refactor(llm): clean up provider layer — fix leak, deduplicate, optimize

CHANGELOG.md

@@ -26,6 +26,12 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 - Replace doom-loop string history with content hash comparison (#645)
 - Return &'static str from detect_image_mime with case-insensitive matching (#646)
 - Replace block_on in history persist with fire-and-forget async spawn (#647)
+- Change `LlmProvider::name()` from `&'static str` to `&str`, eliminating `Box::leak` memory leak in CompatibleProvider (#633)
+- Extract rate-limit retry helper `send_with_retry()` in zeph-llm, deduplicating 3 retry loops (#634)
+- Extract `sse_to_chat_stream()` helpers shared by Claude and OpenAI providers (#635)
+- Replace double `AnyProvider::clone()` in `embed_fn()` with single `Arc` clone (#636)
+- Add `with_client()` builder to ClaudeProvider and OpenAiProvider for shared `reqwest::Client` (#637)
+- Cache `JsonSchema` per `TypeId` in `chat_typed` to avoid per-call schema generation (#638)
 
 ### Fixed
 - False positive: "sudoku" no longer matched by "sudo" blocked pattern (word-boundary matching)

crates/zeph-llm/src/any.rs

@@ -48,11 +48,11 @@ pub enum AnyProvider {
 
 impl AnyProvider {
     /// Return a cloneable closure that calls `embed()` on this provider.
-    pub fn embed_fn(&self) -> impl Fn(&str) -> crate::provider::EmbedFuture {
-        let provider = self.clone();
+    pub fn embed_fn(&self) -> impl Fn(&str) -> crate::provider::EmbedFuture + Send + Sync {
+        let provider = std::sync::Arc::new(self.clone());
         move |text: &str| -> crate::provider::EmbedFuture {
+            let p = std::sync::Arc::clone(&provider);
             let owned = text.to_owned();
-            let p = provider.clone();
             Box::pin(async move { p.embed(&owned).await })
         }
     }
@@ -62,7 +62,7 @@ impl AnyProvider {
     /// Returns an error if the provider fails or the response cannot be parsed.
     pub async fn chat_typed_erased<T>(&self, messages: &[Message]) -> Result<T, crate::LlmError>
     where
-        T: DeserializeOwned + JsonSchema,
+        T: DeserializeOwned + JsonSchema + 'static,
     {
         delegate_provider!(self, |p| p.chat_typed::<T>(messages).await)
     }
@@ -119,7 +119,7 @@ impl LlmProvider for AnyProvider {
         delegate_provider!(self, |p| p.supports_embeddings())
     }
 
-    fn name(&self) -> &'static str {
+    fn name(&self) -> &str {
         delegate_provider!(self, |p| p.name())
     }
 

crates/zeph-llm/src/candle_provider/mod.rs

@@ -163,7 +163,9 @@ impl LlmProvider for CandleProvider {
 
     async fn embed(&self, text: &str) -> Result<Vec<f32>, LlmError> {
         let Some(ref embed_model) = self.embed_model else {
-            return Err(LlmError::EmbedUnsupported { provider: "candle" });
+            return Err(LlmError::EmbedUnsupported {
+                provider: "candle".into(),
+            });
         };
         let model = embed_model.clone();
         let text = text.to_owned();
@@ -176,7 +178,7 @@ impl LlmProvider for CandleProvider {
         self.embed_model.is_some()
     }
 
-    fn name(&self) -> &'static str {
+    fn name(&self) -> &str {
         "candle"
     }
 }