Struct LlmMessageSnapshot 

pub struct LlmMessageSnapshot {
    pub input: Vec<MessageSnapshot>,
    pub output_text: Option<String>,
    pub output_tool_calls: Vec<ToolCallSnapshot>,
    pub system_instructions: Option<String>,
}

Full prompt/completion content for one llm.call, captured and credential-scrubbed at the agent-loop boundary so the OTel exporter can emit gen_ai.input.messages / gen_ai.output.messages / gen_ai.system_instructions.

Populated at the agent-loop capture boundary whenever the observability-otel feature is active; None otherwise (other observers and non-OTel builds leave it None). Capture is policy-agnostic: whether the snapshot is actually exported — and at which privacy level (off / redacted / full) — is decided by the owning OtelObserver’s instance content config at the OTel export boundary, not by the capture path.

Fields

input: Vec<MessageSnapshot>

Non-system input messages, in send order.

output_text: Option<String>

Assistant text output, if any. Empty text is captured as None.

output_tool_calls: Vec<ToolCallSnapshot>

Tool calls the assistant emitted this turn.

system_instructions: Option<String>

System prompt, carried separately from input.

Trait Implementations

impl Clone for LlmMessageSnapshot

fn clone(&self) -> LlmMessageSnapshot

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for LlmMessageSnapshot

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.