Skip to content

Dashscope

DashScopeMultiModal #

Bases: MultiModalLLM

DashScope LLM.

Source code in llama-index-integrations/multi_modal_llms/llama-index-multi-modal-llms-dashscope/llama_index/multi_modal_llms/dashscope/base.py
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
class DashScopeMultiModal(MultiModalLLM):
    """DashScope LLM."""

    model_name: str = Field(
        default=DashScopeMultiModalModels.QWEN_VL_MAX,
        description="The DashScope model to use.",
    )
    incremental_output: Optional[bool] = Field(
        description="Control stream output, If False, the subsequent \
                                                            output will include the content that has been \
                                                            output previously.",
        default=True,
    )
    top_k: Optional[int] = Field(
        description="Sample counter when generate.", default=None
    )
    top_p: Optional[float] = Field(
        description="Sample probability threshold when generate."
    )
    seed: Optional[int] = Field(
        description="Random seed when generate.", default=1234, ge=0
    )
    api_key: Optional[str] = Field(
        default=None, description="The DashScope API key.", exclude=True
    )

    def __init__(
        self,
        model_name: Optional[str] = DashScopeMultiModalModels.QWEN_VL_MAX,
        incremental_output: Optional[int] = True,
        top_k: Optional[int] = None,
        top_p: Optional[float] = None,
        seed: Optional[int] = 1234,
        api_key: Optional[str] = None,
        callback_manager: Optional[CallbackManager] = None,
        **kwargs: Any,
    ):
        super().__init__(
            model_name=model_name,
            incremental_output=incremental_output,
            top_k=top_k,
            top_p=top_p,
            seed=seed,
            api_key=api_key,
            callback_manager=callback_manager,
            kwargs=kwargs,
        )

    @classmethod
    def class_name(cls) -> str:
        return "DashScopeMultiModal_LLM"

    @property
    def metadata(self) -> LLMMetadata:
        return LLMMetadata(
            model_name=self.model_name, **DASHSCOPE_MODEL_META[self.model_name]
        )

    def _get_default_parameters(self) -> Dict:
        params: Dict[Any, Any] = {}
        params["incremental_output"] = self.incremental_output
        if self.top_k is not None:
            params["top_k"] = self.top_k

        if self.top_p is not None:
            params["top_p"] = self.top_p
        if self.seed is not None:
            params["seed"] = self.seed

        return params

    def _get_input_parameters(
        self, prompt: str, image_documents: Sequence[ImageNode], **kwargs: Any
    ) -> Tuple[ChatMessage, Dict]:
        parameters = self._get_default_parameters()
        parameters.update(kwargs)
        parameters["stream"] = False
        if image_documents is None:
            message = ChatMessage(
                role=MessageRole.USER.value, content=[{"text": prompt}]
            )
        else:
            content = []
            for image_document in image_documents:
                content.append({"image": image_document.image_url})
            content.append({"text": prompt})
            message = ChatMessage(role=MessageRole.USER.value, content=content)
        return message, parameters

    def complete(
        self, prompt: str, image_documents: Sequence[ImageNode], **kwargs: Any
    ) -> CompletionResponse:
        message, parameters = self._get_input_parameters(
            prompt, image_documents, **kwargs
        )
        parameters.pop("incremental_output", None)
        parameters.pop("stream", None)
        messages = chat_message_to_dashscope_multi_modal_messages([message])
        response = call_with_messages(
            model=self.model_name,
            messages=messages,
            api_key=self.api_key,
            parameters=parameters,
        )
        return dashscope_response_to_completion_response(response)

    def stream_complete(
        self, prompt: str, image_documents: Sequence[ImageNode], **kwargs: Any
    ) -> CompletionResponseGen:
        message, parameters = self._get_input_parameters(
            prompt, image_documents, **kwargs
        )
        parameters["incremental_output"] = True
        parameters["stream"] = True
        responses = call_with_messages(
            model=self.model_name,
            messages=chat_message_to_dashscope_multi_modal_messages([message]),
            api_key=self.api_key,
            parameters=parameters,
        )

        def gen() -> CompletionResponseGen:
            content = ""
            for response in responses:
                if response.status_code == HTTPStatus.OK:
                    top_choice = response["output"]["choices"][0]
                    incremental_output = top_choice["message"]["content"]
                    if incremental_output:
                        incremental_output = incremental_output[0]["text"]
                    else:
                        incremental_output = ""

                    content += incremental_output
                    yield CompletionResponse(
                        text=content, delta=incremental_output, raw=response
                    )
                else:
                    yield CompletionResponse(text="", raw=response)
                    return

        return gen()

    def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
        parameters = self._get_default_parameters()
        parameters.update({**kwargs})
        parameters.pop("stream", None)
        parameters.pop("incremental_output", None)
        response = call_with_messages(
            model=self.model_name,
            messages=chat_message_to_dashscope_multi_modal_messages(messages),
            api_key=self.api_key,
            parameters=parameters,
        )
        return dashscope_response_to_chat_response(response)

    def stream_chat(
        self, messages: Sequence[ChatMessage], **kwargs: Any
    ) -> ChatResponseGen:
        parameters = self._get_default_parameters()
        parameters.update({**kwargs})
        parameters["stream"] = True
        parameters["incremental_output"] = True
        responses = call_with_messages(
            model=self.model_name,
            messages=chat_message_to_dashscope_multi_modal_messages(messages),
            api_key=self.api_key,
            parameters=parameters,
        )

        def gen() -> ChatResponseGen:
            content = ""
            for response in responses:
                if response.status_code == HTTPStatus.OK:
                    top_choice = response["output"]["choices"][0]
                    incremental_output = top_choice["message"]["content"]
                    if incremental_output:
                        incremental_output = incremental_output[0]["text"]
                    else:
                        incremental_output = ""

                    content += incremental_output
                    role = top_choice["message"]["role"]
                    yield ChatResponse(
                        message=ChatMessage(role=role, content=content),
                        delta=incremental_output,
                        raw=response,
                    )
                else:
                    yield ChatResponse(message=ChatMessage(), raw=response)
                    return

        return gen()

    # TODO: use proper async methods
    async def acomplete(
        self, prompt: str, image_documents: Sequence[ImageNode], **kwargs: Any
    ) -> CompletionResponse:
        return self.complete(prompt, image_documents, **kwargs)

    async def astream_complete(
        self, prompt: str, image_documents: Sequence[ImageNode], **kwargs: Any
    ) -> CompletionResponseAsyncGen:
        raise Exception("Not supported")

    async def achat(
        self,
        messages: Sequence[ChatMessage],
        **kwargs: Any,
    ) -> ChatResponse:
        return self.chat(messages, **kwargs)

    async def astream_chat(
        self,
        messages: Sequence[ChatMessage],
        **kwargs: Any,
    ) -> ChatResponseAsyncGen:
        raise Exception("Not supported")