feat: Removed the server functionality

src/client/bedrock.rs

@@ -518,13 +518,7 @@ fn extract_chat_completions(data: &Value) -> Result<ChatCompletionsOutput> {
         bail!("Invalid response data: {data}");
     }
 
-    let output = ChatCompletionsOutput {
-        text,
-        tool_calls,
-        id: None,
-        input_tokens: data["usage"]["inputTokens"].as_u64(),
-        output_tokens: data["usage"]["outputTokens"].as_u64(),
-    };
+    let output = ChatCompletionsOutput { text, tool_calls };
     Ok(output)
 }
 

src/client/claude.rs

@@ -2,10 +2,10 @@ use super::*;
 
 use crate::utils::strip_think_tag;
 
-use anyhow::{bail, Context, Result};
+use anyhow::{Context, Result, bail};
 use reqwest::RequestBuilder;
 use serde::Deserialize;
-use serde_json::{json, Value};
+use serde_json::{Value, json};
 
 const API_BASE: &str = "https://api.anthropic.com/v1";
 
@@ -301,7 +301,7 @@ pub fn claude_build_chat_completions_body(
 								}
 						})
 					.collect();
-		}
+    }
     Ok(body)
 }
 
@@ -353,9 +353,6 @@ pub fn claude_extract_chat_completions(data: &Value) -> Result<ChatCompletionsOu
     let output = ChatCompletionsOutput {
         text: text.to_string(),
         tool_calls,
-        id: data["id"].as_str().map(|v| v.to_string()),
-        input_tokens: data["usage"]["input_tokens"].as_u64(),
-        output_tokens: data["usage"]["output_tokens"].as_u64(),
     };
     Ok(output)
 }

src/client/cohere.rs

@@ -244,12 +244,6 @@ fn extract_chat_completions(data: &Value) -> Result<ChatCompletionsOutput> {
     if text.is_empty() && tool_calls.is_empty() {
         bail!("Invalid response data: {data}");
     }
-    let output = ChatCompletionsOutput {
-        text,
-        tool_calls,
-        id: data["id"].as_str().map(|v| v.to_string()),
-        input_tokens: data["usage"]["billed_units"]["input_tokens"].as_u64(),
-        output_tokens: data["usage"]["billed_units"]["output_tokens"].as_u64(),
-    };
+    let output = ChatCompletionsOutput { text, tool_calls };
     Ok(output)
 }

src/client/common.rs

@@ -21,7 +21,7 @@ use std::sync::LazyLock;
 use std::time::Duration;
 use tokio::sync::mpsc::unbounded_channel;
 
-const MODELS_YAML: &str = include_str!("../../models.yaml");
+pub const MODELS_YAML: &str = include_str!("../../models.yaml");
 
 pub static ALL_PROVIDER_MODELS: LazyLock<Vec<ProviderModels>> = LazyLock::new(|| {
     Config::local_models_override()
@@ -47,8 +47,6 @@ pub trait Client: Sync + Send {
 
     fn model(&self) -> &Model;
 
-    fn model_mut(&mut self) -> &mut Model;
-
     fn build_client(&self) -> Result<ReqwestClient> {
         let mut builder = ReqwestClient::builder();
         let extra = self.extra_config();
@@ -291,9 +289,6 @@ pub struct ChatCompletionsData {
 pub struct ChatCompletionsOutput {
     pub text: String,
     pub tool_calls: Vec<ToolCall>,
-    pub id: Option<String>,
-    pub input_tokens: Option<u64>,
-    pub output_tokens: Option<u64>,
 }
 
 impl ChatCompletionsOutput {
@@ -341,7 +336,6 @@ pub type RerankOutput = Vec<RerankResult>;
 #[derive(Debug, Deserialize)]
 pub struct RerankResult {
     pub index: usize,
-    pub relevance_score: f64,
 }
 
 pub type PromptAction<'a> = (&'a str, &'a str, Option<&'a str>, bool);

src/client/macros.rs

@@ -159,10 +159,6 @@ macro_rules! client_common_fns {
         fn model(&self) -> &Model {
             &self.model
         }
-
-        fn model_mut(&mut self) -> &mut Model {
-            &mut self.model
-        }
     };
 }
 

src/client/model.rs

@@ -118,14 +118,6 @@ impl Model {
         }
     }
 
-    pub fn data(&self) -> &ModelData {
-        &self.data
-    }
-
-    pub fn data_mut(&mut self) -> &mut ModelData {
-        &mut self.data
-    }
-
     pub fn description(&self) -> String {
         match self.model_type() {
             ModelType::Chat => {

src/client/openai.rs

@@ -389,13 +389,7 @@ pub fn openai_extract_chat_completions(data: &Value) -> Result<ChatCompletionsOu
     } else {
         text.to_string()
     };
-    let output = ChatCompletionsOutput {
-        text,
-        tool_calls,
-        id: data["id"].as_str().map(|v| v.to_string()),
-        input_tokens: data["usage"]["prompt_tokens"].as_u64(),
-        output_tokens: data["usage"]["completion_tokens"].as_u64(),
-    };
+    let output = ChatCompletionsOutput { text, tool_calls };
     Ok(output)
 }
 

src/client/stream.rs

@@ -56,7 +56,6 @@ impl SseHandler {
     }
 
     pub fn tool_call(&mut self, call: ToolCall) -> Result<()> {
-        // debug!("HandleCall: {:?}", call);
         self.tool_calls.push(call);
         Ok(())
     }
@@ -65,10 +64,6 @@ impl SseHandler {
         self.abort_signal.clone()
     }
 
-    pub fn tool_calls(&self) -> &[ToolCall] {
-        &self.tool_calls
-    }
-
     pub fn take(self) -> (String, Vec<ToolCall>) {
         let Self {
             buffer, tool_calls, ..

src/client/vertexai.rs

@@ -296,13 +296,7 @@ fn gemini_extract_chat_completions_text(data: &Value) -> Result<ChatCompletionsO
             bail!("Invalid response data: {data}");
         }
     }
-    let output = ChatCompletionsOutput {
-        text,
-        tool_calls,
-        id: None,
-        input_tokens: data["usageMetadata"]["promptTokenCount"].as_u64(),
-        output_tokens: data["usageMetadata"]["candidatesTokenCount"].as_u64(),
-    };
+    let output = ChatCompletionsOutput { text, tool_calls };
     Ok(output)
 }