hack/tools/scripts/llm-key-hunter/results/_cache/content/ahnuf-karim-chowdhury_car-dealership-ai-voice-assistant.json

{"repo": "ahnuf-karim-chowdhury/car-dealership-ai-voice-assistant", "files": {"ai/Lib/site-packages/livekit/plugins/openai/llm.py": {"c64bfd78ded7309c2e37311407723a1dcc718404": {"t": 1786068484.640317, "n": 29991, "h": "6b942407773ed29b", "c": "# Copyright 2023 LiveKit, Inc.\n#\n\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n#     http://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n\nfrom __future__ import annotations\n\nimport asyncio\nimport datetime\nimport os\nfrom dataclasses import dataclass\nfrom typing import Any, Literal, MutableSet, Union\n\nimport aiohttp\nimport httpx\nfrom livekit.agents import (\n    APIConnectionError,\n    APIStatusError,\n    APITimeoutError,\n    llm,\n)\nfrom livekit.agents.llm import (\n    LLMCapabilities,\n    ToolChoice,\n    _create_ai_function_info,\n)\nfrom livekit.agents.types import DEFAULT_API_CONNECT_OPTIONS, APIConnectOptions\n\nimport openai\nfrom openai.types.chat import ChatCompletionChunk, ChatCompletionMessageParam\nfrom openai.types.chat.chat_completion_chunk import Choice\n\nfrom ._oai_api import build_oai_function_description\nfrom .log import logger\nfrom .models import (\n    CerebrasChatModels,\n    ChatModels,\n    DeepSeekChatModels,\n    GroqChatModels,\n    OctoChatModels,\n    PerplexityChatModels,\n    TelnyxChatModels,\n    TogetherChatModels,\n    VertexModels,\n    XAIChatModels,\n)\nfrom .utils import AsyncAzureADTokenProvider, build_oai_message\n\n\n@dataclass\nclass LLMOptions:\n    model: str | ChatModels\n    user: str | None\n    temperature: float | None\n    parallel_tool_calls: bool | None\n    tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\"\n    store: bool | None = None\n    metadata: dict[str, str] | None = None\n\n\nclass LLM(llm.LLM):\n    def __init__(\n        self,\n        *,\n        model: str | ChatModels = \"gpt-4o\",\n        api_key: str | None = None,\n        base_url: str | None = None,\n        user: str | None = None,\n        client: openai.AsyncClient | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n        store: bool | None = None,\n        metadata: dict[str, str] | None = None,\n    ) -> None:\n        \"\"\"\n        Create a new instance of OpenAI LLM.\n\n        ``api_key`` must be set to your OpenAI API key, either using the argument or by setting the\n        ``OPENAI_API_KEY`` environmental variable.\n        \"\"\"\n        super().__init__(\n            capabilities=LLMCapabilities(\n                supports_choices_on_int=True,\n                requires_persistent_functions=False,\n            )\n        )\n\n        self._opts = LLMOptions(\n            model=model,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n            store=store,\n            metadata=metadata,\n        )\n        self._client = client or openai.AsyncClient(\n            api_key=api_key,\n            base_url=base_url,\n            max_retries=0,\n            http_client=httpx.AsyncClient(\n                timeout=httpx.Timeout(connect=15.0, read=5.0, write=5.0, pool=5.0),\n                follow_redirects=True,\n                limits=httpx.Limits(\n                    max_connections=50,\n                    max_keepalive_connections=50,\n                    keepalive_expiry=120,\n                ),\n            ),\n        )\n        self._running_fncs: MutableSet[asyncio.Task[Any]] = set()\n\n    @staticmethod\n    def with_azure(\n        *,\n        model: str | ChatModels = \"gpt-4o\",\n        azure_endpoint: str | None = None,\n        azure_deployment: str | None = None,\n        api_version: str | None = None,\n        api_key: str | None = None,\n        azure_ad_token: str | None = None,\n        azure_ad_token_provider: AsyncAzureADTokenProvider | None = None,\n        organization: str | None = None,\n        project: str | None = None,\n        base_url: str | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        This automatically infers the following arguments from their corresponding environment variables if they are not provided:\n        - `api_key` from `AZURE_OPENAI_API_KEY`\n        - `organization` from `OPENAI_ORG_ID`\n        - `project` from `OPENAI_PROJECT_ID`\n        - `azure_ad_token` from `AZURE_OPENAI_AD_TOKEN`\n        - `api_version` from `OPENAI_API_VERSION`\n        - `azure_endpoint` from `AZURE_OPENAI_ENDPOINT`\n        \"\"\"\n\n        azure_client = openai.AsyncAzureOpenAI(\n            max_retries=0,\n            azure_endpoint=azure_endpoint,\n            azure_deployment=azure_deployment,\n            api_version=api_version,\n            api_key=api_key,\n            azure_ad_token=azure_ad_token,\n            azure_ad_token_provider=azure_ad_token_provider,\n            organization=organization,\n            project=project,\n            base_url=base_url,\n        )  # type: ignore\n\n        return LLM(\n            model=model,\n            client=azure_client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_cerebras(\n        *,\n        model: str | CerebrasChatModels = \"llama3.1-8b\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.cerebras.ai/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of Cerebras LLM.\n\n        ``api_key`` must be set to your Cerebras API key, either using the argument or by setting\n        the ``CEREBRAS_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"CEREBRAS_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_vertex(\n        *,\n        model: str | VertexModels = \"google/gemini-2.0-flash-exp\",\n        project_id: str | None = None,\n        location: str = \"us-central1\",\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of VertexAI LLM.\n\n        `GOOGLE_APPLICATION_CREDENTIALS` environment variable must be set to the path of the service account key file.\n        \"\"\"\n        logger.warning(\n            \"`openai.LLM.with_vertex()` is deprecated. Use `google.LLM()` instead.\"\n        )\n        project_id = project_id\n        location = location\n        _gac = os.environ.get(\"GOOGLE_APPLICATION_CREDENTIALS\")\n        if _gac is None:\n            logger.warning(\n                \"`GOOGLE_APPLICATION_CREDENTIALS` environment variable is not set. please set it to the path of the service account key file. Otherwise, use any of the other Google Cloud auth methods.\"\n            )\n\n        try:\n            from google.auth._default_async import default_async\n            from google.auth.transport._aiohttp_requests import Request\n        except ImportError:\n            raise ImportError(\n                \"Google Auth dependencies not found. Please install with: `pip install livekit-plugins-openai[vertex]`\"\n            )\n\n        class AuthTokenRefresher(openai.AsyncClient):\n            def __init__(self, **kwargs: Any) -> None:\n                self.creds, self.project = default_async(\n                    scopes=[\"https://www.googleapis.com/auth/cloud-platform\"]\n                )\n                project = project_id or self.project\n                base_url = f\"https://{location}-aiplatform.googleapis.com/v1beta1/projects/{project}/locations/{location}/endpoints/openapi\"\n                kwargs.update({\"base_url\": base_url})\n                super().__init__(api_key=\"DUMMY\", **kwargs)\n                self.refresh_threshold = 600  # 10 minutes\n\n            def _token_needs_refresh(self) -> bool:\n                if not self.creds or not self.creds.valid:\n                    return True\n                expiry = self.creds.expiry\n                if expiry is None:\n                    return True\n                remaining = (expiry - datetime.datetime.utcnow()).total_seconds()\n                return remaining < self.refresh_threshold\n\n            async def _refresh_credentials(self) -> None:\n                if self.creds and self.creds.valid and not self._token_needs_refresh():\n                    return\n                async with aiohttp.ClientSession(auto_decompress=False) as session:\n                    auth_req = Request(session=session)\n                    await self.creds.refresh(auth_req)\n                self.api_key = self.creds.token\n\n        client = AuthTokenRefresher(\n            max_retries=0,\n            http_client=httpx.AsyncClient(\n                timeout=httpx.Timeout(connect=15.0, read=5.0, write=5.0, pool=5.0),\n                follow_redirects=True,\n                limits=httpx.Limits(\n                    max_connections=50,\n                    max_keepalive_connections=50,\n                    keepalive_expiry=120,\n                ),\n            ),\n        )\n\n        vertex_llm = LLM(\n            model=model,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n        vertex_llm._capabilities = llm.LLMCapabilities(supports_choices_on_int=False)\n        return vertex_llm\n\n    @staticmethod\n    def with_fireworks(\n        *,\n        model: str = \"accounts/fireworks/models/llama-v3p3-70b-instruct\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.fireworks.ai/inference/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of Fireworks LLM.\n\n        ``api_key`` must be set to your Fireworks API key, either using the argument or by setting\n        the ``FIREWORKS_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"FIREWORKS_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_x_ai(\n        *,\n        model: str | XAIChatModels = \"grok-2-public\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.x.ai/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of XAI LLM.\n\n        ``api_key`` must be set to your XAI API key, either using the argument or by setting\n        the ``XAI_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"XAI_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_groq(\n        *,\n        model: str | GroqChatModels = \"llama3-8b-8192\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.groq.com/openai/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of Groq LLM.\n\n        ``api_key`` must be set to your Groq API key, either using the argument or by setting\n        the ``GROQ_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"GROQ_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_deepseek(\n        *,\n        model: str | DeepSeekChatModels = \"deepseek-chat\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.deepseek.com/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of DeepSeek LLM.\n\n        ``api_key`` must be set to your DeepSeek API key, either using the argument or by setting\n        the ``DEEPSEEK_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"DEEPSEEK_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_octo(\n        *,\n        model: str | OctoChatModels = \"llama-2-13b-chat\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://text.octoai.run/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of OctoAI LLM.\n\n        ``api_key`` must be set to your OctoAI API key, either using the argument or by setting\n        the ``OCTOAI_TOKEN`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"OCTOAI_TOKEN\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_ollama(\n        *,\n        model: str = \"llama3.1\",\n        base_url: str | None = \"http://localhost:11434/v1\",\n        client: openai.AsyncClient | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of Ollama LLM.\n        \"\"\"\n\n        return LLM(\n            model=model,\n            api_key=\"ollama\",\n            base_url=base_url,\n            client=client,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_perplexity(\n        *,\n        model: str | PerplexityChatModels = \"llama-3.1-sonar-small-128k-chat\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.perplexity.ai\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of PerplexityAI LLM.\n\n        ``api_key`` must be set to your Perplexity API key, either using the argument or by setting\n        the ``PERPLEXITY_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"PERPLEXITY_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_together(\n        *,\n        model: str | TogetherChatModels = \"meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.together.xyz/v1\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of TogetherAI LLM.\n\n        ``api_key`` must be set to your TogetherAI API key, either using the argument or by setting\n        the ``TOGETHER_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"TOGETHER_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def with_telnyx(\n        *,\n        model: str | TelnyxChatModels = \"meta-llama/Meta-Llama-3.1-70B-Instruct\",\n        api_key: str | None = None,\n        base_url: str | None = \"https://api.telnyx.com/v2/ai\",\n        client: openai.AsyncClient | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        \"\"\"\n        Create a new instance of Telnyx LLM.\n\n        ``api_key`` must be set to your Telnyx API key, either using the argument or by setting\n        the ``TELNYX_API_KEY`` environmental variable.\n        \"\"\"\n        api_key = _get_api_key(\"TELNYX_API_KEY\", api_key)\n        return LLM(\n            model=model,\n            api_key=api_key,\n            base_url=base_url,\n            client=client,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    @staticmethod\n    def create_azure_client(\n        *,\n        model: str | ChatModels = \"gpt-4o\",\n        azure_endpoint: str | None = None,\n        azure_deployment: str | None = None,\n        api_version: str | None = None,\n        api_key: str | None = None,\n        azure_ad_token: str | None = None,\n        azure_ad_token_provider: AsyncAzureADTokenProvider | None = None,\n        organization: str | None = None,\n        project: str | None = None,\n        base_url: str | None = None,\n        user: str | None = None,\n        temperature: float | None = None,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]] = \"auto\",\n    ) -> LLM:\n        logger.warning(\"This alias is deprecated. Use LLM.with_azure() instead\")\n        return LLM.with_azure(\n            model=model,\n            azure_endpoint=azure_endpoint,\n            api_version=api_version,\n            api_key=api_key,\n            azure_ad_token=azure_ad_token,\n            azure_ad_token_provider=azure_ad_token_provider,\n            organization=organization,\n            project=project,\n            base_url=base_url,\n            user=user,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n    def chat(\n        self,\n        *,\n        chat_ctx: llm.ChatContext,\n        conn_options: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n        fnc_ctx: llm.FunctionContext | None = None,\n        temperature: float | None = None,\n        n: int | None = 1,\n        parallel_tool_calls: bool | None = None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]]\n        | None = None,\n    ) -> \"LLMStream\":\n        if parallel_tool_calls is None:\n            parallel_tool_calls = self._opts.parallel_tool_calls\n\n        if tool_choice is None:\n            tool_choice = self._opts.tool_choice\n\n        if temperature is None:\n            temperature = self._opts.temperature\n\n        return LLMStream(\n            self,\n            client=self._client,\n            model=self._opts.model,\n            user=self._opts.user,\n            chat_ctx=chat_ctx,\n            fnc_ctx=fnc_ctx,\n            conn_options=conn_options,\n            n=n,\n            temperature=temperature,\n            parallel_tool_calls=parallel_tool_calls,\n            tool_choice=tool_choice,\n        )\n\n\nclass LLMStream(llm.LLMStream):\n    def __init__(\n        self,\n        llm: LLM,\n        *,\n        client: openai.AsyncClient,\n        model: str | ChatModels,\n        user: str | None,\n        chat_ctx: llm.ChatContext,\n        conn_options: APIConnectOptions,\n        fnc_ctx: llm.FunctionContext | None,\n        temperature: float | None,\n        n: int | None,\n        parallel_tool_calls: bool | None,\n        tool_choice: Union[ToolChoice, Literal[\"auto\", \"required\", \"none\"]],\n    ) -> None:\n        super().__init__(\n            llm, chat_ctx=chat_ctx, fnc_ctx=fnc_ctx, conn_options=conn_options\n        )\n        self._client = client\n        self._model = model\n        self._llm: LLM = llm\n\n        self._user = user\n        self._temperature = temperature\n        self._n = n\n        self._parallel_tool_calls = parallel_tool_calls\n        self._tool_choice = tool_choice\n\n    async def _run(self) -> None:\n        if hasattr(self._llm._client, \"_refresh_credentials\"):\n            await self._llm._client._refresh_credentials()\n\n        # current function call that we're waiting for full completion (args are streamed)\n        # (defined inside the _run method to make sure the state is reset for each run/attempt)\n        self._oai_stream: openai.AsyncStream[ChatCompletionChunk] | None = None\n        self._tool_call_id: str | None = None\n        self._fnc_name: str | None = None\n        self._fnc_raw_arguments: str | None = None\n        self._tool_index: int | None = None\n        retryable = True\n\n        try:\n            if self._fnc_ctx and len(self._fnc_ctx.ai_functions) > 0:\n                tools = [\n                    build_oai_function_description(fnc, self._llm._capabilities)\n                    for fnc in self._fnc_ctx.ai_functions.values()\n                ]\n            else:\n                tools = None\n\n            opts: dict[str, Any] = {\n                \"tools\": tools,\n                \"parallel_tool_calls\": self._parallel_tool_calls if tools else None,\n                \"tool_choice\": (\n                    {\"type\": \"function\", \"function\": {\"name\": self._tool_choice.name}}\n                    if isinstance(self._tool_choice, ToolChoice)\n                    else self._tool_choice\n                )\n                if tools is not None\n                else None,\n                \"temperature\": self._temperature,\n                \"metadata\": self._llm._opts.metadata,\n                \"store\": self._llm._opts.store,\n                \"n\": self._n,\n                \"stream\": True,\n                \"stream_options\": {\"include_usage\": True},\n                \"user\": self._user or openai.NOT_GIVEN,\n            }\n            # remove None values from the options\n            opts = _strip_nones(opts)\n\n            messages = _build_oai_context(self._chat_ctx, id(self))\n            stream = await self._client.chat.completions.create(\n                messages=messages,\n                model=self._model,\n                **opts,\n            )\n\n            async with stream:\n                async for chunk in stream:\n                    for choice in chunk.choices:\n                        chat_chunk = self._parse_choice(chunk.id, choice)\n                        if chat_chunk is not None:\n                            retryable = False\n                            self._event_ch.send_nowait(chat_chunk)\n\n                    if chunk.usage is not None:\n                        usage = chunk.usage\n                        self._event_ch.send_nowait(\n                            llm.ChatChunk(\n                                request_id=chunk.id,\n                                usage=llm.CompletionUsage(\n                                    completion_tokens=usage.completion_tokens,\n                                    prompt_tokens=usage.prompt_tokens,\n                                    total_tokens=usage.total_tokens,\n                                ),\n                            )\n                        )\n\n        except openai.APITimeoutError:\n            raise APITimeoutError(retryable=retryable)\n        except openai.APIStatusError as e:\n            raise APIStatusError(\n                e.message,\n                status_code=e.status_code,\n                request_id=e.request_id,\n                body=e.body,\n            )\n        except Exception as e:\n            raise APIConnectionError(retryable=retryable) from e\n\n    def _parse_choice(self, id: str, choice: Choice) -> llm.ChatChunk | None:\n        delta = choice.delta\n\n        # https://github.com/livekit/agents/issues/688\n        # the delta can be None when using Azure OpenAI using content filtering\n        if delta is None:\n            return None\n\n        if delta.tool_calls:\n            # check if we have functions to calls\n            for tool in delta.tool_calls:\n                if not tool.function:\n                    continue  # oai may add other tools in the future\n\n                call_chunk = None\n                if self._tool_call_id and tool.id and tool.index != self._tool_index:\n                    call_chunk = self._try_build_function(id, choice)\n\n                if tool.function.name:\n                    self._tool_index = tool.index\n                    self._tool_call_id = tool.id\n                    self._fnc_name = tool.function.name\n                    self._fnc_raw_arguments = tool.function.arguments or \"\"\n                elif tool.function.arguments:\n                    self._fnc_raw_arguments += tool.function.arguments  # type: ignore\n\n                if call_chunk is not None:\n                    return call_chunk\n\n        if choice.finish_reason in (\"tool_calls\", \"stop\") and self._tool_call_id:\n            # we're done with the tool calls, run the last one\n            return self._try_build_function(id, choice)\n\n        return llm.ChatChunk(\n            request_id=id,\n            choices=[\n                llm.Choice(\n                    delta=llm.ChoiceDelta(content=delta.content, role=\"assistant\"),\n                    index=choice.index,\n                )\n            ],\n        )\n\n    def _try_build_function(self, id: str, choice: Choice) -> llm.ChatChunk | None:\n        if not self._fnc_ctx:\n            logger.warning(\"oai stream tried to run function without function context\")\n            return None\n\n        if self._tool_call_id is None:\n            logger.warning(\n                \"oai stream tried to run function but tool_call_id is not set\"\n            )\n            return None\n\n        if self._fnc_name is None or self._fnc_raw_arguments is None:\n            logger.warning(\n                \"oai stream tried to call a function but raw_arguments and fnc_name are not set\"\n            )\n            return None\n\n        fnc_info = _create_ai_function_info(\n            self._fnc_ctx, self._tool_call_id, self._fnc_name, self._fnc_raw_arguments\n        )\n\n        self._tool_call_id = self._fnc_name = self._fnc_raw_arguments = None\n        self._function_calls_info.append(fnc_info)\n\n        return llm.ChatChunk(\n            request_id=id,\n            choices=[\n                llm.Choice(\n                    delta=llm.ChoiceDelta(\n                        role=\"assistant\",\n                        tool_calls=[fnc_info],\n                        content=choice.delta.content,\n                    ),\n                    index=choice.index,\n                )\n            ],\n        )\n\n\ndef _build_oai_context(\n    chat_ctx: llm.ChatContext, cache_key: Any\n) -> list[ChatCompletionMessageParam]:\n    return [build_oai_message(msg, cache_key) for msg in chat_ctx.messages]  # type: ignore\n\n\ndef _strip_nones(data: dict[str, Any]) -> dict[str, Any]:\n    return {k: v for k, v in data.items() if v is not None}\n\n\ndef _get_api_key(env_var: str, key: str | None) -> str:\n    key = key or os.environ.get(env_var)\n    if not key:\n        raise ValueError(\n            f\"{env_var} is required, either as argument or set {env_var} environmental variable\"\n        )\n    return key\n"}}}}