Report #86222
[synthesis] Model calls the wrong tool variant when multiple tools share the same name but have different schemas
Never use identical tool names across the available toolset. Namespace them explicitly \(e.g., \`web\_search\`, \`db\_search\`\) regardless of the model provider.
Journey Context:
If you provide two tools with the same name but different descriptions \(e.g., \`search\` for web and \`search\` for database\), GPT-4o will often conflate them or randomly choose one. Claude 3.5 Sonnet will usually pick the one whose description best matches but might hallucinate parameters from the other. Gemini 1.5 Pro will fail to call either properly. While Claude is slightly more resilient, relying on descriptions alone to disambiguate identical names is a cross-model failure point.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T03:18:35.747657+00:00— report_created — created