ai-rag
Retrieval Augmented Generation (RAG) for PascalAI. Build knowledge indexes from documents, search semantically, and generate answers grounded in your data.
Install
ppm install ai-rag
Usage Example
uses ai_rag;
// Create a knowledge index
var idx := CreateIndex('company_knowledge');
// Index documents
IndexDocument(idx, 'docs/product.txt');
IndexDocument(idx, 'docs/faq.md');
IndexDocument(idx, 'docs/pricing.html');
// Semantic search
var results := Search(idx, 'How much does the Pro plan cost?');
for r in results do
WriteLn(r.Text + ' (score: ' + FloatToStr(r.Score) + ')');
// Re-rank results
var reranked := ReRank(results, 'Pro plan pricing');
// RAG-powered answer generation
var question := 'What is included in the Pro plan?';
var answer := RAGComplete(idx, question);
WriteLn(answer);
Features
CreateIndex: create a named vector knowledge index
IndexDocument: embed and store documents (txt, md, html)
Search: semantic similarity search over indexed documents
RAGComplete: LLM answer generation grounded in retrieved context
ReRank: cross-encoder re-ranking for higher precision
Chunk size and overlap configuration
Persistent indexes across program restarts