Initial commit

crates/agentos-cli/src/bin/agentos-npu-bridge.rs

@@ -0,0 +1,50 @@
+use agentos_npu::{LoopbackServerConfig, bridge_chat};
+use agentos_protocol::ChatRequest;
+use anyhow::{Context, Result, bail};
+use clap::Parser;
+use std::time::Duration;
+use tokio::io::{AsyncReadExt, AsyncWriteExt};
+
+const MAX_REQUEST_BYTES: u64 = 8 * 1024 * 1024;
+
+#[derive(Debug, Parser)]
+#[command(name = "agentos-npu-bridge", version)]
+struct Args {
+    #[arg(long, required = true)]
+    request_json_stdin: bool,
+}
+
+#[tokio::main]
+async fn main() -> Result<()> {
+    let _args = Args::parse();
+    let mut input = Vec::new();
+    tokio::io::stdin()
+        .take(MAX_REQUEST_BYTES + 1)
+        .read_to_end(&mut input)
+        .await
+        .context("read request from stdin")?;
+    if input.len() as u64 > MAX_REQUEST_BYTES {
+        bail!("request exceeds 8 MiB");
+    }
+    let mut request: ChatRequest =
+        serde_json::from_slice(&input).context("parse agentos.chat.v1 request")?;
+    request.validate()?;
+    request.parallel_tool_calls = false;
+    let config = LoopbackServerConfig {
+        endpoint: std::env::var("AGENTOS_NPU_SERVER_URL")
+            .unwrap_or_else(|_| "http://127.0.0.1:8031/v1/chat/completions".into()),
+        model: std::env::var("AGENTOS_NPU_SERVER_MODEL").unwrap_or_else(|_| "agentos-npu".into()),
+        timeout: Duration::from_secs(
+            std::env::var("AGENTOS_NPU_SERVER_TIMEOUT_S")
+                .ok()
+                .and_then(|value| value.parse().ok())
+                .unwrap_or(240),
+        ),
+    };
+    let response = bridge_chat(&config, &request).await?;
+    let output = serde_json::to_vec(&response)?;
+    let mut stdout = tokio::io::stdout();
+    stdout.write_all(&output).await?;
+    stdout.write_all(b"\n").await?;
+    Ok(())
+}