Feature/summary

README.md

@@ -204,7 +204,23 @@ HUGGINGFACEHUB_API_TOKEN="MY_HF_API_KEY"
 
 You can also pass `huggingfacehub_api_token` as a named parameter.
 
+#### AWS Bedrock
+Create your access keys in security credentials of your user in AWS.
 
+Then write in the files ```~/.aws/config``` and ````~/.aws/credentials```` for Linux and MacOS or ````%USERPROFILE%\.aws\config```` and ````%USERPROFILE%\.aws\credentials```` for Windows:
+
+In credentials:
+```shell
+[default]
+aws_access_key_id = <YOUR_CREATED_AWS_KEY>
+aws_secret_access_key = <YOUR_CREATED_AWS_SECRET_KEY>
+```
+
+In config:
+```shell
+[default]
+region = <AWS_REGION>
+```
 
 ### ⚙️ Install locally
 
@@ -238,6 +254,7 @@ The current available tasks in Promptmeteo are:
 | `CodeGenerator`      | Code generation                   |
 | `ApiGenerator`       | API REST generation               |
 | `ApiFormatter`       | API REST correction               |
+| `Summarizer`         | Text summarization                |
 
 ### ✅ Available Model
 
@@ -259,3 +276,5 @@ The current available `model_name` and `language` values are:
 | google         | text-bison@001       |    en    |
 | google         | text-bison-32k       |    es    |
 | google         | text-bison-32k       |    en    |
+| bedrock        | anthropic.claude-v2  |    en    |
+| bedrock        | anthropic.claude-v2  |    es    |

promptmeteo/__init__.py

@@ -4,3 +4,4 @@
 from .document_classifier import DocumentClassifier
 from .api_generator import APIGenerator
 from .api_formatter import APIFormatter
+from .summarizer import Summarizer

promptmeteo/api_formatter.py

@@ -382,4 +382,4 @@ def replace_values(orig_dict, replace_dict):
             sort_keys=False,
         )
 
-        return api
+        return api

promptmeteo/models/__init__.py

@@ -29,6 +29,7 @@
 from .hf_hub_api import HFHubApiLLM
 from .hf_pipeline import HFPipelineLLM
 from .google_vertexai import GoogleVertexAILLM
+from .bedrock import BedrockLLM
 
 
 class ModelProvider(str, Enum):
@@ -42,6 +43,7 @@ class ModelProvider(str, Enum):
     PROVIDER_2: str = "hf_hub_api"
     PROVIDER_3: str = "hf_pipeline"
     PROVIDER_4: str = "google-vertexai"
+    PROVIDER_5: str = "bedrock"
 
 
 class ModelFactory:
@@ -57,6 +59,7 @@ class ModelFactory:
         ModelProvider.PROVIDER_2: HFHubApiLLM,
         ModelProvider.PROVIDER_3: HFPipelineLLM,
         ModelProvider.PROVIDER_3: GoogleVertexAILLM,
+        ModelProvider.PROVIDER_5: BedrockLLM
     }
 
     @classmethod
@@ -87,6 +90,9 @@ def factory_method(
 
         elif model_provider_name == ModelProvider.PROVIDER_4.value:
             model_cls = GoogleVertexAILLM
+
+        elif model_provider_name == ModelProvider.PROVIDER_5.value:
+            model_cls = BedrockLLM
 
         else:
             raise ValueError(

promptmeteo/models/bedrock.py

@@ -0,0 +1,127 @@
+#!/usr/bin/python3
+
+#  Copyright (c) 2023 Paradigma Digital S.L.
+
+#  Permission is hereby granted, free of charge, to any person obtaining a copy
+#  of this software and associated documentation files (the "Software"), to deal
+#  in the Software without restriction, including without limitation the rights
+#  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+#  copies of the Software, and to permit persons to whom the Software is
+#  furnished to do so, subject to the following conditions:
+
+#  The above copyright notice and this permission notice shall be included in
+#  all copies or substantial portions of the Software.
+
+#  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+#  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+#  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+#  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+#  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+#  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+#  THE SOFTWARE.
+
+from enum import Enum
+from typing import Dict
+from typing import Optional
+import os
+import boto3
+from langchain.llms.bedrock import Bedrock
+from langchain.embeddings import HuggingFaceEmbeddings
+
+from .base import BaseModel
+
+
+class ModelTypes(str, Enum):
+
+    """
+    Enum of available model types.
+    """
+
+    AnthropicClaudeV2: str = "anthropic.claude-v2"
+
+    @classmethod
+    def has_value(
+        cls,
+        value: str,
+    ) -> bool:
+        """
+        Checks if the value is in the enum or not.
+        """
+
+        return value in cls._value2member_map_
+
+
+class ModelEnum(Enum):
+
+    """
+    Model Parameters.
+    """
+
+    class AnthropicClaudeV2:
+
+        """
+        Default parameters for Anthropic Claude V2
+        """
+
+        client = Bedrock
+        embedding = HuggingFaceEmbeddings
+        model_task: str = "text2text-generation"
+        params: dict = {
+            'max_tokens_to_sample': 2048,
+            'temperature': 0.3,
+            'top_k': 250,
+            'top_p': 0.999,
+            'stop_sequences': ['Human:']
+        }
+
+
+class BedrockLLM(BaseModel):
+
+    """
+    Bedrock LLM model.
+    """
+
+    def __init__(
+        self,
+        model_name: Optional[str] = "",
+        model_params: Optional[Dict] = None,
+        model_provider_token: Optional[str] = "",
+        **kwargs
+    ) -> None:
+        """
+        Make predictions using a model from OpenAI.
+        It will use the os environment called OPENAI_ORGANIZATION for instance the LLM
+        """
+
+        if not ModelTypes.has_value(model_name):
+            raise ValueError(
+                f"`model_name`={model_name} not in supported model names: "
+                f"{[i.name for i in ModelTypes]}"
+            )
+        self.boto3_bedrock = boto3.client('bedrock-runtime', **kwargs)
+        super(BedrockLLM, self).__init__()
+
+        # Model name
+        model = ModelTypes(model_name).name
+
+        # Model parameters
+        if not model_params:
+            model_params = (
+                ModelEnum[model].value.params
+                if not model_params
+                else model_params
+            )
+        self.model_params = model_params
+
+        # Model
+        self._llm = ModelEnum[model].value.client(
+            model_id=model_name,
+            model_kwargs=self.model_params,
+            client =  self.boto3_bedrock
+        )
+
+        embedding_name = "sentence-transformers/all-MiniLM-L6-v2"
+        if os.path.exists("/home/models/all-MiniLM-L6-v2"):
+            embedding_name = "/home/models/all-MiniLM-L6-v2"
+
+        self._embeddings = HuggingFaceEmbeddings(model_name=embedding_name)

promptmeteo/parsers/__init__.py

@@ -27,6 +27,7 @@
 from .base import BaseParser
 from .dummy_parser import DummyParser
 from .classification_parser import ClassificationParser
+from .json_parser import JSONParser
 
 
 class ParserTypes(str, Enum):
@@ -41,6 +42,8 @@ class ParserTypes(str, Enum):
     PARSER_4: str = "code-generation"
     PARSER_5: str = "api-generation"
     PARSER_6: str = "api-correction"
+    PARSER_7: str = "json-info-extraction"
+    PARSER_8: str = "summarization"
 
 
 class ParserFactory:
@@ -77,6 +80,12 @@ def factory_method(
 
         elif task_type == ParserTypes.PARSER_6.value:
             parser_cls = ApiParser
+
+        elif task_type == ParserTypes.PARSER_7.value:
+            parser_cls = JSONParser
+
+        elif task_type == ParserTypes.PARSER_8.value:
+            parser_cls = DummyParser
 
         else:
             raise ValueError(

promptmeteo/parsers/json_parser.py

@@ -0,0 +1,67 @@
+#!/usr/bin/python3
+
+#  Copyright (c) 2023 Paradigma Digital S.L.
+
+#  Permission is hereby granted, free of charge, to any person obtaining a copy
+#  of this software and associated documentation files (the "Software"), to deal
+#  in the Software without restriction, including without limitation the rights
+#  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+#  copies of the Software, and to permit persons to whom the Software is
+#  furnished to do so, subject to the following conditions:
+
+#  The above copyright notice and this permission notice shall be included in
+#  all copies or substantial portions of the Software.
+
+#  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+#  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+#  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+#  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+#  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+#  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+#  THE SOFTWARE.
+
+from typing import List
+import re
+from .base import BaseParser
+import regex
+import json
+
+
+class JSONParser(BaseParser):
+
+    """
+    Parser for potential JSON outputs
+    """
+
+    def run(
+        self,
+        text: str,
+    ) -> List[str]:
+        """
+        Given a response string from an LLM, returns the response expected for
+        the task.
+        """
+
+        try:
+            json_output = self._preprocess(text)
+            json_obtanaied = json.loads(json_output)
+            return json_output
+        except:
+            return ""
+
+
+    def _preprocess(
+        self,
+        text: str,
+    ) -> str:
+        """
+        Preprocess output string before parsing result to solve common mistakes
+        such as end-of-line presence and beginning and finishing with empty
+        space.
+        """
+        pattern = regex.compile(r'\{(?:[^{}]|(?R))*\}')
+        str_json = pattern.findall(text)[0]
+
+        str_json = str_json.replace("'",'"')
+
+        return str_json

promptmeteo/prompts/anthropic.claude-v2_en_classification.prompt

@@ -0,0 +1,61 @@
+#  Copyright (c) 2023 Paradigma Digital S.L.
+
+#  Permission is hereby granted, free of charge, to any person obtaining a copy
+#  of this software and associated documentation files (the "Software"), to deal
+#  in the Software without restriction, including without limitation the rights
+#  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+#  copies of the Software, and to permit persons to whom the Software is
+#  furnished to do so, subject to the following conditions:
+
+#  The above copyright notice and this permission notice shall be included in
+#  all copies or substantial portions of the Software.
+
+#  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+#  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+#  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+#  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+#  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+#  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+#  THE SOFTWARE.
+
+
+TEMPLATE:
+    "I need you to help me with a text classification task.
+    {__PROMPT_DOMAIN__}
+    {__PROMPT_LABELS__}
+
+    {__CHAIN_THOUGHT__}
+    {__ANSWER_FORMAT__}
+    {__SHOT_EXAMPLES__}
+    {__PROMPT_SAMPLE__}"
+
+
+PROMPT_DOMAIN:
+    "The texts you will be processing are from the {__DOMAIN__} domain."
+
+
+PROMPT_LABELS:
+    "I want you to classify the texts into one of the following categories:
+    {__LABELS__}."
+
+
+PROMPT_DETAIL:
+    ""
+
+SHOT_EXAMPLES:
+    "Examples:\n\n{__EXAMPLES__}"
+
+PROMPT_SAMPLE:
+    "\n\n{__SAMPLE__}\n"
+
+CHAIN_THOUGHT:
+    "Please provide a step-by-step argument for your answer, explain why you
+    believe your final choice is justified, and make sure to conclude your
+    explanation with the name of the class you have selected as the correct
+    one, in lowercase and without punctuation."
+
+
+ANSWER_FORMAT:
+    "In your response, include only the name of the class as a single word, in
+    lowercase, without punctuation, and without adding any other statements or
+    words."