Primary search
V3 full-text search remains the preferred retrieval strategy.
Typo-Tolerant Retrieval
Search normally. Recover intelligently.
V4 keeps V3 indexed full-text search as the primary retrieval path, then activates PostgreSQL trigram similarity when linguistic search cannot recover useful results.
Try
≈
Fuzzy fallback
V4 introduces typo tolerance without replacing full-text search. PostgreSQL first attempts the normal FTS path, then falls back to trigram similarity when needed.
Detect failure
Fuzzy retrieval activates only when the normal search path cannot satisfy the query.
Trigram recovery
Similar character sequences help recover misspelled brands, categories and names.
Running search pipeline
Testing FTS and inspecting fuzzy fallback.
!
Search failed
Something went wrong.
QUERY
Search report
Query
—
RETRIEVAL STRATEGY
Waiting
Search path
V4 exposes whether PostgreSQL returned normal full-text results or activated fuzzy recovery.
The retrieval mode used for this request.
Whether trigram similarity was required.
PostgreSQL FTS or PostgreSQL pg_trgm.
1
→
Full-Text Search
V3 persisted tsvector + GIN
results?
YES
Return FTS results
NO
Activate pg_trgm
FUZZY ANALYSIS
Trigram similarity
Trigram matching compares overlapping three-character sequences rather than linguistic lexemes.
Input
keybord
key
eyb
ybo
bor
ord
≈
Stored value
keyboard
key
eyb
ybo
boa
oar
ard
Product name
similarity(name, query)
Strong fuzzy signal for product-name misspellings.
Brand
similarity(brand, query)
Useful for inputs such as logtech.
Category
similarity(category, query)
Provides another recovery signal for categorical search terms.
GREATEST()
MAX field similarity
The strongest fuzzy field becomes the product's current fuzzy score.
POSTGRESQL
Waiting
Execution report
Human-readable interpretation of
EXPLAIN ANALYZE and
BUFFERS.
DB
Primary operation
—
Run a search to inspect PostgreSQL's selected execution strategy.
PostgreSQL query planning duration.
PostgreSQL server-side execution.
Rows discarded by filtering conditions.
Blocks already present in shared buffers.
Blocks read into PostgreSQL buffers.
Startup timing for the selected node.
Total execution time for that node.
FUZZY PIPELINE
pg_trgm fallback execution
1
→
Input
raw query text
2
→
% operator
similarity threshold
3
→
similarity()
field scores
4
→
GREATEST()
fuzzy_score
5
Sort
fuzzy_score DESC
Raw SQL query +
Raw EXPLAIN response +
BENCHMARK
Version comparison
Search architecture evolution from substring matching to fuzzy recovery.
| Version | Architecture | Results | App time | DB time | Peak memory | Retrieval |
|---|---|---|---|---|---|---|
| V1 | ILIKE substring | — | — | — | — | Sequential scan |
| V2 | Runtime FTS | — | — | — | — | Linguistic scan |
| V3 | Persisted FTS + GIN | — | — | — | — | Indexed FTS |
| V4 | FTS + fuzzy fallback | — | — | — | — | — |
RESULTS
Search results
Results ranked by the active retrieval strategy.
0
shown
EXPLANATION
What changed in V4?
FTS remains primary
Good queries should still use V3's linguistic retrieval and GIN-backed search path.
search_vector @@ tsquery
Fallback activation
Fuzzy matching is an alternate recovery path rather than the default search engine.
fallback_used = true
Trigram matching
pg_trgm compares overlapping character sequences and can recover many spelling mistakes.
similarity()
Similarity threshold
PostgreSQL's percent operator filters candidates based on the configured trigram similarity threshold.
field % query
Why not replace FTS with fuzzy search? +
Linguistic retrieval
Understands lexemes, stemming, stop words, Boolean query structure and field weighting.
Character similarity
Understands that two strings share many character sequences, making it useful for spelling recovery.
Different problems require different
retrieval strategies.
V4 combines them instead of forcing one algorithm to solve both linguistic relevance and typo tolerance.