Skip to content

Neural Cache

TruthVouch uses an intelligent multi-layer cache to return verified facts in milliseconds instead of seconds. When a query matches something you’ve already verified — exactly or semantically — the cached result is returned instantly, saving time and reducing LLM costs.

How It Works

Layer 1: Exact Match Cache When the same query is asked repeatedly, TruthVouch returns the cached answer in milliseconds.

Layer 2: Semantic Cache When a query is paraphrased differently (“When was TruthVouch founded?” vs “What’s TruthVouch’s founding year?”), the system recognizes it as semantically identical and returns the cached result.

Layer 3: LLM Fallback Completely new queries are sent to LLMs and verified against your truth nuggets. The result is cached for future similar queries.

Benefits

  • Speed: Cached responses return in milliseconds vs 100+ ms for LLM calls
  • Cost Reduction: Up to 80% fewer LLM calls means lower API bills
  • Scalability: Serve more queries with the same infrastructure

Automatic Cache Invalidation

When you update a truth nugget, affected cached results are automatically cleared. The next query will re-verify against the updated truth.

Manual Cache Management

You can manually clear cache through Dashboard → Settings → Cache Performance:

  • Clear specific queries
  • Clear by truth nugget category
  • Full cache reset

Monitoring

View cache performance in your dashboard to see:

  • Hit rate trends
  • Top cached queries
  • Most-verified truth nuggets
  • Cost savings over time

Next Steps

  • Truth Nuggets: Create the knowledge base that powers caching
  • Embeddings: Understand semantic matching
  • Dashboard: Monitor cache performance and configure settings