Cost allocation with compression

Per-team cost tracking

def compress_with_allocation(context, query, team_id, project_id):
    result = comp.compress(context, query)
    log_compression(
        team=team_id,
        project=project_id,
        original_tokens=result.original_tokens,
        kept_tokens=result.kept_tokens,
        saved=result.tokens_removed
    )
    return result

# Later: query the logs for cost allocation
def team_costs(team_id, period):
    entries = query_logs(f"team={team_id} AND date >= {period}")
    total_saved = sum(e.saved * MODEL_COST_PER_TOKEN for e in entries)
    return total_saved

Frequently asked questions

How granular should allocation be?

Per team and per project is usually sufficient. Per-user allocation adds complexity without proportional benefit.

Can I show savings reports to leadership?

Yes. Aggregate compression logs into a monthly report showing total tokens saved and dollar amounts.