Interface PropagateAttributesParams

INTERNAL USE ONLY - For Langfuse experiment framework.

This parameter is used internally by the Langfuse experiment system to propagate experiment context to child spans. It should NOT be used by external code.

If true, propagates attributes using OpenTelemetry baggage for cross-process/service propagation.

Security warning: When enabled, attribute values are added to HTTP headers on ALL outbound requests. Only enable if values are safe to transmit via HTTP headers and you need cross-service tracing.

Attach the processor's skills.available to new spans in this scope. Affects tracing only; explicit skillsAvailable take precedence. Nested scopes can opt out with false. Never sent as an attribute or baggage.

Langfuse environment to associate with all spans in this context.

Must be a lowercase alphanumeric string with optional hyphens or underscores, must be ≤40 characters, and must not start with langfuse. This maps to the first-class langfuse.environment attribute, not to trace metadata.

A propagated environment takes precedence over the default configured on LangfuseSpanProcessor or via LANGFUSE_TRACING_ENVIRONMENT while this propagation scope is active.

Additional key-value metadata to propagate to all spans.

• Every value is JSON-encoded, strings included, the same way as observation metadata (e.g. "prod" → '"prod"', "123" → '"123"', 42 → "42", [1, "a"] → '[1,"a"]'), so the server keeps the string "123" apart from the number 123
• null, undefined, functions and symbols are dropped with a warning
• Encoded values must be ≤200 characters, otherwise they are dropped with a warning
• Use for dimensions like internal correlating identifiers
• AVOID: large payloads, sensitive data

Langfuse prompt to link to observations created within this context.

Accepts a prompt client returned by langfuse.prompt.get(...) or any plain object exposing name (non-empty string) and version (integer) — e.g. { name: "my-prompt", version: 3 }. This is the recommended way to link prompts to generations emitted by auto-instrumentation libraries (e.g. OpenInference, other OTel instrumentations) where you don't create the generation via the Langfuse SDK yourself.

The prompt link is only applied to generation-type observations by the Langfuse backend. Fallback prompts are never linked. An explicit prompt set directly on an observation takes precedence over the propagated one.

Session identifier to associate with all spans in this context. Must be a string ≤200 characters. Use this to group related traces within a user session (e.g., a conversation thread, multi-turn interaction).

Skills available to generations in this scope, provided as names and optional versions. Updates the active observation and propagates to new child spans. Nested lists replace the outer list; an empty array clears it. Local to this process, even when asBaggage is enabled.

List of tags to categorize the group of observations

Trace name to associate with all spans in this context. Must be a string ≤200 characters.

User identifier to associate with all spans in this context. Must be a string ≤200 characters. Use this to track which user generated each trace and enable e.g. per-user cost/performance analysis.

Version identifier for parts of your application that are independently versioned, e.g. agents