AI-Powered Blockchain Analytics

Why Do Traditional Blockchain Analytics Fall Short?

What Blockchain Intelligence Can You Extract with AI Embeddings?

This guide demonstrates building enterprise-grade blockchain analytics using SettleMint Integration Studio with OpenAI embeddings and vector search. Transform blockchain data into business intelligence that drives decisions.

What you'll build:

• Semantic smart contract search - Find contracts with similar functionality across all blockchains
• Transaction pattern detection - Identify suspicious activities and compliance violations automatically
• DeFi protocol correlation analysis - Discover hidden risks and opportunities across protocols
• Real-time fraud detection - Match new transactions against known malicious patterns

Enterprise Infrastructure Requirements

Production-Ready AI Analytics Workflows

Pre-built Enterprise Capabilities:

• Real-time fraud detection processing 100,000+ transactions/hour
• Smart contract vulnerability scanning across 50+ blockchain networks
• DeFi protocol correlation analysis tracking systemic risk indicators
• Compliance reporting automation for AML and KYC requirements

Part 1: Building Enterprise Blockchain Intelligence Infrastructure

Step 1: Design Enterprise Vector Database Schema

Step 2: Configure High-Performance Integration Studio Flow

// Enterprise-grade contract processing
{
  "model": "text-embedding-3-large",
  "input": "Contract: ${contract_name}\nNetwork: ${blockchain}\nCode: ${source_code}\nSecurity: ${audit_results}\nGas: ${gas_analysis}",
  "dimensions": 1536,
  "encoding_format": "float"
}

Step 4: Vectorize Data with OpenAI Node

1. Insert an OpenAI Node in the workflow:
   • Use this node to generate vector embeddings for the text data using OpenAI's Embedding API.
   • Configure the OpenAI node to use the appropriate model and input data, such as text-embedding-ada-002.

Step 5: Store Vectors in Hasura with pgvector

1. Add a GraphQL Node to save the vector embeddings and data id in Hasura.
2. Set up a GraphQL Mutation to store the vectors and associated IDs in a table enabled with pgvector.

Example Mutation:

mutation insertVector($id: uuid!, $vector: [Float!]!) {
  insert_vectors(objects: { id: $id, vector: $vector }) {
    affected_rows
  }
}

3. Ensure correct data mapping from the fetched data and vectorized output.

Step 6: Deploy and Test the Workflow

1. Deploy the Flow within Integration Studio and run it to confirm that data is fetched, vectorized, and stored in Hasura.
2. Verify Hasura Data by checking the table to ensure vectorized entries and corresponding IDs are stored correctly.

Part 2: Setting Up a Similarity Search Endpoint

Step 1: Create a POST Endpoint

1. Add an HTTP POST Node to accept a JSON payload with a query string to be vectorized and compared to stored data.

Payload Example:

{
  "query": "input string for similarity search"
}

2. Parse the Request by adding a JSON node to extract the query field from the incoming POST request.

Step 2: Vectorize the Input Query

1. Add an OpenAI Node to convert the incoming query string into a vector representation.

Example Configuration:

Model: text-embedding-ada-002
Input: {{msg.payload.query}}

Step 3: Perform a Similarity Search with Hasura

1. Add a GraphQL Node to perform a vector similarity search within Hasura using the pgvector plugin.
2. Use a GraphQL Query to order results by similarity, returning the top 5 most similar records.

Example Query:

query searchVectors($vector: [Float!]!) {
  vectors(order_by: { vector: { _vector_distance: $vector } }, limit: 5) {
    id
    vector
  }
}

3. Map the vector from the OpenAI node output as the vector input for the Hasura query.

Step 4: Format and Return the Results

1. Add a Function Node to format the response, listing the top 5 matches in a structured JSON format.

Step 5: Test the Flow

1. Deploy the Flow and send a POST request to confirm the similarity search functionality.
2. Verify Response to ensure that the flow accurately returns the top 5 matches from the vectorized data in Hasura.

Next Steps

Now that you have built an AI-powered workflow, here are some blockchain-specific applications you can explore:

Vectorize On-Chain Data

• Index and vectorize smart contract events for similarity-based event monitoring
• Create embeddings from transaction data to detect patterns or anomalies
• Vectorize NFT metadata for content-based recommendations
• Build semantic search for on-chain attestations

Advanced Use Cases

• Combine transaction data with natural language descriptions for enhanced search
• Create AI-powered analytics dashboards using vectorized blockchain metrics
• Implement fraud detection by vectorizing transaction patterns
• Build a semantic search engine for smart contract code and documentation

Integration Ideas

• Connect to multiple blockchain indexers to vectorize data across networks
• Combine off-chain and on-chain data vectors for comprehensive analysis
• Set up automated alerts based on similarity to known patterns
• Create a knowledge base from vectorized blockchain documentation

For further resources, check out:

• SettleMint Integration Studio Documentation
• Node-RED Documentation
• OpenAI API Documentation
• Hasura pgvector Documentation

This guide should enable you to build AI-powered workflows with SettleMint's new OpenAI nodes and pgvector support in Hasura for efficient similarity searches.