openai: Don't send prompt_cache_key for OpenAI-compatible models (#36231)

Some APIs fail when they get this parameter

Closes #36215

Release Notes:

- Fixed OpenAI-compatible providers that don't support prompt caching
and/or reasoning

crates/language_models/src/provider/cloud.rs

@@ -941,6 +941,7 @@ impl LanguageModel for CloudLanguageModel {
                     request,
                     model.id(),
                     model.supports_parallel_tool_calls(),
+                    model.supports_prompt_cache_key(),
                     None,
                     None,
                 );

crates/language_models/src/provider/open_ai.rs

@@ -370,6 +370,7 @@ impl LanguageModel for OpenAiLanguageModel {
             request,
             self.model.id(),
             self.model.supports_parallel_tool_calls(),
+            self.model.supports_prompt_cache_key(),
             self.max_output_tokens(),
             self.model.reasoning_effort(),
         );
@@ -386,6 +387,7 @@ pub fn into_open_ai(
     request: LanguageModelRequest,
     model_id: &str,
     supports_parallel_tool_calls: bool,
+    supports_prompt_cache_key: bool,
     max_output_tokens: Option<u64>,
     reasoning_effort: Option<ReasoningEffort>,
 ) -> open_ai::Request {
@@ -477,7 +479,11 @@ pub fn into_open_ai(
         } else {
             None
         },
-        prompt_cache_key: request.thread_id,
+        prompt_cache_key: if supports_prompt_cache_key {
+            request.thread_id
+        } else {
+            None
+        },
         tools: request
             .tools
             .into_iter()

crates/language_models/src/provider/open_ai_compatible.rs

@@ -355,10 +355,13 @@ impl LanguageModel for OpenAiCompatibleLanguageModel {
             LanguageModelCompletionError,
         >,
     > {
+        let supports_parallel_tool_call = true;
+        let supports_prompt_cache_key = false;
         let request = into_open_ai(
             request,
             &self.model.name,
-            true,
+            supports_parallel_tool_call,
+            supports_prompt_cache_key,
             self.max_output_tokens(),
             None,
         );

crates/language_models/src/provider/vercel.rs

@@ -355,6 +355,7 @@ impl LanguageModel for VercelLanguageModel {
             request,
             self.model.id(),
             self.model.supports_parallel_tool_calls(),
+            self.model.supports_prompt_cache_key(),
             self.max_output_tokens(),
             None,
         );

crates/language_models/src/provider/x_ai.rs

@@ -359,6 +359,7 @@ impl LanguageModel for XAiLanguageModel {
             request,
             self.model.id(),
             self.model.supports_parallel_tool_calls(),
+            self.model.supports_prompt_cache_key(),
             self.max_output_tokens(),
             None,
         );