chore(docs): extend transformer scrubbing examples

## About

Clarify that a context transformer can scrub tool returns before they
are sent back to the model, not just plain string prompts.

## Changes

- Extend the ScrubPII example in the README to handle Array prompts.
- Show ScrubPII rebuilding an LLM::Function::Return with scrubbed value data.
- Update the deep dive to call out tool-return scrubbing explicitly.

Author: 0x1eef

README.md

@@ -212,9 +212,20 @@ class ScrubPII
   def scrub(prompt)
     case prompt
     when String then prompt.gsub(EMAIL, "[REDACTED_EMAIL]")
+    when Array then prompt.map { scrub(_1) }
+    when LLM::Function::Return then LLM::Function::Return.new(prompt.id, prompt.name, scrub_value(prompt.value))
     else prompt
     end
   end
+
+  def scrub_value(value)
+    case value
+    when String then value.gsub(EMAIL, "[REDACTED_EMAIL]")
+    when Array then value.map { scrub_value(_1) }
+    when Hash then value.transform_values { scrub_value(_1) }
+    else value
+    end
+  end
 end
 
 ctx = LLM::Context.new(llm)

resources/deepdive.md

@@ -1188,7 +1188,8 @@ This is especially useful when you want context-wide behavior without
 rewriting every `talk` and `respond` call site by hand.
 
 For example, a simple transformer can scrub email addresses before the model
-ever sees them:
+ever sees them. It can also handle `Array` prompts and
+`LLM::Function::Return` objects:
 
 ```ruby
 class ScrubPII
@@ -1203,9 +1204,20 @@ class ScrubPII
   def scrub(prompt)
     case prompt
     when String then prompt.gsub(EMAIL, "[REDACTED_EMAIL]")
+    when Array then prompt.map { scrub(_1) }
+    when LLM::Function::Return then LLM::Function::Return.new(prompt.id, prompt.name, scrub_value(prompt.value))
     else prompt
     end
   end
+
+  def scrub_value(value)
+    case value
+    when String then value.gsub(EMAIL, "[REDACTED_EMAIL]")
+    when Array then value.map { scrub_value(_1) }
+    when Hash then value.transform_values { scrub_value(_1) }
+    else value
+    end
+  end
 end
 
 ctx = LLM::Context.new(llm)