AI Features

Hypernext integrates with OpenAI-compatible APIs for AI-powered content features.

Configuration

ai:
  enabled: true
  openai:
    baseUrl: "http://localhost:11434/v1"
    apiKey: ""
  models:
    embedding: "nomic-embed-text-v2-moe:latest"
    utility: "llama3.2:1b"
    vision: "llava:7b"
    reasoning: "llama3.1:8b"
  vectorDimensions: 768
  features:
    altText: true
    autoTagging: true
    seoMeta: true
    moderation: true

The default configuration points at a local Ollama instance. For OpenAI, set baseUrl to https://api.openai.com/v1 and provide an API key.

Features

Semantic Search (RAG)

When AI is enabled, documents are embedded into a docs_vec vector table using vec0. The talk_to_docs MCP tool performs RAG search:

1. Embeds the user's query
2. Performs KNN search against document embeddings
3. Generates a natural language answer using the utility model

Auto Alt Text

Generates descriptive alt text for images using the vision model.

Auto Tagging

Suggests tags for new documents based on content analysis.

SEO Meta

Generates SEO-optimized meta descriptions.

Content Moderation

Analyzes comments for semantic spam using the utility model.

Vector Search

The vector table is initialized on startup when AI is enabled:

CREATE VIRTUAL TABLE IF NOT EXISTS docs_vec USING vec0(
  slug TEXT PRIMARY KEY,
  embedding FLOAT[768]
);

MCP Tool

talk_to_docs(query)

Ask a natural language question and get an answer based on document content.