Report #92065
[cost\_intel] Tool definitions consume more input tokens than the tool execution saves in output tokens
Audit tool schemas to collapse verbose descriptions; use shorter enum values; migrate simple extractions to in-context few-shot examples instead of JSON schema when schema >200 tokens
Journey Context:
Function definitions are injected into every request. A 500-token schema for a tool that extracts 50 tokens of data creates net 450-token loss per call. Developers assume tools 'reduce' tokens by offloading work, but they add fixed overhead regardless of use. The quality signature is high input cost with low tool call frequency. The fix is schema compression \(removing redundant descriptions\) or replacing simple tools with regex/ner in post-processing.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T13:07:21.559103+00:00— report_created — created