Revert "fix: Add schema-aware numeric coercion for MCP tool arguments…

[jamadeo]

This is a good change to make but as implemented it calls .list_tools() on every token chunk response from a provider. .list_tools() is potentially very expensive, as it queries all extensions for their tools. This is especially noticeable if you have remote extensions, as they tend to have a higher latency.

reverts #5478

This reverts commit 78fd31a.

[tlongwell-block]

Would this work instead?

diff --git a/crates/goose/src/agents/agent.rs b/crates/goose/src/agents/agent.rs
index 088fba588a7..d8a12c0ea34 100644
--- a/crates/goose/src/agents/agent.rs
+++ b/crates/goose/src/agents/agent.rs
@@ -283,11 +283,11 @@ impl Agent {
     async fn categorize_tools(
         &self,
         response: &Message,
-        _tools: &[rmcp::model::Tool],
+        tools: &[rmcp::model::Tool],
     ) -> ToolCategorizeResult {
         // Categorize tool requests
         let (frontend_requests, remaining_requests, filtered_response) =
-            self.categorize_tool_requests(response).await;
+            self.categorize_tool_requests(response, tools).await;
 
         ToolCategorizeResult {
             frontend_requests,
diff --git a/crates/goose/src/agents/reply_parts.rs b/crates/goose/src/agents/reply_parts.rs
index 40b1a99d6a8..9132364cb25 100644
--- a/crates/goose/src/agents/reply_parts.rs
+++ b/crates/goose/src/agents/reply_parts.rs
@@ -269,9 +269,8 @@ impl Agent {
     pub(crate) async fn categorize_tool_requests(
         &self,
         response: &Message,
+        tools: &[Tool],
     ) -> (Vec<ToolRequest>, Vec<ToolRequest>, Message) {
-        let tools = self.list_tools(None).await;
-
         // First collect all tool requests with coercion applied
         let tool_requests: Vec<ToolRequest> = response
             .content

[jamadeo]

@tlongwell-block yes that would do it. The important thing to avoid is calls to list_tools() that query all MCP servers.

[jamadeo]

@tlongwell-block did you end up making the above change? if so we can close this PR, if not let's merge this if it's still an active issue because it could be quite bad for anyone connected to remote MCPs

[tlongwell-block]

@tlongwell-block did you end up making the above change? if so we can close this PR, if not let's merge this if it's still an active issue because it could be quite bad for anyone connected to remote MCPs

I can open that PR today. I'll link it here when its ready

[alexhancock]

@jamadeo @tlongwell-block Is this one still relevant, or replaced with another PR?

[tlongwell-block]

@alexhancock @jamadeo #6138 should fix this issue