Logfire should let tag LLM spans with a call-type and use-case label, then provide built-in aggregation (count, total/avg tokens, total/avg cost) for each label across traces and time ranges, not just within a single trace as today. This would let AI Engineering see cost and volume per prompt/call type and per overall AI use case without writing custom SQL or exporting data.
Logfire should let tag LLM spans with a call-type and use-case label, then provide built-in aggregation (count, total/avg tokens, total/avg cost) for each label across traces and time ranges, not just within a single trace as today. This would let AI Engineering see cost and volume per prompt/call type and per overall AI use case without writing custom SQL or exporting data.