Community · E2 · artifact verified
Mine repeating patterns from noisy sequences
Finding and extracting repeating patterns from noisy sequences with Jev: instead of a hand-tuned distance metric, typed questions decide what counts as the same pattern, and matches come back with probabilities instead of thresholds you guess.
01 · Role in the system
What Jev does here
Pattern mining over noisy data usually lives or dies on its similarity metric. Here that judgment is outsourced to typed questions: candidate repeats are asked about - is this the same pattern, does this segment continue it - and answers arrive as calibrated choices and scores, so a match carries its own confidence instead of sitting above an arbitrary cutoff. The TypeScript client wraps the systemone protocol, and the README frames the swap precisely: the model does not generate text, it answers whether structure repeats, which is exactly the question classical algorithms approximate with edit distance. Sequences in, extracted patterns with probabilities out.
02 · Control boundary
Where Jev sits
Jev as the similarity judge in a pattern-mining pipeline: candidate repeats become typed same-pattern questions, calibrated answers replace a hand-tuned distance threshold, and extraction proceeds over the confident matches.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- Per-comparison calls add cost on long sequences; no published scaling numbers.
- Pattern definitions are question-shaped; exotic structures need new criteria.
- New project (0 stars); accuracy not benchmarked against edit-distance baselines.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- hazlema ↗Community · github · public · checked 2026-10-03