Pre-filtering
A technique that uses deterministic rules, such as path regular expressions and file-type checks, to exclude diff portions that do not require model processing before invoking a large model, thereby reducing cost and latency.
Timeline (last 90 days)
Pre-filtering rules (such as lock file detection and path skipping) should stay on the client side, but model calls cannot.
Cache pre-filtering decisions are safe; caching verdicts is safe only when all three inputs—modelId, templateHash, and diffHash—are present, and the model identifier must come from configuration with no default value.
There is a significant difference in result quality between pre-filtering (filter then search) and post-filtering (search then filter); post-filtering may result in insufficient recall because a large number of candidates are eliminated after the fact.
All Facts (3)
Pre-filtering rules (such as lock file detection and path skipping) should stay on the client side, but model calls cannot.
50%UnverifiedCache pre-filtering decisions are safe; caching verdicts is safe only when all three inputs—modelId, templateHash, and diffHash—are present, and the model identifier must come from configuration with no default value.
50%UnverifiedThere is a significant difference in result quality between pre-filtering (filter then search) and post-filtering (search then filter); post-filtering may result in insufficient recall because a large number of candidates are eliminated after the fact.
50%