Use custom axum fork to fix sse leading space.

Cargo.toml

@@ -14,10 +14,10 @@ exclude = ["assets"]
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
-axum = "0.6.11"
+axum = { git = "https://github.com/cryscan/axum", branch = "sse-leading-space" }
 tower-http = { version = "0.4", features = ["full"] }
 tokio = { version = "1", features = ["full"] }
-web-rwkv = { git = "https://github.com/cryscan/web-rwkv" }
+web-rwkv = "0.1.12"
 memmap = "0.7"
 clap = { version = "4.3", features = ["derive"] }
 serde = { version = "1.0", features = ["derive"] }

src/chat.rs

@@ -223,7 +223,7 @@ pub async fn chat_completions_stream(
                 finish_reason,
                 ..Default::default()
             },
-            Token::Done => return Ok(Event::default().data(" [DONE]")),
+            Token::Done => return Ok(Event::default().data("[DONE]")),
             _ => unreachable!(),
         };
 
@@ -232,7 +232,7 @@ pub async fn chat_completions_stream(
             model: model_name.clone(),
             choices: vec![choice],
         })?;
-        Ok(Event::default().data(format!(" {json}")))
+        Ok(Event::default().data(json))
     });
 
     Sse::new(stream)

src/completion.rs

@@ -176,7 +176,7 @@ pub async fn completions_stream(
                 finish_reason,
                 ..Default::default()
             },
-            Token::Done => return Ok(Event::default().data(" [DONE]")),
+            Token::Done => return Ok(Event::default().data("[DONE]")),
             _ => unreachable!(),
         };
 
@@ -185,7 +185,7 @@ pub async fn completions_stream(
             model: model_name.clone(),
             choices: vec![choice],
         })?;
-        Ok(Event::default().data(format!(" {json}")))
+        Ok(Event::default().data(json))
     });
 
     Sse::new(stream)

src/main.rs

@@ -16,6 +16,7 @@ use std::{
     ffi::OsStr,
     fs::File,
     io::{BufReader, Read, Write},
+    net::SocketAddr,
     path::PathBuf,
     str::FromStr,
 };
@@ -324,9 +325,10 @@ async fn main() -> Result<()> {
         .route("/embeddings", post(embedding::embeddings))
         .route("/v1/embeddings", post(embedding::embeddings))
         .with_state(ThreadState { sender, model_name });
-    axum::Server::bind(&format!("0.0.0.0:{0}", args.port).parse().unwrap())
-        .serve(app.into_make_service())
-        .await?;
+
+    let addr = SocketAddr::from(([127, 0, 0, 1], args.port));
+    let listener = tokio::net::TcpListener::bind(addr).await?;
+    axum::serve(listener, app).await?;
 
     Ok(())
 }