- use unified response codes for measurement recommend handler
- reject recommend inputs with continuous dots as invalid params
- fix exact and continuation match offsets for recommend results
- preserve correct offsets for fallback recommend scenarios
- use hash-based confirmation for measurement token7 recommendations
- write component and bay token7 hash indexes for HEXISTS validation
- add tests for input validation, hash keys, and recommend offset cases
- return a single measurement recommend payload instead of grouped payloads
- remove recommended_type from response schema and regenerated swagger docs
- normalize Redis recommend results in model layer with stable type ordering
- deduplicate trimmed recommend terms and preserve the maximum response offset