides comprehensive case conclusion

• Node.js v18+ and npm
• SAP AI Core account with:
  • OAuth credentials (client ID and secret)
  • RPT-1 model deployment
  • Grounding service pipeline configured
• OpenAI API key or SAP AI Core foundation model access

npm install

Copy the example environment file and fill in your credentials:

cp .env.example .env

Edit .env with your actual values:

# SAP AI Core Configuration
AICORE_CLIENT_ID=your_client_id
AICORE_CLIENT_SECRET=your_client_secret
AICORE_AUTH_URL=https://your-auth-url.authentication.eu10.hana.ondemand.com/oauth/token
AICORE_RESOURCE_GROUP=default

# RPT-1 Model Deployment
RPT1_DEPLOYMENT_URL=https://your-deployment-url/v2/inference/deployments/your-deployment-id

# Grounding Service
GROUNDING_PIPELINE_ID=your-pipeline-id

# LLM Configuration
MODEL_NAME=gpt-4
OPENAI_API_KEY=your_openai_api_key

npm run build

npm run dev

Or using the compiled version:

npm start

The system will:

1. Initialize the three agents
2. Run the appraiser agent to predict missing insurance values
3. Run the evidence analyst to search for suspect evidence
4. Run the lead detective to synthesize findings and identify the culprit
5. Output a comprehensive investigation report

solution/
├── src/
│   ├── investigatorCrew.ts   # LangGraph orchestration and agent definitions
│   ├── main.ts                # Entry point
│   ├── payload.ts             # Stolen items data
│   ├── rptClient.ts           # RPT-1 API client
│   ├── tools.ts               # LangGraph tools for RPT-1 and grounding
│   └── types.ts               # TypeScript type definitions
├── .env.example               # Environment variable template
├── package.json               # Dependencies and scripts
├── tsconfig.json              # TypeScript configuration
└── README.md                  # This file

The investigation workflow is implemented as a LangGraph state graph with sequential execution:

START → Appraiser → Evidence Analyst → Lead Detective → END

Each node in the graph:

• Receives the current state
• Executes its specialized task
• Updates the state with its findings
• Passes control to the next node

The agent state includes:

• payload: Stolen items data with prediction placeholders
• suspect_names: List of suspects to investigate
• appraisal_result: Output from the appraiser agent
• evidence_analysis: Output from the evidence analyst
• final_conclusion: Final report from the lead detective
• messages: Conversation history

1. call_rpt1: Calls the RPT-1 model deployment via SAP AI Core
2. call_grounding_service: Queries the document grounding service for evidence

This TypeScript solution provides equivalent functionality to the Python CrewAI implementation:

If you see OAuth token errors:

• Verify your AICORE_CLIENT_ID and AICORE_CLIENT_SECRET
• Check that AICORE_AUTH_URL is correct
• Ensure your credentials have the necessary permissions

If the RPT-1 call fails:

• Verify the RPT1_DEPLOYMENT_URL is correct
• Check that the deployment is running in SAP AI Core
• Ensure the AICORE_RESOURCE_GROUP matches your deployment

If evidence search fails:

• Verify the GROUNDING_PIPELINE_ID is correct
• Ensure the grounding pipeline is deployed and running
• Check that documents are uploaded to the vector database

• LangGraph Documentation
• SAP AI SDK Documentation
• SAP AI Core Documentation

See the LICENSE file in the root of the repository.

This is an educational project for SAP CodeJam. Feel free to explore and learn from the code!