edsl 0.1.30.dev5__py3-none-any.whl → 0.1.31__py3-none-any.whl

edsl/__version__.py

@@ -1 +1 @@
-__version__ = "0.1.30.dev5"
+__version__ = "0.1.31"

edsl/agents/Invigilator.py

@@ -18,7 +18,12 @@ class InvigilatorAI(PromptConstructorMixin, InvigilatorBase):
     """An invigilator that uses an AI model to answer questions."""
 
     async def async_answer_question(self) -> AgentResponseDict:
-        """Answer a question using the AI model."""
+        """Answer a question using the AI model.
+
+        >>> i = InvigilatorAI.example()
+        >>> i.answer_question()
+        {'message': '{"answer": "SPAM!"}'}
+        """
         params = self.get_prompts() | {"iteration": self.iteration}
         raw_response = await self.async_get_response(**params)
         data = {
@@ -29,6 +34,7 @@ class InvigilatorAI(PromptConstructorMixin, InvigilatorBase):
             "raw_model_response": raw_response["raw_model_response"],
         }
         response = self._format_raw_response(**data)
+        # breakpoint()
         return AgentResponseDict(**response)
 
     async def async_get_response(
@@ -97,7 +103,6 @@ class InvigilatorAI(PromptConstructorMixin, InvigilatorBase):
         answer = question._translate_answer_code_to_answer(
             response["answer"], combined_dict
         )
-        # breakpoint()
         data = {
             "answer": answer,
             "comment": response.get(

edsl/agents/PromptConstructionMixin.py

@@ -275,8 +275,25 @@ class PromptConstructorMixin:
                     if (new_question := question.split("_comment")[0]) in d:
                         d[new_question].comment = answer
 
+            question_data = self.question.data.copy()
+
+            # check to see if the questio_options is actuall a string
+            if "question_options" in question_data:
+                if isinstance(self.question.data["question_options"], str):
+                    from jinja2 import Environment, meta
+
+                    env = Environment()
+                    parsed_content = env.parse(self.question.data["question_options"])
+                    question_option_key = list(
+                        meta.find_undeclared_variables(parsed_content)
+                    )[0]
+                    question_data["question_options"] = self.scenario.get(
+                        question_option_key
+                    )
+
+            # breakpoint()
             rendered_instructions = question_prompt.render(
-                self.question.data | self.scenario | d | {"agent": self.agent}
+                question_data | self.scenario | d | {"agent": self.agent}
             )
 
             undefined_template_variables = (

edsl/config.py

@@ -65,6 +65,10 @@ CONFIG_MAP = {
     #     "default": None,
     #     "info": "This env var holds your Anthropic API key (https://www.anthropic.com/).",
     # },
+    # "GROQ_API_KEY": {
+    #     "default": None,
+    #     "info": "This env var holds your GROQ API key (https://console.groq.com/login).",
+    # },
 }
 
 

edsl/conjure/Conjure.py

@@ -35,6 +35,12 @@ class Conjure:
         # The __init__ method in Conjure won't be called because __new__ returns a different class instance.
         pass
 
+    @classmethod
+    def example(cls):
+        from edsl.conjure.InputData import InputDataABC
+
+        return InputDataABC.example()
+
 
 if __name__ == "__main__":
     pass

edsl/coop/coop.py

@@ -465,6 +465,7 @@ class Coop:
         description: Optional[str] = None,
         status: RemoteJobStatus = "queued",
         visibility: Optional[VisibilityType] = "unlisted",
+        iterations: Optional[int] = 1,
     ) -> dict:
         """
         Send a remote inference job to the server.
@@ -473,6 +474,7 @@ class Coop:
         :param optional description: A description for this entry in the remote cache.
         :param status: The status of the job. Should be 'queued', unless you are debugging.
         :param visibility: The visibility of the cache entry.
+        :param iterations: The number of times to run each interview.
 
         >>> job = Jobs.example()
         >>> coop.remote_inference_create(job=job, description="My job")
@@ -488,6 +490,7 @@ class Coop:
                 ),
                 "description": description,
                 "status": status,
+                "iterations": iterations,
                 "visibility": visibility,
                 "version": self._edsl_version,
             },
@@ -498,6 +501,7 @@ class Coop:
             "uuid": response_json.get("jobs_uuid"),
             "description": response_json.get("description"),
             "status": response_json.get("status"),
+            "iterations": response_json.get("iterations"),
             "visibility": response_json.get("visibility"),
             "version": self._edsl_version,
         }

edsl/coop/utils.py

@@ -2,6 +2,7 @@ from edsl import (
     Agent,
     AgentList,
     Cache,
+    ModelList,
     Notebook,
     Results,
     Scenario,
@@ -9,6 +10,7 @@ from edsl import (
     Survey,
     Study,
 )
+from edsl.language_models import LanguageModel
 from edsl.questions import QuestionBase
 from typing import Literal, Optional, Type, Union
 
@@ -16,6 +18,8 @@ EDSLObject = Union[
     Agent,
     AgentList,
     Cache,
+    LanguageModel,
+    ModelList,
     Notebook,
     Type[QuestionBase],
     Results,
@@ -29,6 +33,8 @@ ObjectType = Literal[
     "agent",
     "agent_list",
     "cache",
+    "model",
+    "model_list",
     "notebook",
     "question",
     "results",
@@ -62,8 +68,10 @@ class ObjectRegistry:
         {"object_type": "agent", "edsl_class": Agent},
         {"object_type": "agent_list", "edsl_class": AgentList},
         {"object_type": "cache", "edsl_class": Cache},
-        {"object_type": "question", "edsl_class": QuestionBase},
+        {"object_type": "model", "edsl_class": LanguageModel},
+        {"object_type": "model_list", "edsl_class": ModelList},
         {"object_type": "notebook", "edsl_class": Notebook},
+        {"object_type": "question", "edsl_class": QuestionBase},
         {"object_type": "results", "edsl_class": Results},
         {"object_type": "scenario", "edsl_class": Scenario},
         {"object_type": "scenario_list", "edsl_class": ScenarioList},

edsl/data/CacheHandler.py

@@ -41,7 +41,7 @@ class CacheHandler:
         old_data = self.from_old_sqlite_cache()
         self.cache.add_from_dict(old_data)
 
-    def create_cache_directory(self) -> None:
+    def create_cache_directory(self, notify = False) -> None:
         """
         Create the cache directory if one is required and it does not exist.
         """
@@ -49,9 +49,8 @@ class CacheHandler:
         dir_path = os.path.dirname(path)
         if dir_path and not os.path.exists(dir_path):
             os.makedirs(dir_path)
-            import warnings
-
-            warnings.warn(f"Created cache directory: {dir_path}")
+            if notify:
+                print(f"Created cache directory: {dir_path}")
 
     def gen_cache(self) -> Cache:
         """

edsl/enums.py

@@ -59,6 +59,7 @@ class InferenceServiceType(EnumWithChecks):
     GOOGLE = "google"
     TEST = "test"
     ANTHROPIC = "anthropic"
+    GROQ = "groq"
 
 
 service_to_api_keyname = {
@@ -69,6 +70,7 @@ service_to_api_keyname = {
     InferenceServiceType.GOOGLE.value: "GOOGLE_API_KEY",
     InferenceServiceType.TEST.value: "TBD",
     InferenceServiceType.ANTHROPIC.value: "ANTHROPIC_API_KEY",
+    InferenceServiceType.GROQ.value: "GROQ_API_KEY",
 }
 
 

edsl/inference_services/DeepInfraService.py

@@ -2,102 +2,17 @@ import aiohttp
 import json
 import requests
 from typing import Any, List
-from edsl.inference_services.InferenceServiceABC import InferenceServiceABC
+
+# from edsl.inference_services.InferenceServiceABC import InferenceServiceABC
 from edsl.language_models import LanguageModel
 
+from edsl.inference_services.OpenAIService import OpenAIService
+
 
-class DeepInfraService(InferenceServiceABC):
+class DeepInfraService(OpenAIService):
     """DeepInfra service class."""
 
     _inference_service_ = "deep_infra"
     _env_key_name_ = "DEEP_INFRA_API_KEY"
-
+    _base_url_ = "https://api.deepinfra.com/v1/openai"
     _models_list_cache: List[str] = []
-
-    @classmethod
-    def available(cls):
-        text_models = cls.full_details_available()
-        return [m["model_name"] for m in text_models]
-
-    @classmethod
-    def full_details_available(cls, verbose=False):
-        if not cls._models_list_cache:
-            url = "https://api.deepinfra.com/models/list"
-            response = requests.get(url)
-            if response.status_code == 200:
-                text_generation_models = [
-                    r for r in response.json() if r["type"] == "text-generation"
-                ]
-                cls._models_list_cache = text_generation_models
-
-                from rich import print_json
-                import json
-
-                if verbose:
-                    print_json(json.dumps(text_generation_models))
-                return text_generation_models
-            else:
-                return f"Failed to fetch data: Status code {response.status_code}"
-        else:
-            return cls._models_list_cache
-
-    @classmethod
-    def create_model(cls, model_name: str, model_class_name=None) -> LanguageModel:
-        base_url = "https://api.deepinfra.com/v1/inference/"
-        if model_class_name is None:
-            model_class_name = cls.to_class_name(model_name)
-        url = f"{base_url}{model_name}"
-
-        class LLM(LanguageModel):
-            _inference_service_ = cls._inference_service_
-            _model_ = model_name
-            _parameters_ = {
-                "temperature": 0.7,
-                "top_p": 0.2,
-                "top_k": 0.1,
-                "max_new_tokens": 512,
-                "stopSequences": [],
-            }
-
-            async def async_execute_model_call(
-                self, user_prompt: str, system_prompt: str = ""
-            ) -> dict[str, Any]:
-                self.url = url
-                headers = {
-                    "Content-Type": "application/json",
-                    "Authorization": f"bearer {self.api_token}",
-                }
-                # don't mess w/ the newlines
-                data = {
-                    "input": f"""
-                    [INST]<<SYS>>
-                    {system_prompt}
-                    <<SYS>>{user_prompt}[/INST]
-                    """,
-                    "stream": False,
-                    "temperature": self.temperature,
-                    "top_p": self.top_p,
-                    "top_k": self.top_k,
-                    "max_new_tokens": self.max_new_tokens,
-                }
-                async with aiohttp.ClientSession() as session:
-                    async with session.post(
-                        self.url, headers=headers, data=json.dumps(data)
-                    ) as response:
-                        raw_response_text = await response.text()
-                        return json.loads(raw_response_text)
-
-            def parse_response(self, raw_response: dict[str, Any]) -> str:
-                if "results" not in raw_response:
-                    raise Exception(
-                        f"Deep Infra response does not contain 'results' key: {raw_response}"
-                    )
-                if "generated_text" not in raw_response["results"][0]:
-                    raise Exception(
-                        f"Deep Infra response does not contain 'generate_text' key: {raw_response['results'][0]}"
-                    )
-                return raw_response["results"][0]["generated_text"]
-
-        LLM.__name__ = model_class_name
-
-        return LLM

edsl/inference_services/GroqService.py

@@ -0,0 +1,18 @@
+from typing import Any, List
+from edsl.inference_services.OpenAIService import OpenAIService
+
+import groq
+
+
+class GroqService(OpenAIService):
+    """DeepInfra service class."""
+
+    _inference_service_ = "groq"
+    _env_key_name_ = "GROQ_API_KEY"
+
+    _sync_client_ = groq.Groq
+    _async_client_ = groq.AsyncGroq
+
+    # _base_url_ = "https://api.deepinfra.com/v1/openai"
+    _base_url_ = None
+    _models_list_cache: List[str] = []

edsl/inference_services/InferenceServicesCollection.py

@@ -15,15 +15,19 @@ class InferenceServicesCollection:
         cls.added_models[service_name].append(model_name)
 
     @staticmethod
-    def _get_service_available(service) -> list[str]:
+    def _get_service_available(service, warn: bool = False) -> list[str]:
         from_api = True
         try:
             service_models = service.available()
         except Exception as e:
-            warnings.warn(
-                f"Error getting models for {service._inference_service_}. Relying on cache.",
-                UserWarning,
-            )
+            if warn:
+                warnings.warn(
+                    f"""Error getting models for {service._inference_service_}. 
+                    Check that you have properly stored your Expected Parrot API key and activated remote inference, or stored your own API keys for the language models that you want to use.
+                    See https://docs.expectedparrot.com/en/latest/api_keys.html for instructions on storing API keys.
+                    Relying on cache.""",
+                    UserWarning,
+                )
             from edsl.inference_services.models_available_cache import models_available
 
             service_models = models_available.get(service._inference_service_, [])
@@ -57,4 +61,8 @@ class InferenceServicesCollection:
                 if service_name is None or service_name == service._inference_service_:
                     return service.create_model(model_name)
 
+        # if model_name == "test":
+        #     from edsl.language_models import LanguageModel
+        #     return LanguageModel(test = True)
+
         raise Exception(f"Model {model_name} not found in any of the services")

edsl/inference_services/OpenAIService.py

@@ -1,6 +1,9 @@
 from typing import Any, List
 import re
-from openai import AsyncOpenAI
+import os
+
+# from openai import AsyncOpenAI
+import openai
 
 from edsl.inference_services.InferenceServiceABC import InferenceServiceABC
 from edsl.language_models import LanguageModel
@@ -12,6 +15,22 @@ class OpenAIService(InferenceServiceABC):
 
     _inference_service_ = "openai"
     _env_key_name_ = "OPENAI_API_KEY"
+    _base_url_ = None
+
+    _sync_client_ = openai.OpenAI
+    _async_client_ = openai.AsyncOpenAI
+
+    @classmethod
+    def sync_client(cls):
+        return cls._sync_client_(
+            api_key=os.getenv(cls._env_key_name_), base_url=cls._base_url_
+        )
+
+    @classmethod
+    def async_client(cls):
+        return cls._async_client_(
+            api_key=os.getenv(cls._env_key_name_), base_url=cls._base_url_
+        )
 
     # TODO: Make this a coop call
     model_exclude_list = [
@@ -31,16 +50,24 @@ class OpenAIService(InferenceServiceABC):
     ]
     _models_list_cache: List[str] = []
 
+    @classmethod
+    def get_model_list(cls):
+        raw_list = cls.sync_client().models.list()
+        if hasattr(raw_list, "data"):
+            return raw_list.data
+        else:
+            return raw_list
+
     @classmethod
     def available(cls) -> List[str]:
-        from openai import OpenAI
+        # from openai import OpenAI
 
         if not cls._models_list_cache:
             try:
-                client = OpenAI()
+                # client = OpenAI(api_key = os.getenv(cls._env_key_name_), base_url = cls._base_url_)
                 cls._models_list_cache = [
                     m.id
-                    for m in client.models.list()
+                    for m in cls.get_model_list()
                     if m.id not in cls.model_exclude_list
                 ]
             except Exception as e:
@@ -78,15 +105,24 @@ class OpenAIService(InferenceServiceABC):
                 "top_logprobs": 3,
             }
 
+            def sync_client(self):
+                return cls.sync_client()
+
+            def async_client(self):
+                return cls.async_client()
+
             @classmethod
             def available(cls) -> list[str]:
-                client = openai.OpenAI()
-                return client.models.list()
+                # import openai
+                # client = openai.OpenAI(api_key = os.getenv(cls._env_key_name_), base_url = cls._base_url_)
+                # return client.models.list()
+                return cls.sync_client().models.list()
 
             def get_headers(self) -> dict[str, Any]:
-                from openai import OpenAI
+                # from openai import OpenAI
 
-                client = OpenAI()
+                # client = OpenAI(api_key = os.getenv(cls._env_key_name_), base_url = cls._base_url_)
+                client = self.sync_client()
                 response = client.chat.completions.with_raw_response.create(
                     messages=[
                         {
@@ -124,8 +160,8 @@ class OpenAIService(InferenceServiceABC):
                 encoded_image=None,
             ) -> dict[str, Any]:
                 """Calls the OpenAI API and returns the API response."""
-                content = [{"type": "text", "text": user_prompt}]
                 if encoded_image:
+                    content = [{"type": "text", "text": user_prompt}]
                     content.append(
                         {
                             "type": "image_url",
@@ -134,21 +170,28 @@ class OpenAIService(InferenceServiceABC):
                             },
                         }
                     )
-                self.client = AsyncOpenAI()
-                response = await self.client.chat.completions.create(
-                    model=self.model,
-                    messages=[
+                else:
+                    content = user_prompt
+                # self.client = AsyncOpenAI(
+                #     api_key = os.getenv(cls._env_key_name_),
+                #     base_url = cls._base_url_
+                #     )
+                client = self.async_client()
+                params = {
+                    "model": self.model,
+                    "messages": [
                         {"role": "system", "content": system_prompt},
                         {"role": "user", "content": content},
                     ],
-                    temperature=self.temperature,
-                    max_tokens=self.max_tokens,
-                    top_p=self.top_p,
-                    frequency_penalty=self.frequency_penalty,
-                    presence_penalty=self.presence_penalty,
-                    logprobs=self.logprobs,
-                    top_logprobs=self.top_logprobs if self.logprobs else None,
-                )
+                    "temperature": self.temperature,
+                    "max_tokens": self.max_tokens,
+                    "top_p": self.top_p,
+                    "frequency_penalty": self.frequency_penalty,
+                    "presence_penalty": self.presence_penalty,
+                    "logprobs": self.logprobs,
+                    "top_logprobs": self.top_logprobs if self.logprobs else None,
+                }
+                response = await client.chat.completions.create(**params)
                 return response.model_dump()
 
             @staticmethod

edsl/inference_services/registry.py

@@ -6,7 +6,8 @@ from edsl.inference_services.OpenAIService import OpenAIService
 from edsl.inference_services.AnthropicService import AnthropicService
 from edsl.inference_services.DeepInfraService import DeepInfraService
 from edsl.inference_services.GoogleService import GoogleService
+from edsl.inference_services.GroqService import GroqService
 
 default = InferenceServicesCollection(
-    [OpenAIService, AnthropicService, DeepInfraService, GoogleService]
+    [OpenAIService, AnthropicService, DeepInfraService, GoogleService, GroqService]
 )