Method _build_prompt_plan_locked

examples/server/server.py:10998–11226

(
        self,
        *,
        messages: List[ChatCompletionRequestMessage],
        media_inputs: List[MediaInput],
        functions: Optional[List[ChatTemplateFunctionDefinition]],
        function_call: Optional[Union[str, ChatTemplateFunctionCall]],
        tools: Optional[List[ChatTemplateTool]],
        tool_choice: Optional[Union[str, ChatTemplateToolChoice]],
        reasoning_effort: Optional[str],
    )

            )

    def _build_prompt_plan_locked(
        self,
        *,
        messages: List[ChatCompletionRequestMessage],
        media_inputs: List[MediaInput],
        functions: Optional[List[ChatTemplateFunctionDefinition]],
        function_call: Optional[Union[str, ChatTemplateFunctionCall]],
        tools: Optional[List[ChatTemplateTool]],
        tool_choice: Optional[Union[str, ChatTemplateToolChoice]],
        reasoning_effort: Optional[str],
    ) -> PromptPlan:
        prompt, generation_prompt, _ = self.chat_formatter.format(
            messages=messages,
            media_marker=self.media_marker,
            functions=functions,
            function_call=function_call,
            tools=tools,
            tool_choice=tool_choice,
            reasoning_effort=reasoning_effort,
        )
        media_bytes_by_index = [self.load_media(media) for media in media_inputs]
        loaded_media: List[MTMDLoadedMedia] = []
        chunks: Optional[Any] = None
        try:
            loaded_media = [
                self._create_loaded_media(media, media_bytes)
                for media, media_bytes in zip(media_inputs, media_bytes_by_index)
            ]
            loaded_media_by_key = {media.key: media for media in loaded_media}
            video_media = [media for media in loaded_media if media.media.kind == "video"]
            if len(video_media) > 1 and any(media.video_frame_count <= 0 for media in video_media):
                raise CompletionRequestValidationError(
                    "multiple videos require MTMD to report frame counts"
                )
            input_text = mtmd_cpp.mtmd_input_text()
            input_text.text = prompt.encode("utf-8")
            input_text.add_special = False
            input_text.parse_special = True
            chunks = mtmd_cpp.mtmd_input_chunks_init()
            if chunks is None:
                raise CompletionRequestValidationError("failed to create MTMD input chunks")
            bitmap_array = (mtmd_cpp.mtmd_bitmap_p_ctypes * len(loaded_media))(
                *(media.bitmap for media in loaded_media)
            )
            result = int(
                mtmd_cpp.mtmd_tokenize(
                    self.ctx,
                    chunks,
                    ctypes.byref(input_text),
                    bitmap_array,
                    len(loaded_media),
                )
            )
            if result != 0:
                raise CompletionRequestValidationError(
                    f"failed to tokenize MTMD prompt: error code {result}"
                )
            parsed_chunks: List[MTMDProcessor.ParsedChunk] = []