AI & LLM Integration

Surgewave provides native connectors for AI/LLM workloads, enabling real-time AI pipelines, RAG (Retrieval-Augmented Generation) systems, and multi-agent architectures.

Overview

Surgewave's AI connectors enable:

  • Embedding Generation - Convert text to vectors using OpenAI or local models (Ollama)
  • LLM Processing - Enrich messages with AI-generated content
  • Vector Storage - Store embeddings in vector databases (Qdrant, pgvector)
  • Agent Communication - Event-driven multi-agent architectures

Available Connectors

LLM Providers

Connector Embeddings Completions Description
OpenAI Yes Yes OpenAI API and compatible providers (Azure OpenAI)
Ollama Yes Yes Local LLM inference with no API costs

Vector Databases

Connector Description
Qdrant High-performance vector database with filtering
PostgreSQL pgvector Vector extension for PostgreSQL

Architecture

flowchart TB
  subgraph Cluster["Surgewave Cluster"]
    Docs[documents topic]
    Embs[embeddings topic]
    Enr[enriched-docs topic]
  end
  OpenAI["OpenAI Connector (embeddings)"]
  Qdrant["Qdrant Connector (vectors)"]
  App[App Consumer]

  Docs --> OpenAI --> Qdrant
  Embs --> Qdrant
  Enr --> App

Use Cases

RAG Pipeline

Stream documents through embedding generation and into a vector database for AI retrieval:

{
  "connectors": [
    {
      "name": "embed-documents",
      "config": {
        "connector.class": "OpenAISinkConnector",
        "mode": "embeddings",
        "input.field": "content",
        "output.field": "embedding",
        "topics": "documents"
      }
    },
    {
      "name": "store-vectors",
      "config": {
        "connector.class": "QdrantSinkConnector",
        "collection": "documents",
        "vector.field": "embedding",
        "topics": "embeddings"
      }
    }
  ]
}

Content Enrichment

Add AI-generated summaries, classifications, or translations to streaming data:

{
  "name": "summarize-articles",
  "config": {
    "connector.class": "OllamaSinkConnector",
    "mode": "completions",
    "completions.model": "llama3",
    "system.prompt": "Summarize this article in 2 sentences:",
    "input.field": "body",
    "output.field": "summary",
    "topics": "articles"
  }
}

Multi-Agent Systems

Use Surgewave as the communication backbone for AI agents:

flowchart LR
  A[Agent A] -->|publish| T1[Surgewave Topic]
  T1 -->|consume| B[Agent B]
  B -->|publish| T2[Surgewave Topic]
  T2 -->|consume| A

See Agent Integration for detailed patterns.

Quick Start

1. Generate Embeddings with OpenAI

surgewave connect create openai-embeddings --config '{
  "connector.class": "OpenAISinkConnector",
  "openai.api.key": "${OPENAI_API_KEY}",
  "mode": "embeddings",
  "topics": "documents"
}'

2. Use Local LLM with Ollama

# Start Ollama locally
ollama serve

# Create connector
surgewave connect create ollama-summaries --config '{
  "connector.class": "OllamaSinkConnector",
  "ollama.base.url": "http://localhost:11434",
  "mode": "completions",
  "completions.model": "llama3",
  "topics": "articles"
}'

3. Store Vectors in Qdrant

surgewave connect create qdrant-vectors --config '{
  "connector.class": "QdrantSinkConnector",
  "qdrant.host": "localhost",
  "qdrant.port": "6334",
  "collection": "documents",
  "topics": "embeddings"
}'

Configuration Reference

Common Options

Option Type Default Description
topics string Required Input topics (comma-separated)
mode string embeddings Processing mode: embeddings or completions
input.field string text JSON field containing input text
output.field string embedding JSON field for output
batch.size int 100 Records to batch before processing
retry.max int 3 Maximum retry attempts
retry.backoff.ms int 1000 Backoff between retries

Next Steps

Edit this page