Event Schema
All events share a common envelope. Provider-specific fields are merged in depending on the event type and endpoint.
Common fields (all events)
| Field |
Type |
Description |
event_id |
uuid |
Groups .before, .after, .error for the same call |
event_type |
string |
e.g. chat.completions.create.after |
model |
string |
LLM model identifier as passed by the caller |
tags |
object |
Caller-supplied key/value metadata |
After events (after)
| Field |
Type |
Description |
elapsed_ms |
float |
Wall-clock latency of the LLM call in milliseconds |
prompt_tokens |
int | null |
Input token count (chat, text completions, embeddings) |
completion_tokens |
int | null |
Output token count (chat, text completions) |
total_tokens |
int | null |
Total tokens (embeddings) |
input_tokens |
int | null |
Input tokens (Responses API) |
output_tokens |
int | null |
Output tokens (Responses API) |
cached_tokens |
int | null |
Cached input tokens (Responses API) |
char_count |
int | null |
Input character count (TTS — billed by character) |