Honor configured rerank top_k in web_search defaults
invoke() always uses args.top_k (defaulting to 5) when reranking, so WebSearchToolConfig.rerank.top_k from tools_config is effectively ignored unless the model explicitly supplies top_k in each tool call. This causes configured rerank depth to be silently bypassed, which can skew result quality and performance in production setups relying on declarative tool config.
Useful? React with 👍 / 👎.
Originally posted by @chatgpt-codex-connector[bot] in #58 (comment)