Update version 1.2.1

Co-authored-by: Yunmo Koo <[email protected]>
Co-authored-by: Taebum Kim <[email protected]>
Co-authored-by: Jiwoong Kwon <[email protected]>

friendli/auth.py

@@ -15,6 +15,7 @@
 import friendli
 from friendli.di.injector import get_injector
 from friendli.errors import APIError, AuthorizationError, AuthTokenNotFoundError
+from friendli.logging import logger
 from friendli.utils.fs import get_friendli_directory
 from friendli.utils.request import DEFAULT_REQ_TIMEOUT, decode_http_err
 from friendli.utils.url import URLProvider
@@ -52,16 +53,30 @@ def get_auth_header(
     """
     token_: Optional[str]
 
+    token_from_cfg = get_token(TokenType.ACCESS)
+    token_from_env = friendli.token
+
     if token is not None:
         token_ = token
-    elif friendli.token:
-        token_ = friendli.token
+    elif token_from_env:
+        if token_from_cfg:
+            logger.warning(
+                "You've entered your login information in two places - through the "
+                "'FRIENDLI_TOKEN' environment variable and the 'friendli login' CLI "
+                "command. We will use the access token from the 'FRIENDLI_TOKEN' "
+                "environment variable and ignore the login session details. This might "
+                "lead to unexpected authorization errors. If you prefer to use the "
+                "login session instead, unset the 'FRIENDLI_TOKEN' environment "
+                "variable. If you don't want to see this warning again, run "
+                "'friendli logout' to remove the login session."
+            )
+        token_ = token_from_env
     else:
-        token_ = get_token(TokenType.ACCESS)
+        token_ = token_from_cfg
 
     if token_ is None:
         raise AuthTokenNotFoundError(
-            "Should set FRIENDLI_TOKEN environment variable or sign in with 'friendli login'."
+            "Should set 'FRIENDLI_TOKEN' environment variable or sign in with 'friendli login'."
         )
 
     headers = {"Authorization": f"Bearer {token_}"}

friendli/cli/api/chat_completions.py

@@ -58,6 +58,16 @@ def create(
         min=1,
         help="The maximum number of tokens to generate.",
     ),
+    stop: Optional[List[str]] = typer.Option(
+        None,
+        "--stop",
+        "-S",
+        help=(
+            "When one of the stop phrases appears in the generation result, the API "
+            "will stop generation. The stop phrases are excluded from the result. "
+            "Repeat this option to use multiple stop phrases."
+        ),
+    ),
     temperature: Optional[float] = typer.Option(
         None,
         "--temperature",
@@ -120,6 +130,7 @@ def create(
             presence_penalty=presence_penalty,
             max_tokens=max_tokens,
             n=n,
+            stop=stop,
             temperature=temperature,
             top_p=top_p,
         )
@@ -137,6 +148,7 @@ def create(
             presence_penalty=presence_penalty,
             max_tokens=max_tokens,
             n=n,
+            stop=stop,
             temperature=temperature,
             top_p=top_p,
         )

friendli/cli/api/completions.py

@@ -6,7 +6,7 @@
 
 from __future__ import annotations
 
-from typing import Optional
+from typing import List, Optional
 
 import typer
 
@@ -53,6 +53,16 @@ def create(
         min=1,
         help="The maximum number of tokens to generate.",
     ),
+    stop: Optional[List[str]] = typer.Option(
+        None,
+        "--stop",
+        "-S",
+        help=(
+            "When one of the stop phrases appears in the generation result, the API "
+            "will stop generation. The stop phrases are excluded from the result. "
+            "Repeat this option to use multiple stop phrases."
+        ),
+    ),
     temperature: Optional[float] = typer.Option(
         None,
         "--temperature",
@@ -113,6 +123,7 @@ def create(
             presence_penalty=presence_penalty,
             max_tokens=max_tokens,
             n=n,
+            stop=stop,
             temperature=temperature,
             top_p=top_p,
         )
@@ -130,6 +141,7 @@ def create(
             presence_penalty=presence_penalty,
             max_tokens=max_tokens,
             n=n,
+            stop=stop,
             temperature=temperature,
             top_p=top_p,
         )

friendli/cli/login.py

@@ -0,0 +1,127 @@
+# Copyright (c) 2024-present, FriendliAI Inc. All rights reserved.
+
+"""CLI command to sign in Friendli."""
+
+from __future__ import annotations
+
+import threading
+import time
+import webbrowser
+from contextlib import contextmanager
+from typing import Iterator, Tuple
+
+import typer
+import uvicorn
+from fastapi import FastAPI, HTTPException, Request
+from fastapi.responses import HTMLResponse
+
+from friendli.client.login import LoginClient
+from friendli.di.injector import get_injector
+from friendli.utils.url import URLProvider
+
+server_app = FastAPI()
+
+
+@contextmanager
+def run_server(port: int) -> Iterator[None]:
+    """Run temporary local server to handle SSO redirection."""
+    config = uvicorn.Config(
+        app=server_app, host="127.0.0.1", port=port, log_level="error"
+    )
+    server = uvicorn.Server(config)
+    thread = threading.Thread(target=server.run)
+    thread.start()
+    try:
+        yield
+    finally:
+        server.should_exit = True
+        thread.join()
+
+
+def oauth2_login() -> Tuple[str, str]:
+    """Login with SSO."""
+    injector = get_injector()
+    url_provider = injector.get(URLProvider)
+    authorization_url = url_provider.get_suite_uri("/login/cli")
+
+    access_token = None
+    refresh_token = None
+
+    @server_app.get("/sso")
+    async def callback(request: Request) -> HTMLResponse:
+        nonlocal access_token
+        nonlocal refresh_token
+
+        access_token = request.query_params.get("access_token")
+        refresh_token = request.query_params.get("refresh_token")
+
+        if not access_token:
+            raise HTTPException(
+                status_code=400, detail="Access token not found in cookies"
+            )
+
+        success_page = r"""
+<html lang="en">
+<head>
+  <meta charset="UTF-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1.0">
+  <title>SSO Login Success</title>
+  <style>
+    body {
+      font-family: Arial, sans-serif;
+      display: flex;
+      justify-content: center;
+      align-items: center;
+      height: 100vh;
+      background-color: #f0f0f0;
+    }
+    .message-box {
+      text-align: center;
+      padding: 40px;
+      border: 1px solid #d0d0d0;
+      background-color: #ffffff;
+      box-shadow: 0 0 10px rgba(0,0,0,0.1);
+    }
+  </style>
+</head>
+<body>
+  <div class="message-box">
+    <h1>Authentication was successful</h1>
+    <p>You can now close this window and return to CLI.</p>
+    <p>Redirecting to <a href="https://docs.periflow.ai/">Friendli Documentation</a> in <span id="countdown">10</span> seconds.</p>
+  </div>
+  <script>
+    var timeLeft = 10;
+    var countdownElement = document.getElementById("countdown");
+
+    var timerId = setInterval(function() {
+      timeLeft--;
+      countdownElement.innerHTML = timeLeft;
+      if (timeLeft <= 0) {
+        clearInterval(timerId);
+        window.location.href = "https://docs.periflow.ai/";
+      }
+    }, 1000);
+  </script>
+</body>
+</html>
+"""
+        return HTMLResponse(content=success_page, status_code=200)
+
+    typer.secho(
+        f"Opening browser for authentication: {authorization_url}", fg=typer.colors.BLUE
+    )
+
+    webbrowser.open(authorization_url)
+
+    with run_server(33333):
+        while access_token is None or refresh_token is None:
+            time.sleep(1)
+
+    return access_token, refresh_token
+
+
+def pwd_login(email: str, pwd: str) -> Tuple[str, str]:
+    """Login with email and password."""
+    client = LoginClient()
+    return client.login(email, pwd)