[8.0.3] fix tools bug + OpenAIProtocol class

Author: yashbonde

docs/changelog.rst

@@ -7,6 +7,13 @@ minor versions.
 
 All relevant steps to be taken will be mentioned here.
 
+8.0.3
+-----
+
+- Fix bug in ``tools`` that was causing ever increasing number of tools in the ``Thread`` object
+- OpenAI protocol abstracted away as a ``OpenAIProtocol`` class in ``tuneapi.apis.openai``. This is to make it easier to
+  add new endpoints in the future.
+
 8.0.2
 -----
 

docs/conf.py

@@ -13,7 +13,7 @@
 project = "tuneapi"
 copyright = "2024-2025, Frello Technologies"
 author = "Frello Technologies"
-release = "8.0.2"
+release = "8.0.3"
 
 # -- General configuration ---------------------------------------------------
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#general-configuration

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "tuneapi"
-version = "8.0.2"
+version = "8.0.3"
 description = "Tune AI APIs."
 authors = ["Frello Technology Private Limited <[email protected]>"]
 license = "MIT"

tuneapi/apis/__init__.py

@@ -1,7 +1,7 @@
 # Copyright © 2024-2025 Frello Technology Private Limited
 
 # model APIs
-from tuneapi.apis.model_openai import Openai, TuneModel, Groq, Mistral
+from tuneapi.apis.model_openai import Openai, TuneModel, Groq, Mistral, OpenAIProtocol
 from tuneapi.apis.model_anthropic import Anthropic
 from tuneapi.apis.model_gemini import Gemini
 from tuneapi.apis.turbo import distributed_chat, distributed_chat_async

tuneapi/apis/model_anthropic.py

@@ -143,7 +143,7 @@ def _process_input(
         # return headers, system.strip(), claude_messages
 
         tools = []
-        if isinstance(chats, tt.Thread):
+        if isinstance(chats, tt.Thread) and chats.tools:
             tools = [x.to_dict() for x in chats.tools]
             for t in tools:
                 t["input_schema"] = t.pop("parameters")
@@ -236,7 +236,7 @@ def chat(
         self,
         chats: tt.Thread | str,
         model: Optional[str] = None,
-        max_tokens: int = 1024,
+        max_tokens: int = 4096,
         temperature: Optional[float] = None,
         token: Optional[str] = None,
         usage: bool = False,
@@ -317,7 +317,7 @@ async def chat_async(
         self,
         chats: tt.Thread | str,
         model: Optional[str] = None,
-        max_tokens: int = 1024,
+        max_tokens: int = 4096,
         temperature: Optional[float] = None,
         token: Optional[str] = None,
         usage: bool = False,

tuneapi/apis/model_gemini.py

@@ -138,7 +138,7 @@ def _process_input(
 
         # create the body
         tools = []
-        if isinstance(chats, tt.Thread):
+        if isinstance(chats, tt.Thread) and chats.tools:
             tools = [x.to_dict() for x in chats.tools]
         extra_headers = extra_headers or self.extra_headers
         if extra_headers:

tuneapi/apis/model_openai.py

@@ -17,38 +17,26 @@
 from tuneapi.apis.turbo import distributed_chat, distributed_chat_async
 
 
-class Openai(tt.ModelInterface):
+class OpenAIProtocol(tt.ModelInterface):
     def __init__(
         self,
-        id: str = "gpt-4o",
-        base_url: str = "https://api.openai.com/v1/chat/completions",
-        extra_headers: Optional[Dict[str, str]] = None,
-        api_token: Optional[str] = None,
-        emebdding_url: Optional[str] = None,
-        image_gen_url: Optional[str] = None,
-        audio_transcribe: Optional[str] = None,
-        audio_gen_url: Optional[str] = None,
+        id: str,
+        base_url: str,
+        extra_headers: Optional[Dict[str, str]],
+        api_token: Optional[str],
+        emebdding_url: Optional[str],
+        image_gen_url: Optional[str],
+        audio_transcribe_url: Optional[str],
+        audio_gen_url: Optional[str],
     ):
         self.model_id = id
         self.base_url = base_url
-        self.api_token = api_token or tu.ENV.OPENAI_TOKEN("")
+        self.api_token = api_token
         self.extra_headers = extra_headers
-        self.emebdding_url = emebdding_url or base_url.replace(
-            "/chat/completions",
-            "/embeddings",
-        )
-        self.image_gen_url = image_gen_url or base_url.replace(
-            "/chat/completions",
-            "/images/generations",
-        )
-        self.audio_transcribe_url = audio_transcribe or base_url.replace(
-            "/chat/completions",
-            "/audio/transcriptions",
-        )
-        self.audio_gen_url = audio_gen_url or base_url.replace(
-            "/chat/completions",
-            "/audio/speech",
-        )
+        self.emebdding_url = emebdding_url
+        self.image_gen_url = image_gen_url
+        self.audio_transcribe_url = audio_transcribe_url
+        self.audio_gen_url = audio_gen_url
 
     def set_api_token(self, token: str) -> None:
         self.api_token = token
@@ -875,7 +863,47 @@ async def text_to_speech_async(
 # Other OpenAI compatible models
 
 
-class Mistral(Openai):
+class Openai(OpenAIProtocol):
+    def __init__(
+        self,
+        id: str = "gpt-4o",
+        base_url: str = "https://api.openai.com/v1/chat/completions",
+        extra_headers: Optional[Dict[str, str]] = None,
+        api_token: Optional[str] = None,
+        emebdding_url: Optional[str] = None,
+        image_gen_url: Optional[str] = None,
+        audio_transcribe: Optional[str] = None,
+        audio_gen_url: Optional[str] = None,
+    ):
+        super().__init__(
+            id=id,
+            base_url=base_url,
+            api_token=api_token or tu.ENV.OPENAI_TOKEN(""),
+            extra_headers=extra_headers,
+            emebdding_url=emebdding_url
+            or base_url.replace(
+                "/chat/completions",
+                "/embeddings",
+            ),
+            image_gen_url=image_gen_url
+            or base_url.replace(
+                "/chat/completions",
+                "/images/generations",
+            ),
+            audio_transcribe_url=audio_transcribe
+            or base_url.replace(
+                "/chat/completions",
+                "/audio/transcriptions",
+            ),
+            audio_gen_url=audio_gen_url
+            or base_url.replace(
+                "/chat/completions",
+                "/audio/speech",
+            ),
+        )
+
+
+class Mistral(OpenAIProtocol):
     """
     A class to interact with Mistral's Large Language Models (LLMs) via their API. Note this class does not contain the
     `embedding` method.
@@ -905,13 +933,17 @@ def __init__(
             base_url=base_url,
             extra_headers=extra_headers,
             api_token=api_token or tu.ENV.MISTRAL_TOKEN(),
+            emebdding_url=None,
+            image_gen_url=None,
+            audio_transcribe_url=None,
+            audio_gen_url=None,
         )
 
     def embedding(*a, **k):
         raise NotImplementedError("Mistral does not support embeddings")
 
 
-class Groq(Openai):
+class Groq(OpenAIProtocol):
     """
     A class to interact with Groq's Large Language Models (LLMs) via their API. Note this class does not contain the
     `embedding` method.
@@ -938,13 +970,17 @@ def __init__(
             base_url=base_url,
             extra_headers=extra_headers,
             api_token=api_token or tu.ENV.GROQ_TOKEN(),
+            emebdding_url=None,
+            image_gen_url=None,
+            audio_transcribe_url=None,
+            audio_gen_url=None,
         )
 
     def embedding(*a, **k):
         raise NotImplementedError("Groq does not support embeddings")
 
 
-class TuneModel(Openai):
+class TuneModel(OpenAIProtocol):
     """
     A class to interact with Groq's Large Language Models (LLMs) via their API.
 
@@ -978,6 +1014,9 @@ def __init__(
             extra_headers=extra_headers,
             api_token=api_token or tu.ENV.TUNEAPI_TOKEN(),
             emebdding_url="https://proxy.tune.app/v1/embeddings",
+            image_gen_url=None,
+            audio_transcribe_url=None,
+            audio_gen_url=None,
         )
 
     def embedding(

tuneapi/types/chats.py

@@ -483,7 +483,7 @@ def __init__(
         self,
         input_tokens: int,
         output_tokens: int,
-        cached_tokens: int,
+        cached_tokens: Optional[int] = 0,
         **kwargs,
     ):
         self.input_tokens = input_tokens
@@ -498,6 +498,20 @@ def __getitem__(self, x):
     def __repr__(self) -> str:
         return f"<Usage: {self.input_tokens} [Cached: {self.cached_tokens}] -> {self.output_tokens}>"
 
+    def __radd__(self, other: "Usage"):
+        return Usage(
+            input_tokens=self.input_tokens + other.input_tokens,
+            output_tokens=self.output_tokens + other.output_tokens,
+            cached_tokens=self.cached_tokens + other.cached_tokens,
+        )
+
+    def __add__(self, other: "Usage"):
+        return Usage(
+            input_tokens=self.input_tokens + other.input_tokens,
+            output_tokens=self.output_tokens + other.output_tokens,
+            cached_tokens=self.cached_tokens + other.cached_tokens,
+        )
+
     def to_json(self, *a, **k) -> str:
         return tu.to_json(self.__dict__, *a, **k)