fabro/nodes/implement/response.md

All 390 tests pass and clippy is clean with zero warnings.

Here's a summary of the changes made to lib/crates/fabro-llm/src/providers/anthropic.rs:

1. Added output_config field to ApiRequest struct (after thinking):

   #[serde(skip_serializing_if = "Option::is_none")]
   output_config: Option<serde_json::Value>,
   

2. Mapped reasoning_effort → output_config in build_api_request — converts request.reasoning_effort into {"effort": "<value>"} JSON, mirroring how the OpenAI provider maps this field:

   let output_config = request
       .reasoning_effort
       .as_ref()
       .map(|effort| serde_json::json!({"effort": effort}));
   

3. Added two tests:

   • build_api_request_maps_reasoning_effort_to_output_config — verifies reasoning_effort: Some("medium") produces output_config: Some({"effort": "medium"})
   • build_api_request_omits_output_config_when_no_reasoning_effort — verifies reasoning_effort: None produces output_config: None

4. Updated all existing ApiRequest construction sites in tests to include output_config: None.