Talon
Use Cases chevron_right RAG Hybrid Search
compare_arrows HYBRID SEARCH ENGINE

Enterprise Knowledge Base RAG Hybrid Search

Fuse BM25 keyword matching with HNSW vector semantic search. RRF algorithm for precision ranking, solving proper noun and semantic ambiguity.

inventory_2 Preprocessing Pipeline

Document Input CHUNKING
Embedding EMBEDDING

text-embedding-3-large

Index Write INDEXING

Auto-build parallel index structures

search
USER QUERY
"Talon RAG 系统的延迟是多少?"
find_in_page

Full-Text Search [FTS]

Algorithm BM25 / TF-IDF
Tokenizer Jieba / IK
Feature High-precision keyword hit
"Talon" "延迟" "是"
hub

Vector Search [Vector]

Index HNSW / IVF
Metric Cosine Similarity
Feature High-recall semantic match
0.892
merge

RRF Weighted Fusion

SCORE FORMULA
1 / (k + rank_i)
Top K
50 Candidates
Final Result
Optimized List
search_query.sql
SELECT * FROM hybrid_search(
    query_text => 'Talon RAG 系统的延迟是多少?',
    vector_col => 'embedding',
    fts_col => 'content_tsvector',
    limit_n => 10,
    rrf_k => 60,
    alpha => 0.5  -- balance keyword vs semantic weight
);

Ready to build your enterprise RAG pipeline?