Patch changes
- Updated dependencies: posthog@7.27.0
$feature_flag_called events now send a reduced property set for flags not linked to an experiment when the server enables minimization. Custom flag_definition_cache providers receive an additional minimal_flag_called_events key, and all $feature_flag_called events carry a $feature_flag_has_experiment boolean when the server reports it.
Added a label option to Prompts.get() to fetch the prompt version a label currently points to, with separate caching and a resolved label field on the result. Older PostHog servers ignore the parameter and return the latest version.
Events captured with capture_exception now include the active OpenTelemetry span's $trace_id and $span_id.
Feature flag events now include a $feature_flag_has_experiment boolean property when the server reports whether the flag is linked to an experiment. The property is omitted on older server deployments.
Backend services can now record statsd-style metrics through the posthog.metrics API, with pre-aggregation, OTLP/JSON flushing, and a cardinality guardrail. No OpenTelemetry setup required.
Django middleware now sends the request user agent as $raw_user_agent, the standardized property PostHog's server-side classification reads.
Malformed flag-dependency conditions now evaluate locally as no-match, reverting a 7.22.1 regression that fell back to the /flags endpoint on every evaluation and could massively increase billable request volume.
Pre-calculated total cost from OpenAI Agents Responses API usage is now captured automatically.
Fixed a bug where local evaluation of flag dependencies with a "flag_evaluates_to: false" condition never matched, forcing the dependent flag to false for every locally-evaluated user.
Added a README file so PyPI renders a project description instead of showing none.