Skip to content

Service

dandy.llm.service.service.LlmService

Bases: BaseHttpService

Source code in dandy/llm/service/service.py 
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
def __init__(
        self,
        llm_config: BaseLlmConfig,
        llm_options: LlmConfigOptions,
):
    super().__init__(config=llm_config.http_config)

    self.event_id = generate_new_recorder_event_id()

    self._llm_config = llm_config
    self._llm_options = llm_options
    self._intel = None
    self._intel_json_schema = None
    self._request_body = self._llm_config.generate_request_body(
        max_input_tokens=self._llm_options.max_input_tokens,
        max_output_tokens=self._llm_options.max_output_tokens,
        seed=self._llm_options.seed,
        temperature=self._llm_options.temperature,
    )
    self._response_content = None
    self._retry_attempt = 0

event_id = generate_new_recorder_event_id() instance-attribute

has_retry_attempts_available property

process_prompt_to_intel

Source code in dandy/llm/service/service.py 
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 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
def process_prompt_to_intel(
        self,
        prompt: PromptOrStr,
        intel_class: Union[Type[IntelType], None] = None,
        intel_object: Union[IntelType, None] = None,
        images: Union[List[str], None] = None,
        include_fields: Union[IncEx, None] = None,
        exclude_fields: Union[IncEx, None] = None,
        system_prompt: PromptOrStrOrNone = None,
        message_history: Union[MessageHistory, None] = None,
) -> IntelType:

    if intel_class and intel_object:
        raise LlmCriticalException('Cannot specify both intel_class and intel_object.')

    if intel_class is None and intel_object is None:
        raise LlmCriticalException('Must specify either intel_class or intel_object.')

    self._intel = intel_class or intel_object

    self._intel_json_schema = IntelFactory.intel_to_json_inc_ex_schema(
        intel=self._intel,
        include=include_fields,
        exclude=exclude_fields
    )

    self._request_body.set_format_to_json_schema(
        self._intel_json_schema
    )

    self._request_body.add_message(
        role='system',
        content=service_system_prompt(
            system_prompt=system_prompt
        ).to_str()
    )

    if message_history:
        for message in message_history.messages:
            self._request_body.add_message(
                role=message.role,
                content=message.content,
                images=message.images
            )

    self._request_body.add_message(
        role='user',
        content=service_user_prompt(
            prompt if isinstance(prompt, Prompt) else Prompt(prompt)
        ).to_str(),
        images=images,
    )

    return self._process_request_to_intel()

retry_process_request_to_intel

Source code in dandy/llm/service/service.py 
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
def retry_process_request_to_intel(
        self,
        retry_event_description: str,
        retry_user_prompt: PromptOrStr,
) -> IntelType:
    if self.has_retry_attempts_available:
        self._retry_attempt += 1

        recorder_add_llm_retry_event(
            retry_event_description,
            self.event_id,
            remaining_attempts=self._llm_config.options.prompt_retry_count - self._retry_attempt
        )

        self._request_body.add_message(
            role='assistant',
            content=self._response_content
        )

        self._request_body.add_message(
            role='user',
            content=Prompt(retry_user_prompt).to_str()
        )

        return self._process_request_to_intel(
        )

    else:
        raise LlmRecoverableException(
            f'Failed to get the correct response from the LlmService after {self._llm_config.options.prompt_retry_count} attempts.')