Skip to content

anthropic

AnthropicAPI

Bases: AsyncAPI

Class for querying the Anthropic API asynchronously.

Parameters:

Name Type Description Default
settings Settings

The settings for the pipeline/experiment

required
log_file str

The path to the log file

required
Source code in src/prompto/apis/anthropic/anthropic.py
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 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
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
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
class AnthropicAPI(AsyncAPI):
    """
    Class for querying the Anthropic API asynchronously.

    Parameters
    ----------
    settings : Settings
        The settings for the pipeline/experiment
    log_file : str
        The path to the log file

    """

    def __init__(
        self,
        settings: Settings,
        log_file: str,
        *args: Any,
        **kwargs: Any,
    ):
        super().__init__(settings=settings, log_file=log_file, *args, **kwargs)

    @staticmethod
    def check_environment_variables() -> list[Exception]:
        """
        For Anthropic, there are some optional environment variables:
        - ANTHROPIC_API_KEY

        These are optional only if the model_name is passed
        in the prompt dictionary. If the model_name is not
        passed, then the default values are taken from these
        environment variables.

        These are checked in the check_prompt_dict method to ensure that
        the required environment variables are set.

        Returns
        -------
        list[Exception]
            A list of exceptions or warnings if the environment variables
            are not set
        """
        issues = []

        # check the optional environment variables are set and warn if not
        issues.extend(check_optional_env_variables_set([API_KEY_VAR_NAME]))

        return issues

    @staticmethod
    def check_prompt_dict(prompt_dict: dict) -> list[Exception]:
        """
        For Anthropic, we make the following model-specific checks:
        - "prompt" key must be of type str, list[str], or list[dict[str,str]]
        - model-specific environment variable (ANTHROPIC_API_KEY_{identifier})
          (where identifier is the model name with invalid characters replaced
          by underscores obtained using get_model_name_identifier function)
          can be set or the default environment variable must be set

        Parameters
        ----------
        prompt_dict : dict
            The prompt dictionary to check

        Returns
        -------
        list[Exception]
            A list of exceptions or warnings if the prompt dictionary
            is not valid
        """
        issues = []

        # check prompt is of the right type
        if isinstance(prompt_dict["prompt"], str):
            pass
        elif isinstance(prompt_dict["prompt"], list):
            if all([isinstance(message, str) for message in prompt_dict["prompt"]]):
                pass
            elif (
                all(isinstance(message, dict) for message in prompt_dict["prompt"])
                and (
                    set(prompt_dict["prompt"][0].keys()) == {"role", "content"}
                    and prompt_dict["prompt"][0]["role"]
                    in list(anthropic_chat_roles) + ["system"]
                )
                and all(
                    [
                        set(d.keys()) == {"role", "content"}
                        and d["role"] in anthropic_chat_roles
                        for d in prompt_dict["prompt"][1:]
                    ]
                )
            ):
                pass
            else:
                issues.append(TYPE_ERROR)
        else:
            issues.append(TYPE_ERROR)

        # use the model specific environment variables if they exist
        model_name = prompt_dict["model_name"]
        # replace any invalid characters in the model name
        identifier = get_model_name_identifier(model_name)

        # check the required environment variables are set
        # must either have the model specific key or the default key set
        issues.extend(
            check_either_required_env_variables_set(
                [
                    [f"{API_KEY_VAR_NAME}_{identifier}", API_KEY_VAR_NAME],
                ]
            )
        )

        return issues

    async def _obtain_model_inputs(
        self, prompt_dict: dict
    ) -> tuple[str, str, AsyncAnthropic, dict]:
        """
        Async method to obtain the model inputs from the prompt dictionary.

        Parameters
        ----------
        prompt_dict : dict
            The prompt dictionary to use for querying the model

        Returns
        -------
        tuple[str, str, AsyncAnthropic, dict]
            A tuple containing the prompt, model name, AsyncAnthropic client object,
            the generation config, and mode to use for querying the model
        """
        # obtain the prompt from the prompt dictionary
        prompt = prompt_dict["prompt"]

        # obtain model name
        model_name = prompt_dict["model_name"]
        api_key = get_environment_variable(
            env_variable=API_KEY_VAR_NAME, model_name=model_name
        )

        # create the AsyncAnthropic client object
        client = AsyncAnthropic(api_key=api_key, max_retries=1)

        # get parameters dict (if any)
        generation_config = prompt_dict.get("parameters", None)
        if generation_config is None:
            generation_config = {}
        if type(generation_config) is not dict:
            raise TypeError(
                f"parameters must be a dictionary, not {type(generation_config)}"
            )

        return prompt, model_name, client, generation_config

    async def _query_string(self, prompt_dict: dict, index: int | str) -> dict:
        """
        Async method for querying the model with a string prompt
        (prompt_dict["prompt"] is a string),
        i.e. single-turn completion or chat.
        """
        prompt, model_name, client, generation_config = await self._obtain_model_inputs(
            prompt_dict
        )

        try:
            response = await client.messages.create(
                model=model_name,
                messages=[{"role": "user", "content": prompt}],
                **generation_config,
            )

            response_text = process_response(response)

            log_success_response_query(
                index=index,
                model=f"Anthropic ({model_name})",
                prompt=prompt,
                response_text=response_text,
                id=prompt_dict.get("id", "NA"),
            )

            prompt_dict["response"] = response_text
            return prompt_dict
        except Exception as err:
            error_as_string = f"{type(err).__name__} - {err}"
            log_message = log_error_response_query(
                index=index,
                model=f"Anthropic ({model_name})",
                prompt=prompt,
                error_as_string=error_as_string,
                id=prompt_dict.get("id", "NA"),
            )
            async with FILE_WRITE_LOCK:
                write_log_message(
                    log_file=self.log_file,
                    log_message=log_message,
                    log=True,
                )
            raise err

    async def _query_chat(self, prompt_dict: dict, index: int | str) -> dict:
        """
        Async method for querying the model with a chat prompt
        (prompt_dict["prompt"] is a list of strings to sequentially send to the model),
        i.e. multi-turn chat with history.
        """
        prompt, model_name, client, generation_config = await self._obtain_model_inputs(
            prompt_dict
        )

        messages = []
        response_list = []

        try:
            for message_index, message in enumerate(prompt):
                # add the user message to the list of messages
                messages.append({"role": "user", "content": message})
                # obtain the response from the model
                response = await client.messages.create(
                    model=model_name,
                    messages=messages,
                    **generation_config,
                )
                # parse the response to obtain the response text
                response_text = process_response(response)
                # add the response to the list of responses
                response_list.append(response_text)
                # add the response message to the list of messages
                messages.append({"role": "assistant", "content": response_text})

                log_success_response_chat(
                    index=index,
                    model=f"Anthropic ({model_name})",
                    message_index=message_index,
                    n_messages=len(prompt),
                    message=message,
                    response_text=response_text,
                    id=prompt_dict.get("id", "NA"),
                )

            logging.info(
                f"Chat completed (i={index}, id={prompt_dict.get('id', 'NA')})"
            )

            prompt_dict["response"] = response_list
            return prompt_dict
        except Exception as err:
            error_as_string = f"{type(err).__name__} - {err}"
            log_message = log_error_response_chat(
                index=index,
                model=f"Anthropic ({model_name})",
                message_index=message_index,
                n_messages=len(prompt),
                message=message,
                responses_so_far=response_list,
                error_as_string=error_as_string,
                id=prompt_dict.get("id", "NA"),
            )
            async with FILE_WRITE_LOCK:
                write_log_message(
                    log_file=self.log_file,
                    log_message=log_message,
                    log=True,
                )
            raise err

    async def _query_history(self, prompt_dict: dict, index: int | str) -> dict:
        """
        Async method for querying the model with a chat prompt with history
        (prompt_dict["prompt"] is a list of dictionaries with keys "role" and "content",
        where "role" is one of "user", "assistant", or "system" and "content" is the message),
        i.e. multi-turn chat with history.

        The "system" role is not handled the same way as in the OpenAI API.
        There is no "system role". Instead, it is handled in a separate parameter
        outside of the dictionary. This argument accepts the system role in the prompt_dict,
        but extracts it from the dictionary and passes it as a separate argument.
        """
        prompt, model_name, client, generation_config = await self._obtain_model_inputs(
            prompt_dict
        )

        # pop the "system" role from the prompt
        system = [
            message_dict["content"]
            for message_dict in prompt
            if message_dict["role"] == "system"
        ]

        # remove the system messages from prompt
        messages = [
            message_dict for message_dict in prompt if message_dict["role"] != "system"
        ]

        # if system message is present, then it must be the only one
        if len(system) == 0:
            system = ""
        elif len(system) == 1:
            system = system[0]
        else:
            raise ValueError(
                f"There are {len(system)} system messages. Only one system message is supported"
            )

        try:
            response = await client.messages.create(
                model=model_name,
                messages=[
                    convert_dict_to_input(
                        content_dict=x, media_folder=self.settings.media_folder
                    )
                    for x in messages
                ],
                system=system,
                **generation_config,
            )

            response_text = process_response(response)

            log_success_response_query(
                index=index,
                model=f"Anthropic ({model_name})",
                prompt=prompt,
                response_text=response_text,
                id=prompt_dict.get("id", "NA"),
            )

            prompt_dict["response"] = response_text
            return prompt_dict
        except Exception as err:
            error_as_string = f"{type(err).__name__} - {err}"
            log_message = log_error_response_query(
                index=index,
                model=f"Anthropic ({model_name})",
                prompt=prompt,
                error_as_string=error_as_string,
                id=prompt_dict.get("id", "NA"),
            )
            async with FILE_WRITE_LOCK:
                write_log_message(
                    log_file=self.log_file,
                    log_message=log_message,
                    log=True,
                )
            raise err

    async def query(self, prompt_dict: dict, index: int | str = "NA") -> dict:
        """
        Async Method for querying the API/model asynchronously.

        Parameters
        ----------
        prompt_dict : dict
            The prompt dictionary to use for querying the model
        index : int | str
            The index of the prompt in the experiment

        Returns
        -------
        dict
            Completed prompt_dict with "response" key storing the response(s)
            from the LLM

        Raises
        ------
        Exception
            If an error occurs during the querying process
        """
        if isinstance(prompt_dict["prompt"], str):
            return await self._query_string(
                prompt_dict=prompt_dict,
                index=index,
            )
        elif isinstance(prompt_dict["prompt"], list):
            if all([isinstance(message, str) for message in prompt_dict["prompt"]]):
                return await self._query_chat(
                    prompt_dict=prompt_dict,
                    index=index,
                )
            elif (
                all(isinstance(message, dict) for message in prompt_dict["prompt"])
                and (
                    set(prompt_dict["prompt"][0].keys()) == {"role", "content"}
                    and prompt_dict["prompt"][0]["role"]
                    in list(anthropic_chat_roles) + ["system"]
                )
                and all(
                    [
                        set(d.keys()) == {"role", "content"}
                        and d["role"] in anthropic_chat_roles
                        for d in prompt_dict["prompt"][1:]
                    ]
                )
            ):
                return await self._query_history(
                    prompt_dict=prompt_dict,
                    index=index,
                )

        raise TYPE_ERROR

check_environment_variables staticmethod

check_environment_variables() -> list[Exception]

For Anthropic, there are some optional environment variables: - ANTHROPIC_API_KEY

These are optional only if the model_name is passed in the prompt dictionary. If the model_name is not passed, then the default values are taken from these environment variables.

These are checked in the check_prompt_dict method to ensure that the required environment variables are set.

Returns:

Type Description
list[Exception]

A list of exceptions or warnings if the environment variables are not set

Source code in src/prompto/apis/anthropic/anthropic.py
@staticmethod
def check_environment_variables() -> list[Exception]:
    """
    For Anthropic, there are some optional environment variables:
    - ANTHROPIC_API_KEY

    These are optional only if the model_name is passed
    in the prompt dictionary. If the model_name is not
    passed, then the default values are taken from these
    environment variables.

    These are checked in the check_prompt_dict method to ensure that
    the required environment variables are set.

    Returns
    -------
    list[Exception]
        A list of exceptions or warnings if the environment variables
        are not set
    """
    issues = []

    # check the optional environment variables are set and warn if not
    issues.extend(check_optional_env_variables_set([API_KEY_VAR_NAME]))

    return issues

check_prompt_dict staticmethod

check_prompt_dict(prompt_dict: dict) -> list[Exception]

For Anthropic, we make the following model-specific checks: - “prompt” key must be of type str, list[str], or list[dict[str,str]] - model-specific environment variable (ANTHROPIC_API_KEY_{identifier}) (where identifier is the model name with invalid characters replaced by underscores obtained using get_model_name_identifier function) can be set or the default environment variable must be set

Parameters:

Name Type Description Default
prompt_dict dict

The prompt dictionary to check

required

Returns:

Type Description
list[Exception]

A list of exceptions or warnings if the prompt dictionary is not valid

Source code in src/prompto/apis/anthropic/anthropic.py
@staticmethod
def check_prompt_dict(prompt_dict: dict) -> list[Exception]:
    """
    For Anthropic, we make the following model-specific checks:
    - "prompt" key must be of type str, list[str], or list[dict[str,str]]
    - model-specific environment variable (ANTHROPIC_API_KEY_{identifier})
      (where identifier is the model name with invalid characters replaced
      by underscores obtained using get_model_name_identifier function)
      can be set or the default environment variable must be set

    Parameters
    ----------
    prompt_dict : dict
        The prompt dictionary to check

    Returns
    -------
    list[Exception]
        A list of exceptions or warnings if the prompt dictionary
        is not valid
    """
    issues = []

    # check prompt is of the right type
    if isinstance(prompt_dict["prompt"], str):
        pass
    elif isinstance(prompt_dict["prompt"], list):
        if all([isinstance(message, str) for message in prompt_dict["prompt"]]):
            pass
        elif (
            all(isinstance(message, dict) for message in prompt_dict["prompt"])
            and (
                set(prompt_dict["prompt"][0].keys()) == {"role", "content"}
                and prompt_dict["prompt"][0]["role"]
                in list(anthropic_chat_roles) + ["system"]
            )
            and all(
                [
                    set(d.keys()) == {"role", "content"}
                    and d["role"] in anthropic_chat_roles
                    for d in prompt_dict["prompt"][1:]
                ]
            )
        ):
            pass
        else:
            issues.append(TYPE_ERROR)
    else:
        issues.append(TYPE_ERROR)

    # use the model specific environment variables if they exist
    model_name = prompt_dict["model_name"]
    # replace any invalid characters in the model name
    identifier = get_model_name_identifier(model_name)

    # check the required environment variables are set
    # must either have the model specific key or the default key set
    issues.extend(
        check_either_required_env_variables_set(
            [
                [f"{API_KEY_VAR_NAME}_{identifier}", API_KEY_VAR_NAME],
            ]
        )
    )

    return issues

query async

query(prompt_dict: dict, index: int | str = 'NA') -> dict

Async Method for querying the API/model asynchronously.

Parameters:

Name Type Description Default
prompt_dict dict

The prompt dictionary to use for querying the model

required
index int | str

The index of the prompt in the experiment

'NA'

Returns:

Type Description
dict

Completed prompt_dict with “response” key storing the response(s) from the LLM

Raises:

Type Description
Exception

If an error occurs during the querying process

Source code in src/prompto/apis/anthropic/anthropic.py
async def query(self, prompt_dict: dict, index: int | str = "NA") -> dict:
    """
    Async Method for querying the API/model asynchronously.

    Parameters
    ----------
    prompt_dict : dict
        The prompt dictionary to use for querying the model
    index : int | str
        The index of the prompt in the experiment

    Returns
    -------
    dict
        Completed prompt_dict with "response" key storing the response(s)
        from the LLM

    Raises
    ------
    Exception
        If an error occurs during the querying process
    """
    if isinstance(prompt_dict["prompt"], str):
        return await self._query_string(
            prompt_dict=prompt_dict,
            index=index,
        )
    elif isinstance(prompt_dict["prompt"], list):
        if all([isinstance(message, str) for message in prompt_dict["prompt"]]):
            return await self._query_chat(
                prompt_dict=prompt_dict,
                index=index,
            )
        elif (
            all(isinstance(message, dict) for message in prompt_dict["prompt"])
            and (
                set(prompt_dict["prompt"][0].keys()) == {"role", "content"}
                and prompt_dict["prompt"][0]["role"]
                in list(anthropic_chat_roles) + ["system"]
            )
            and all(
                [
                    set(d.keys()) == {"role", "content"}
                    and d["role"] in anthropic_chat_roles
                    for d in prompt_dict["prompt"][1:]
                ]
            )
        ):
            return await self._query_history(
                prompt_dict=prompt_dict,
                index=index,
            )

    raise TYPE_ERROR