Complete a prefix

Rank next-word candidates with Complete().

const results = chain.Complete("goodbye", {
  limit: 5,
  minCount: 1,
});

Flow

1. probe = Create(prefix).glyph
2. tokens = CreateUnigrams(prefix)   // TextStrip — letters/digits only
3. if tokens.length < order - 1 → return []
4. stateKey = last (order - 1) tokens joined
5. candidates = transitions from stateKey
6. score each candidate by weighted glyph similarity to probe
7. sort by score desc → count desc → token asc
8. apply minCount, then limit

Scoring

Markov structure defines which tokens are valid. Glyph similarity defines rank.

glyphScore = sum(weight_i * CompareGlyphs(probe, source_i).similarity) / sum(weight_i)

Rank factor	Role
score	Primary sort (glyph-guided)
count	Tiebreak when scores are equal
token	Lexical tiebreak for stability

Occurrence count does not drive primary rank. It only filters candidates (minCount) and breaks ties.

Completions vs query

Use	API
Find similar documents	Query
Suggest the next word in a prefix	Completions