From ae9e4b4b24368403a1cdc94d3be9ef2c9d9d4088 Mon Sep 17 00:00:00 2001
From: Hunter Bown <hmbown@gmail.com>
Date: Thu, 14 May 2026 14:48:24 -0500
Subject: [PATCH] fix(client): omit strict OpenAI-incompatible fields

---
 crates/tui/src/client.rs      | 46 ++++++++++++++++++++++++++++++++---
 crates/tui/src/client/chat.rs |  9 -------
 2 files changed, 43 insertions(+), 12 deletions(-)

diff --git a/crates/tui/src/client.rs b/crates/tui/src/client.rs
index 4744b6b0..50998854 100644
--- a/crates/tui/src/client.rs
+++ b/crates/tui/src/client.rs
@@ -887,10 +887,10 @@ pub(super) fn apply_reasoning_effort(
             | ApiProvider::DeepseekCN
             | ApiProvider::Openrouter
             | ApiProvider::Novita
-            | ApiProvider::Fireworks
             | ApiProvider::Sglang => {
                 body["thinking"] = json!({ "type": "disabled" });
             }
+            ApiProvider::Fireworks => {}
             // vLLM is an OpenAI-protocol server, not an Anthropic-protocol one.
             // For Qwen3 / DeepSeek-R1 / other reasoning models hosted via vLLM,
             // the canonical OpenAI extension to disable thinking is
@@ -917,11 +917,13 @@ pub(super) fn apply_reasoning_effort(
             | ApiProvider::DeepseekCN
             | ApiProvider::Openrouter
             | ApiProvider::Novita
-            | ApiProvider::Fireworks
             | ApiProvider::Sglang => {
                 body["reasoning_effort"] = json!("high");
                 body["thinking"] = json!({ "type": "enabled" });
             }
+            ApiProvider::Fireworks => {
+                body["reasoning_effort"] = json!("high");
+            }
             ApiProvider::Vllm => {
                 body["chat_template_kwargs"] = json!({
                     "enable_thinking": true,
@@ -941,11 +943,13 @@ pub(super) fn apply_reasoning_effort(
             | ApiProvider::DeepseekCN
             | ApiProvider::Openrouter
             | ApiProvider::Novita
-            | ApiProvider::Fireworks
             | ApiProvider::Sglang => {
                 body["reasoning_effort"] = json!("max");
                 body["thinking"] = json!({ "type": "enabled" });
             }
+            ApiProvider::Fireworks => {
+                body["reasoning_effort"] = json!("max");
+            }
             ApiProvider::Vllm => {
                 body["chat_template_kwargs"] = json!({
                     "enable_thinking": true,
@@ -1918,6 +1922,21 @@ mod tests {
         );
     }
 
+    #[test]
+    fn reasoning_effort_uses_openai_compatible_shape_for_fireworks() {
+        let mut body = json!({});
+        apply_reasoning_effort(&mut body, Some("max"), ApiProvider::Fireworks);
+
+        assert_eq!(
+            body.get("reasoning_effort").and_then(Value::as_str),
+            Some("max")
+        );
+        assert!(
+            body.get("thinking").is_none(),
+            "Fireworks strict-validates OpenAI-compatible requests and rejects top-level thinking"
+        );
+    }
+
     #[test]
     fn chat_parser_accepts_nvidia_nim_reasoning_field() -> Result<()> {
         let response = parse_chat_message(&json!({
@@ -2055,6 +2074,27 @@ mod tests {
         }
     }
 
+    #[test]
+    fn chat_tool_wire_shape_omits_anthropic_only_metadata() {
+        let tool = Tool {
+            tool_type: Some("function".to_string()),
+            name: "mcp_read_resource".to_string(),
+            description: "Read resource".to_string(),
+            input_schema: json!({"type": "object", "properties": {}}),
+            allowed_callers: Some(vec!["direct".to_string()]),
+            defer_loading: Some(false),
+            input_examples: Some(vec![json!({"uri": "file://example"})]),
+            strict: None,
+            cache_control: None,
+        };
+
+        let encoded = tool_to_chat_for_base_url(&tool, "https://api.fireworks.ai/inference/v1");
+
+        assert!(encoded.get("allowed_callers").is_none());
+        assert!(encoded.get("defer_loading").is_none());
+        assert!(encoded.get("input_examples").is_none());
+    }
+
     #[test]
     fn chat_messages_drop_thinking_only_assistant_for_non_reasoning_model() {
         let message = Message {
diff --git a/crates/tui/src/client/chat.rs b/crates/tui/src/client/chat.rs
index 9f3b8e10..770cbdef 100644
--- a/crates/tui/src/client/chat.rs
+++ b/crates/tui/src/client/chat.rs
@@ -1396,15 +1396,6 @@ pub(super) fn tool_to_chat(tool: &Tool) -> Value {
             "parameters": tool.input_schema,
         }
     });
-    if let Some(allowed_callers) = &tool.allowed_callers {
-        value["allowed_callers"] = json!(allowed_callers);
-    }
-    if let Some(defer_loading) = tool.defer_loading {
-        value["defer_loading"] = json!(defer_loading);
-    }
-    if let Some(input_examples) = &tool.input_examples {
-        value["input_examples"] = json!(input_examples);
-    }
     if let Some(strict) = tool.strict
         && let Some(function) = value.get_mut("function")
     {