From b6bacd6b4a78849f07cf439fd9f59b06dad20699 Mon Sep 17 00:00:00 2001
From: YoVinchen <gzh298255@gmail.com>
Date: Sat, 11 Apr 2026 16:46:41 +0800
Subject: [PATCH] Stop sending prompt cache keys on Claude chat conversions

Responses conversions still use promptCacheKey, but chat completions now stay a pure shape transform. This keeps Claude -> chat requests aligned with providers that do not understand the field and keeps stream checks consistent with production behavior.

Constraint: Issue #1919 requires removing prompt_cache_key from Claude -> OpenAI Chat requests
Rejected: Add a runtime toggle for chat injection | requested behavior is unconditional removal
Confidence: high
Scope-risk: narrow
Reversibility: clean
Directive: Keep promptCacheKey limited to Claude -> Responses conversions unless a provider-specific contract is proven
Tested: cargo test anthropic_to_openai
Tested: cargo test anthropic_to_responses_with_cache_key
Tested: cargo test transform_claude_request_for_api_format_responses
Not-tested: Full src-tauri test suite
Related: #1919
---
 src-tauri/src/provider.rs                  |  6 +--
 src-tauri/src/proxy/providers/claude.rs    | 13 +++--
 src-tauri/src/proxy/providers/transform.rs | 59 ++++++++--------------
 src-tauri/src/services/stream_check.rs     |  2 +-
 src/types.ts                               |  2 +-
 5 files changed, 31 insertions(+), 51 deletions(-)

diff --git a/src-tauri/src/provider.rs b/src-tauri/src/provider.rs
index 67616853c..4170dd3f9 100644
--- a/src-tauri/src/provider.rs
+++ b/src-tauri/src/provider.rs
@@ -282,9 +282,9 @@ pub struct ProviderMeta {
     /// 是否将 base_url 视为完整 API 端点（不拼接 endpoint 路径）
     #[serde(rename = "isFullUrl", skip_serializing_if = "Option::is_none")]
     pub is_full_url: Option<bool>,
-    /// Prompt cache key for OpenAI-compatible endpoints.
-    /// When set, injected into converted requests to improve cache hit rate.
-    /// If not set, provider ID is used automatically during format conversion.
+    /// Prompt cache key for OpenAI Responses-compatible endpoints.
+    /// When set, injected into converted Responses requests to improve cache hit rate.
+    /// If not set, provider ID is used automatically during Claude -> Responses conversion.
     #[serde(rename = "promptCacheKey", skip_serializing_if = "Option::is_none")]
     pub prompt_cache_key: Option<String>,
     /// 累加模式应用中，该 provider 是否已写入 live config。
diff --git a/src-tauri/src/proxy/providers/claude.rs b/src-tauri/src/proxy/providers/claude.rs
index 52689d3b5..aa9949995 100644
--- a/src-tauri/src/proxy/providers/claude.rs
+++ b/src-tauri/src/proxy/providers/claude.rs
@@ -81,14 +81,13 @@ pub fn transform_claude_request_for_api_format(
     provider: &Provider,
     api_format: &str,
 ) -> Result<serde_json::Value, ProxyError> {
-    let cache_key = provider
-        .meta
-        .as_ref()
-        .and_then(|m| m.prompt_cache_key.as_deref())
-        .unwrap_or(&provider.id);
-
     match api_format {
         "openai_responses" => {
+            let cache_key = provider
+                .meta
+                .as_ref()
+                .and_then(|m| m.prompt_cache_key.as_deref())
+                .unwrap_or(&provider.id);
             // Codex OAuth (ChatGPT Plus/Pro 反代) 需要在请求体里强制 store: false
             // + include: ["reasoning.encrypted_content"]，由 transform 层统一处理。
             let is_codex_oauth = provider
@@ -102,7 +101,7 @@ pub fn transform_claude_request_for_api_format(
                 is_codex_oauth,
             )
         }
-        "openai_chat" => super::transform::anthropic_to_openai(body, Some(cache_key)),
+        "openai_chat" => super::transform::anthropic_to_openai(body),
         _ => Ok(body),
     }
 }
diff --git a/src-tauri/src/proxy/providers/transform.rs b/src-tauri/src/proxy/providers/transform.rs
index 7380fb860..a41d18501 100644
--- a/src-tauri/src/proxy/providers/transform.rs
+++ b/src-tauri/src/proxy/providers/transform.rs
@@ -71,10 +71,8 @@ pub fn resolve_reasoning_effort(body: &Value) -> Option<&'static str> {
     }
 }
 
-/// Anthropic 请求 → OpenAI 请求
-///
-/// `cache_key`: optional prompt_cache_key to inject for improved cache routing
-pub fn anthropic_to_openai(body: Value, cache_key: Option<&str>) -> Result<Value, ProxyError> {
+/// Anthropic 请求 → OpenAI Chat Completions 请求
+pub fn anthropic_to_openai(body: Value) -> Result<Value, ProxyError> {
     let mut result = json!({});
 
     // NOTE: 模型映射由上游统一处理（proxy::model_mapper），格式转换层只做结构转换。
@@ -175,11 +173,6 @@ pub fn anthropic_to_openai(body: Value, cache_key: Option<&str>) -> Result<Value
         result["tool_choice"] = v.clone();
     }
 
-    // Inject prompt_cache_key for improved cache routing on OpenAI-compatible endpoints
-    if let Some(key) = cache_key {
-        result["prompt_cache_key"] = json!(key);
-    }
-
     Ok(result)
 }
 
@@ -569,7 +562,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["model"], "claude-3-opus");
         assert_eq!(result["max_tokens"], 1024);
         assert_eq!(result["messages"][0]["role"], "user");
@@ -585,7 +578,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["messages"][0]["role"], "system");
         assert_eq!(
             result["messages"][0]["content"],
@@ -607,7 +600,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["tools"][0]["type"], "function");
         assert_eq!(result["tools"][0]["function"]["name"], "get_weather");
     }
@@ -627,7 +620,7 @@ mod tests {
             ]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["messages"].as_array().unwrap().len(), 2);
         assert_eq!(result["messages"][0]["role"], "system");
         assert_eq!(
@@ -651,7 +644,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         let msg = &result["messages"][0];
         assert_eq!(msg["role"], "assistant");
         assert!(msg.get("tool_calls").is_some());
@@ -671,7 +664,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         let msg = &result["messages"][0];
         assert_eq!(msg["role"], "tool");
         assert_eq!(msg["tool_call_id"], "call_123");
@@ -743,31 +736,19 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["model"], "gpt-4o");
     }
 
     #[test]
-    fn test_anthropic_to_openai_with_cache_key() {
+    fn test_anthropic_to_openai_does_not_inject_prompt_cache_key() {
         let input = json!({
             "model": "claude-3-opus",
             "max_tokens": 1024,
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, Some("provider-123")).unwrap();
-        assert_eq!(result["prompt_cache_key"], "provider-123");
-    }
-
-    #[test]
-    fn test_anthropic_to_openai_no_cache_key() {
-        let input = json!({
-            "model": "claude-3-opus",
-            "max_tokens": 1024,
-            "messages": [{"role": "user", "content": "Hello"}]
-        });
-
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert!(result.get("prompt_cache_key").is_none());
     }
 
@@ -793,7 +774,7 @@ mod tests {
             }]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         // System message cache_control preserved
         assert_eq!(result["messages"][0]["cache_control"]["type"], "ephemeral");
         // Text block cache_control preserved
@@ -1047,7 +1028,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert!(result.get("reasoning_effort").is_none());
     }
 
@@ -1060,7 +1041,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["reasoning_effort"], "medium");
     }
 
@@ -1073,7 +1054,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["reasoning_effort"], "xhigh");
     }
 
@@ -1086,7 +1067,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["reasoning_effort"], "low");
     }
 
@@ -1099,7 +1080,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["reasoning_effort"], "xhigh");
     }
 
@@ -1111,7 +1092,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert!(result.get("reasoning_effort").is_none());
     }
 
@@ -1124,7 +1105,7 @@ mod tests {
                 "messages": [{"role": "user", "content": "Hello"}]
             });
 
-            let result = anthropic_to_openai(input, None).unwrap();
+            let result = anthropic_to_openai(input).unwrap();
             assert!(
                 result.get("max_tokens").is_none(),
                 "{model} should not have max_tokens"
@@ -1144,7 +1125,7 @@ mod tests {
             "messages": [{"role": "user", "content": "Hello"}]
         });
 
-        let result = anthropic_to_openai(input, None).unwrap();
+        let result = anthropic_to_openai(input).unwrap();
         assert_eq!(result["max_tokens"], 1024);
         assert!(result.get("max_completion_tokens").is_none());
     }
diff --git a/src-tauri/src/services/stream_check.rs b/src-tauri/src/services/stream_check.rs
index 5e2203a05..7fd758efd 100644
--- a/src-tauri/src/services/stream_check.rs
+++ b/src-tauri/src/services/stream_check.rs
@@ -372,7 +372,7 @@ impl StreamCheckService {
             anthropic_to_responses(anthropic_body, Some(&provider.id), is_codex_oauth)
                 .map_err(|e| AppError::Message(format!("Failed to build test request: {e}")))?
         } else if is_openai_chat {
-            anthropic_to_openai(anthropic_body, Some(&provider.id))
+            anthropic_to_openai(anthropic_body)
                 .map_err(|e| AppError::Message(format!("Failed to build test request: {e}")))?
         } else {
             anthropic_body
diff --git a/src/types.ts b/src/types.ts
index 90cb06a22..ec6e9ea5f 100644
--- a/src/types.ts
+++ b/src/types.ts
@@ -166,7 +166,7 @@ export interface ProviderMeta {
   apiKeyField?: ClaudeApiKeyField;
   // 是否将 base_url 视为完整 API 端点（代理直接使用此 URL，不拼接路径）
   isFullUrl?: boolean;
-  // Prompt cache key for OpenAI-compatible endpoints (improves cache hit rate)
+  // Prompt cache key for OpenAI Responses-compatible endpoints (improves cache hit rate)
   promptCacheKey?: string;
   // 供应商类型（用于识别 Copilot 等特殊供应商）
   providerType?: string;