0.39.0

Internals

Removed MetadataValuesRequest and MetadataValuesResponse from roboto.query.
Introduced @experimental decorator for SDK classes, methods and functions whose functionality is incomplete, preview-only or subject to change or removal without notice.

Search and query operations (e.g., RobotoSearch.find_files(), RobotoSearch.find_datasets()) now use eventually consistent reads for improved performance and scalability. Results may not immediately reflect very recent writes (typically within 1 second). If you create or update data and immediately query for it, you may need to add retry logic. See the Eventual Consistency Migration Guide for complete details, affected endpoints, and code examples.

find_similar_signals now supports rate-invariant (multi-scale) search via a new scale parameter accepting a Scale object. Finds a query pattern regardless of how fast or slow it occurs in the target. Each Match carries a scale field for the matched time-scale factor. Use Scale(min, max, steps, spacing) to configure the search grid, or Scale.any() for a wide default range. Distances are normalised so existing max_distance thresholds apply without adjustment.

Fixed _derive_session_status unconditionally returning CLIENT_TOOL_TURN for any completed assistant message ending with a tool_use block. Status derivation now requires explicit client_tool_names to distinguish client tools from server tools, defaulting to ROBOTO_TURN for unrecognised tools.
find_similar_signals no longer raises when a DataFrame contains non-numeric string values (e.g. header rows). Non-convertible rows are silently dropped and logged at INFO; a ValueError is raised only if the entire needle or a target topic collapses entirely to non-numeric data.
Condition.matches() now correctly handles comparing a datetime field in the target dictionary to an ISO 8601 timestamp.