Skip to content
TruthVouch Docs

Embeddings & Vector Search

Text embeddings convert text into semantic vectors that capture meaning. This allows TruthVouch to understand that “When was TruthVouch founded?” and “What’s TruthVouch’s founding year?” mean the same thing — enabling fast, intelligent fact-checking even when questions are phrased differently.

How TruthVouch Uses Embeddings

1. Neural Cache (L2)

Semantic search via pgvector:

  • Query converted to embedding
  • Compared against cached truth nugget embeddings
  • Sub-10ms retrieval of semantically similar results

2. Hallucination Detection

NLI scoring uses embeddings:

  • Premise converted to embedding
  • Hypothesis converted to embedding
  • Relationship (entailment/neutral/contradiction) determined
  • 94%+ accuracy hallucination detection

3. Fact Matching

Finding relevant truth nuggets:

  • User query converted to embedding
  • Matched against truth nugget embeddings
  • Returns top K most relevant facts

4. Query Recommendations

Suggest related queries:

  • Current query embedding computed
  • Similar past queries found via vector search
  • Top related queries recommended

How It Works

TruthVouch automatically generates embeddings for your truth nuggets and incoming AI responses. When a new claim arrives, it’s converted to an embedding and compared against your knowledge base using semantic similarity. This happens in milliseconds, enabling real-time hallucination detection.

Enterprise Customization

For organizations with domain-specific terminology, TruthVouch supports custom embedding models. Contact your account team to configure a model optimized for your industry or specialized knowledge domain.

Next Steps