feat: add cli backend request specs

crates/localllm-cli/src/main.rs

@@ -668,12 +668,74 @@ fn backend_invocation(cli: &Cli) -> BackendInvocation {
 fn backend_request(cli: &Cli) -> BackendRequest {
     match backend_invocation(cli) {
         BackendInvocation::SummaryOnly { command } => BackendRequest::SummaryOnly { command },
-        BackendInvocation::HttpCall { .. } => BackendRequest::SummaryOnly {
+        BackendInvocation::HttpCall { method, url, .. } if method == "GET" => {
-            command: "request-unimplemented".to_owned(),
+            BackendRequest::HttpRequest {
+                method,
+                url,
+                headers: bearer_headers(cli),
+                body: None,
+            }
+        }
+        BackendInvocation::HttpCall { method, url, .. } if method == "POST" => {
+            BackendRequest::HttpRequest {
+                method,
+                url,
+                headers: json_headers(cli),
+                body: Some(bench_request_body(cli)),
+            }
+        }
+        BackendInvocation::HttpCall {
+            method, url, body, ..
+        } => BackendRequest::HttpRequest {
+            method,
+            url,
+            headers: bearer_headers(cli),
+            body,
         },
     }
 }
 
+fn bearer_headers(cli: &Cli) -> Vec<(String, String)> {
+    cli.api_key
+        .as_ref()
+        .map(|api_key| vec![("authorization".to_owned(), format!("Bearer {}", api_key))])
+        .unwrap_or_default()
+}
+
+fn json_headers(cli: &Cli) -> Vec<(String, String)> {
+    let mut headers = bearer_headers(cli);
+    headers.push(("content-type".to_owned(), "application/json".to_owned()));
+    headers
+}
+
+fn bench_request_body(cli: &Cli) -> String {
+    match &cli.command {
+        Commands::Bench(bench) => match &bench.command {
+            BenchSubcommand::Throughput(args) => serde_json::json!({
+                "model": args.model,
+                "stream": false,
+                "temperature": 0.0,
+                "top_p": 1.0,
+                "max_completion_tokens": args.completion_tokens,
+                "messages": [{
+                    "role": "user",
+                    "content": benchmark_prompt(args.prompt_tokens),
+                }],
+            })
+            .to_string(),
+            _ => "{}".to_owned(),
+        },
+        _ => "{}".to_owned(),
+    }
+}
+
+fn benchmark_prompt(prompt_tokens: u32) -> String {
+    std::iter::repeat("benchmark token")
+        .take((prompt_tokens / 2) as usize)
+        .collect::<Vec<_>>()
+        .join(" ")
+}
+
 fn command_summary(command: &Commands) -> String {
     match command {
         Commands::Status => "status".to_owned(),