use hermesllm in arch gateway for both stream and non stream messages

crates/common/Cargo.toml

@@ -18,6 +18,7 @@ serde_json = "1.0"
 hex = "0.4.3"
 urlencoding = "2.1.3"
 url = "2.5.4"
+hermesllm = { version = "0.1.0", path = "../hermesllm" }
 
 [dev-dependencies]
 pretty_assertions = "1.4.1"

crates/common/src/errors.rs

@@ -1,6 +1,7 @@
 use proxy_wasm::types::Status;
 
 use crate::{api::open_ai::ChatCompletionChunkResponseError, ratelimit};
+use hermesllm::providers::openai::types::{OpenAIError};
 
 #[derive(thiserror::Error, Debug)]
 pub enum ClientError {
@@ -39,4 +40,6 @@ pub enum ServerError {
     BadRequest { why: String },
     #[error("error in streaming response")]
     Streaming(#[from] ChatCompletionChunkResponseError),
+    #[error("error parsing openai message: {0}")]
+    OpenAIPError(#[from] OpenAIError),
 }