llama-stack/llama_stack/providers/remote/inference/databricks/databricks.py

# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.

from typing import AsyncGenerator, List, Optional

from openai import OpenAI

from llama_stack.apis.common.content_types import (
    InterleavedContent,
    InterleavedContentItem,
)
from llama_stack.apis.inference import (
    ChatCompletionRequest,
    ChatCompletionResponse,
    EmbeddingsResponse,
    EmbeddingTaskType,
    Inference,
    LogProbConfig,
    Message,
    ResponseFormat,
    SamplingParams,
    TextTruncation,
    ToolChoice,
    ToolConfig,
    ToolDefinition,
    ToolPromptFormat,
)
from llama_stack.models.llama.sku_types import CoreModelId
from llama_stack.providers.utils.inference.model_registry import (
    ModelRegistryHelper,
    build_hf_repo_model_entry,
)
from llama_stack.providers.utils.inference.openai_compat import (
    OpenAIChatCompletionUnsupportedMixin,
    OpenAICompletionUnsupportedMixin,
    get_sampling_options,
    process_chat_completion_response,
    process_chat_completion_stream_response,
)
from llama_stack.providers.utils.inference.prompt_adapter import (
    chat_completion_request_to_prompt,
)

from .config import DatabricksImplConfig

model_entries = [
    build_hf_repo_model_entry(
        "databricks-meta-llama-3-1-70b-instruct",
        CoreModelId.llama3_1_70b_instruct.value,
    ),
    build_hf_repo_model_entry(
        "databricks-meta-llama-3-1-405b-instruct",
        CoreModelId.llama3_1_405b_instruct.value,
    ),
]


class DatabricksInferenceAdapter(
    ModelRegistryHelper,
    Inference,
    OpenAIChatCompletionUnsupportedMixin,
    OpenAICompletionUnsupportedMixin,
):
    def __init__(self, config: DatabricksImplConfig) -> None:
        ModelRegistryHelper.__init__(self, model_entries=model_entries)
        self.config = config

    async def initialize(self) -> None:
        return

    async def shutdown(self) -> None:
        pass

    async def completion(
        self,
        model: str,
        content: InterleavedContent,
        sampling_params: Optional[SamplingParams] = None,
        response_format: Optional[ResponseFormat] = None,
        stream: Optional[bool] = False,
        logprobs: Optional[LogProbConfig] = None,
    ) -> AsyncGenerator:
        raise NotImplementedError()

    async def chat_completion(
        self,
        model: str,
        messages: List[Message],
        sampling_params: Optional[SamplingParams] = None,
        response_format: Optional[ResponseFormat] = None,
        tools: Optional[List[ToolDefinition]] = None,
        tool_choice: Optional[ToolChoice] = ToolChoice.auto,
        tool_prompt_format: Optional[ToolPromptFormat] = None,
        stream: Optional[bool] = False,
        logprobs: Optional[LogProbConfig] = None,
        tool_config: Optional[ToolConfig] = None,
    ) -> AsyncGenerator:
        if sampling_params is None:
            sampling_params = SamplingParams()
        request = ChatCompletionRequest(
            model=model,
            messages=messages,
            sampling_params=sampling_params,
            tools=tools or [],
            stream=stream,
            logprobs=logprobs,
            tool_config=tool_config,
        )

        client = OpenAI(base_url=self.config.url, api_key=self.config.api_token)
        if stream:
            return self._stream_chat_completion(request, client)
        else:
            return await self._nonstream_chat_completion(request, client)

    async def _nonstream_chat_completion(
        self, request: ChatCompletionRequest, client: OpenAI
    ) -> ChatCompletionResponse:
        params = self._get_params(request)
        r = client.completions.create(**params)
        return process_chat_completion_response(r, request)

    async def _stream_chat_completion(self, request: ChatCompletionRequest, client: OpenAI) -> AsyncGenerator:
        params = self._get_params(request)

        async def _to_async_generator():
            s = client.completions.create(**params)
            for chunk in s:
                yield chunk

        stream = _to_async_generator()
        async for chunk in process_chat_completion_stream_response(stream, request):
            yield chunk

    def _get_params(self, request: ChatCompletionRequest) -> dict:
        return {
            "model": request.model,
            "prompt": chat_completion_request_to_prompt(request, self.get_llama_model(request.model)),
            "stream": request.stream,
            **get_sampling_options(request.sampling_params),
        }

    async def embeddings(
        self,
        model_id: str,
        contents: List[str] | List[InterleavedContentItem],
        text_truncation: Optional[TextTruncation] = TextTruncation.none,
        output_dimension: Optional[int] = None,
        task_type: Optional[EmbeddingTaskType] = None,
    ) -> EmbeddingsResponse:
        raise NotImplementedError()