Merge branch 'main' into timkpaine-patch-2

Author: timkpaine

README.md

@@ -4,6 +4,34 @@
   </a>
 </div>
 
+> [!NOTE]
+> ## 🚀 Together Python SDK 2.0 is now available!
+>
+> Check out the new SDK: **[together-py](https://github.com/togethercomputer/together-py)**
+>
+> 📖 **Migration Guide:** [https://docs.together.ai/docs/pythonv2-migration-guide](https://docs.together.ai/docs/pythonv2-migration-guide)
+>
+> ### Install the Beta
+>
+> **Using uv (Recommended):**
+> ```bash
+> # Install uv if you haven't already
+> curl -LsSf https://astral.sh/uv/install.sh | sh
+>
+> # Install together python SDK
+> uv add together --prerelease allow
+>
+> # Or upgrade an existing installation
+> uv sync --upgrade-package together --prerelease allow
+> ```
+>
+> **Using pip:**
+> ```bash
+> pip install --pre together
+> ```
+>
+> This package will be maintained until January 2026.
+
 # Together Python API library
 
 [![PyPI version](https://img.shields.io/pypi/v/together.svg)](https://pypi.org/project/together/)

pyproject.toml

@@ -12,9 +12,9 @@ build-backend = "poetry.masonry.api"
 
 [tool.poetry]
 name = "together"
-version = "1.5.30"
+version = "1.5.33"
 authors = ["Together AI <support@together.ai>"]
-description = "Python client for Together's Cloud Platform!"
+description = "Python client for Together's Cloud Platform! Note: SDK 2.0 is now available at https://github.com/togethercomputer/together-py"
 readme = "README.md"
 license = "Apache-2.0"
 classifiers = [

src/together/__init__.py

@@ -1,5 +1,56 @@
 from __future__ import annotations
 
+import os
+import sys
+
+# =============================================================================
+# SDK 2.0 ANNOUNCEMENT
+# =============================================================================
+_ANNOUNCEMENT_MESSAGE = """
+================================================================================
+Together Python SDK 2.0 is now available!
+
+Install: pip install --pre together
+New SDK: https://github.com/togethercomputer/together-py
+Migration guide: https://docs.together.ai/docs/pythonv2-migration-guide
+
+This package will be maintained until January 2026.
+================================================================================
+"""
+
+# Show info banner (unless suppressed)
+if not os.environ.get("TOGETHER_NO_BANNER"):
+    try:
+        from rich.console import Console
+        from rich.panel import Panel
+
+        console = Console(stderr=True)
+        console.print(
+            Panel(
+                "[bold cyan]Together Python SDK 2.0 is now available![/bold cyan]\n\n"
+                "Install the beta:\n"
+                "[green]pip install --pre together[/green]  or  "
+                "[green]uv add together --prerelease allow[/green]\n\n"
+                "New SDK: [link=https://github.com/togethercomputer/together-py]"
+                "https://github.com/togethercomputer/together-py[/link]\n"
+                "Migration guide: [link=https://docs.together.ai/docs/pythonv2-migration-guide]"
+                "https://docs.together.ai/docs/pythonv2-migration-guide[/link]\n\n"
+                "[dim]This package will be maintained until January 2026.\n"
+                "Set TOGETHER_NO_BANNER=1 to hide this message.[/dim]",
+                title="🚀 New SDK Available",
+                border_style="cyan",
+            )
+        )
+    except Exception:
+        # Fallback for any error (ImportError, OSError in daemons, rich errors, etc.)
+        # Banner display should never break module imports
+        try:
+            print(_ANNOUNCEMENT_MESSAGE, file=sys.stderr)
+        except Exception:
+            pass  # Silently ignore if even stderr is unavailable
+
+# =============================================================================
+
 from contextvars import ContextVar
 from typing import TYPE_CHECKING, Callable
 

src/together/cli/api/chat.py

@@ -2,7 +2,7 @@
 
 import cmd
 import json
-from typing import List, Tuple
+from typing import Any, Dict, List, Tuple
 
 import click
 
@@ -181,6 +181,12 @@ def interactive(
     "--frequency-penalty", type=float, help="Frequency penalty sampling method"
 )
 @click.option("--min-p", type=float, help="Min p sampling")
+@click.option(
+    "--audio-url",
+    type=str,
+    multiple=True,
+    help="Audio URL to attach to the last user message",
+)
 @click.option("--no-stream", is_flag=True, help="Disable streaming")
 @click.option("--logprobs", type=int, help="Return logprobs. Only works with --raw.")
 @click.option("--echo", is_flag=True, help="Echo prompt. Only works with --raw.")
@@ -200,6 +206,7 @@ def chat(
     presence_penalty: float | None = None,
     frequency_penalty: float | None = None,
     min_p: float | None = None,
+    audio_url: List[str] | None = None,
     no_stream: bool = False,
     logprobs: int | None = None,
     echo: bool | None = None,
@@ -210,7 +217,22 @@ def chat(
     """Generate chat completions from messages"""
     client: Together = ctx.obj
 
-    messages = [{"role": msg[0], "content": msg[1]} for msg in message]
+    messages: List[Dict[str, Any]] = [
+        {"role": msg[0], "content": msg[1]} for msg in message
+    ]
+
+    if audio_url and messages:
+        last_msg = messages[-1]
+        if last_msg["role"] == "user":
+            # Convert content to list if it is string
+            if isinstance(last_msg["content"], str):
+                last_msg["content"] = [{"type": "text", "text": last_msg["content"]}]
+
+            # Append audio URLs
+            for url in audio_url:
+                last_msg["content"].append(
+                    {"type": "audio_url", "audio_url": {"url": url}}
+                )
 
     response = client.chat.completions.create(
         model=model,

src/together/cli/api/endpoints.py

@@ -137,8 +137,7 @@ def endpoints(ctx: click.Context) -> None:
     help="Start endpoint in specified availability zone (e.g., us-central-4b)",
 )
 @click.option(
-    "--wait",
-    is_flag=True,
+    "--wait/--no-wait",
     default=True,
     help="Wait for the endpoint to be ready after creation",
 )
@@ -284,7 +283,9 @@ def fetch_and_print_hardware_options(
 @endpoints.command()
 @click.argument("endpoint-id", required=True)
 @click.option(
-    "--wait", is_flag=True, default=True, help="Wait for the endpoint to stop"
+    "--wait/--no-wait",
+    default=True,
+    help="Wait for the endpoint to stop",
 )
 @click.pass_obj
 @handle_api_errors
@@ -307,7 +308,9 @@ def stop(client: Together, endpoint_id: str, wait: bool) -> None:
 @endpoints.command()
 @click.argument("endpoint-id", required=True)
 @click.option(
-    "--wait", is_flag=True, default=True, help="Wait for the endpoint to start"
+    "--wait/--no-wait",
+    default=True,
+    help="Wait for the endpoint to start",
 )
 @click.pass_obj
 @handle_api_errors

src/together/cli/api/finetune.py

@@ -9,14 +9,17 @@
 import click
 from click.core import ParameterSource  # type: ignore[attr-defined]
 from rich import print as rprint
+from rich.json import JSON
 from tabulate import tabulate
 
 from together import Together
-from together.cli.api.utils import BOOL_WITH_AUTO, INT_WITH_MAX
+from together.cli.api.utils import BOOL_WITH_AUTO, INT_WITH_MAX, generate_progress_bar
 from together.types.finetune import (
     DownloadCheckpointType,
     FinetuneEventType,
     FinetuneTrainingLimits,
+    FullTrainingType,
+    LoRATrainingType,
 )
 from together.utils import (
     finetune_price_to_dollars,
@@ -29,13 +32,21 @@
 
 _CONFIRMATION_MESSAGE = (
     "You are about to create a fine-tuning job. "
-    "The cost of your job will be determined by the model size, the number of tokens "
+    "The estimated price of this job is {price}. "
+    "The actual cost of your job will be determined by the model size, the number of tokens "
     "in the training file, the number of tokens in the validation file, the number of epochs, and "
-    "the number of evaluations. Visit https://www.together.ai/pricing to get a price estimate.\n"
+    "the number of evaluations. Visit https://www.together.ai/pricing to learn more about fine-tuning pricing.\n"
+    "{warning}"
     "You can pass `-y` or `--confirm` to your command to skip this message.\n\n"
     "Do you want to proceed?"
 )
 
+_WARNING_MESSAGE_INSUFFICIENT_FUNDS = (
+    "The estimated price of this job is significantly greater than your current credit limit and balance combined. "
+    "It will likely get cancelled due to insufficient funds. "
+    "Consider increasing your credit limit at https://api.together.xyz/settings/profile\n"
+)
+
 
 class DownloadCheckpointTypeChoice(click.Choice):
     def __init__(self) -> None:
@@ -357,12 +368,36 @@ def create(
             "You have specified a number of evaluation loops but no validation file."
         )
 
-    if confirm or click.confirm(_CONFIRMATION_MESSAGE, default=True, show_default=True):
+    finetune_price_estimation_result = client.fine_tuning.estimate_price(
+        training_file=training_file,
+        validation_file=validation_file,
+        model=model,
+        n_epochs=n_epochs,
+        n_evals=n_evals,
+        training_type="lora" if lora else "full",
+        training_method=training_method,
+    )
+
+    price = click.style(
+        f"${finetune_price_estimation_result.estimated_total_price:.2f}",
+        bold=True,
+    )
+
+    if not finetune_price_estimation_result.allowed_to_proceed:
+        warning = click.style(_WARNING_MESSAGE_INSUFFICIENT_FUNDS, fg="red", bold=True)
+    else:
+        warning = ""
+
+    confirmation_message = _CONFIRMATION_MESSAGE.format(
+        price=price,
+        warning=warning,
+    )
+
+    if confirm or click.confirm(confirmation_message, default=True, show_default=True):
         response = client.fine_tuning.create(
             **training_args,
             verbose=True,
         )
-
         report_string = f"Successfully submitted a fine-tuning job {response.id}"
         if response.created_at is not None:
             created_time = datetime.strptime(
@@ -401,6 +436,9 @@ def list(ctx: click.Context) -> None:
                 "Price": f"""${
                     finetune_price_to_dollars(float(str(i.total_price)))
                 }""",  # convert to string for mypy typing
+                "Progress": generate_progress_bar(
+                    i, datetime.now().astimezone(), use_rich=False
+                ),
             }
         )
     table = tabulate(display_list, headers="keys", tablefmt="grid", showindex=True)
@@ -420,7 +458,15 @@ def retrieve(ctx: click.Context, fine_tune_id: str) -> None:
     # remove events from response for cleaner output
     response.events = None
 
-    click.echo(json.dumps(response.model_dump(exclude_none=True), indent=4))
+    rprint(JSON.from_data(response.model_dump(exclude_none=True)))
+    progress_text = generate_progress_bar(
+        response, datetime.now().astimezone(), use_rich=True
+    )
+    status = "Unknown"
+    if response.status is not None:
+        status = response.status.value
+    prefix = f"Status: [bold]{status}[/bold],"
+    rprint(f"{prefix} {progress_text}")
 
 
 @fine_tuning.command()