tokevenTokivenWrapper
The main class that wraps your AI provider client.
from tokeven import TokivenWrapper
tracked = TokivenWrapper(
client, # Your Anthropic/OpenAI/Google client
token="tkv_ing_...", # Ingest token (or set TOKEVEN_INGEST_TOKEN env)
session_id="my-session", # Optional: group calls into sessions
batch_size=10, # Optional: events per batch (default 10)
flush_interval=5, # Optional: seconds between flushes (default 5)
)Configuration options
| Option | Default | Description |
|---|---|---|
token | env var | Your write-only ingest token |
session_id | auto | Groups related calls together |
batch_size | 10 | Events batched before sending |
flush_interval | 5 | Seconds between batch flushes |
enabled | true | Toggle tracking on/off |
Event schema
Each tracked call generates an event with:
- •
event_id— client-generated UUID for idempotency - •
model— model ID used - •
input_tokens— prompt token count - •
output_tokens— completion token count - •
cache_read_tokens— tokens served from cache - •
cache_write_tokens— tokens written to cache - •
latency_ms— request duration - •
occurred_at— ISO 8601 timestamp