tools deprecation and new format support (#89)

* tools deprecation and new format support

* Add tool for arxiv (#88)

* enable advanced=True for Google search

* add ConfigurableAction in base

* yahoo tool added

* pubmed tool added

---------

Co-authored-by: lucifertrj <[email protected]>
Co-authored-by: Adhvaith Hundi <[email protected]>

src/openagi/actions/base.py

@@ -4,7 +4,7 @@
 
 from openagi.llms.base import LLMBaseModel
 from openagi.memory.memory import Memory
-
+from typing import ClassVar, Dict, Any
 
 class BaseAction(BaseModel):
     """Base Actions class to be inherited by other actions, providing basic functionality and structure."""
@@ -24,8 +24,7 @@ class BaseAction(BaseModel):
     )
 
     def execute(self):
-        """Executes the action
-        """
+        """Executes the action"""
         raise NotImplementedError("Subclasses must implement this method.")
 
     @classmethod
@@ -43,3 +42,19 @@ def cls_doc(cls):
                 if field_name not in default_exclude_doc_fields
             },
         }
+
+class ConfigurableAction(BaseAction):
+    config: ClassVar[Dict[str, Any]] = {}
+
+    @classmethod
+    def set_config(cls, *args, **kwargs):
+        if args:
+            if len(args) == 1 and isinstance(args[0], dict):
+                cls.config.update(args[0])
+            else:
+                raise ValueError("If using positional arguments, a single dictionary must be provided.")
+        cls.config.update(kwargs)
+
+    @classmethod
+    def get_config(cls, key: str, default: Any = None) -> Any:
+        return cls.config.get(key, default)

src/openagi/actions/files.py

@@ -1,7 +1,6 @@
 import logging
 from pathlib import Path
 from typing import Dict, Optional
-
 from pydantic import Field
 
 from openagi.actions.base import BaseAction

src/openagi/actions/tools/arxiv_search.py

@@ -0,0 +1,57 @@
+from openagi.actions.base import ConfigurableAction
+from pydantic import Field
+from openagi.exception import OpenAGIException
+from typing import ClassVar, Dict, Any
+
+try:
+    import arxiv
+except ImportError:
+    raise OpenAGIException("Install arxiv with cmd `pip install arxiv`")
+
+
+class ArxivSearch(ConfigurableAction):
+    """
+    Arxiv Search is a tool used to search articles in Physics, Mathematics, Computer Science, Quantitative Biology, Quantitative Finance, and Statistics
+    """
+    query: str = Field(..., description="User query or question")
+    max_results: int = Field(10, description="Total results, in int, to be executed from the search. Defaults to 10.")
+
+    def execute(self):
+        search = arxiv.Search(
+        query = self.query,
+        max_results = self.max_results,
+                              )
+        client = arxiv.Client()
+        results = client.results(search)
+        meta_data = ""
+        for result in results:
+            meta_data += f"title : {result.title}\n "
+            meta_data += f"summary : {result.summary}\n "
+            meta_data += f"published : {result.published}\n "
+            meta_data += f"authors : {result.authors}\n "
+            meta_data += f"pdf_url : {result.pdf_url}\n "
+            meta_data += f"entry_id : {result.entry_id}\n\n "
+        return meta_data.strip()
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+

src/openagi/actions/tools/ddg_search.py

@@ -1,11 +1,11 @@
 import json
 from typing import Any
-from openagi.actions.base import BaseAction
+from openagi.actions.base import ConfigurableAction
 from pydantic import Field
 from duckduckgo_search import DDGS
 import logging
 
-class DuckDuckGoSearch(BaseAction):
+class DuckDuckGoSearch(ConfigurableAction):
     """Use this Action to search DuckDuckGo for a query."""
 
     name: str = Field(
@@ -39,14 +39,4 @@ def execute(self):
             self.query,
             max_results=self.max_results,
         )
-        return json.dumps(result)
-
-
-class DuckDuckGoNewsSearch(DuckDuckGoSearch):
-    """Use this Action to get the latest news from DuckDuckGo."""
-
-    def execute(self):
-        ddgs = self._get_ddgs()
-        return json.dumps(
-            ddgs.news(keywords=self.query, max_results=(self.max_results)), indent=2
-        )
+        return json.dumps(result)

src/openagi/actions/tools/document_loader.py

@@ -1,39 +1,64 @@
-from openagi.actions.base import BaseAction
+from typing import Any
+from openagi.actions.base import ConfigurableAction
 from langchain_community.document_loaders import TextLoader
 from langchain_community.document_loaders.csv_loader import CSVLoader
+from langchain_community.document_loaders.pdf import PyPDFLoader
 from pydantic import Field
 
+class TextLoaderTool(ConfigurableAction):
+    """Load content from a text file.
+    
+    This action loads and processes content from .txt files, combining
+    metadata and content into a single context string.
+    """
+
+    def execute(self) -> str:
+        file_path: str = self.get_config('filename')
+        loader = TextLoader(file_path=file_path)
+        documents = loader.load()
+
+        if not documents:
+            return ""
+
+        page_content = documents[0].page_content
+        source = documents[0].metadata["source"]
+        return f"{source} {page_content}"
 
-class DocumentLoader(BaseAction):
-    """Use this Action to extract content from documents"""
+class PDFLoaderTool(ConfigurableAction):
+    """Load content from a PDF file.
+    
+    This action loads and processes content from .pdf files, combining
+    metadata and content into a single context string.
+    """
+
+    def execute(self) -> str:
+        file_path: str = self.get_config('filename')
+        loader = PyPDFLoader(file_path=file_path)
+        documents = loader.load()
+
+        if not documents:
+            return ""
+
+        page_content = documents[0].page_content
+        source = documents[0].metadata["source"]
+        return f"{source} {page_content}"
 
-    file_path: str = Field(
-        default_factory=str,
-        description="File from which content is extracted",
-    )
-
-    def text_loader(self):
-        loader = TextLoader(file_path=self.file_path)
-        data = loader.load()
-        page_content = data[0].page_content
-        meta_data = data[0].metadata["source"]
-        context = meta_data + " " + page_content
-        return context
-
-    def csv_loader(self):
-        content = ""
-        loader = CSVLoader(file_path=self.file_path)
-        data = loader.load()
-
-        for i in range(len(data)):
-            row_content = data[i].page_content
-            row_no = data[i].metadata["row"]
-            content += "row_no" + " " + str(row_no) + ": " + str(row_content)
-        return content
-
-    def execute(self):
-        if self.file_path.endswith(".txt"):
-            context = self.text_loader()
-        elif self.file_path.endswith(".csv"):
-            context = self.csv_loader()
-        return context
+class CSVLoaderTool(ConfigurableAction):
+    """Load content from a CSV file.
+    
+    This action loads and processes content from .csv files, combining
+    row numbers and content into a formatted string representation.
+    """
+
+    def execute(self) -> str:
+        file_path: str = self.get_config('filename')
+        loader = CSVLoader(file_path=file_path)
+        documents = loader.load()
+
+        content_parts = []
+        for idx, doc in enumerate(documents):
+            row_content = doc.page_content
+            row_number = doc.metadata["row"]
+            content_parts.append(f"row_no {row_number}: {row_content}")
+
+        return "".join(content_parts)

src/openagi/actions/tools/exasearch.py

@@ -1,30 +1,63 @@
-from openagi.actions.base import BaseAction
-import os
+from openagi.actions.base import ConfigurableAction
 from pydantic import Field
 from openagi.exception import OpenAGIException
+import os
+import warnings
 
 try:
-   from exa_py import Exa
+    from exa_py import Exa
 except ImportError:
-  raise OpenAGIException("Install Exa Py with cmd `pip install exa_py`")
+    raise OpenAGIException("Install Exa Py with cmd `pip install exa_py`")
 
-class ExaSearch(BaseAction):
-    """
-    Exa Search is a tool used when user needs to ask the question in terms of query to get response 
+class ExaSearch(ConfigurableAction):
+    """Exa Search tool for querying and retrieving information.
+    
+    This action uses the Exa API to perform searches and retrieve relevant content
+    based on user queries. Requires an API key to be configured before use.
     """
-    query: str = Field(..., description="User query or question ")
-
-    def execute(self):
-        api_key = os.environ["EXA_API_KEY"]
+    query: str = Field(..., description="User query or question")
+
+    def __init__(self, **data):
+        super().__init__(**data)
+        self._check_deprecated_usage()
+
+    def _check_deprecated_usage(self):
+        if 'EXA_API_KEY' in os.environ and not self.get_config('api_key'):
+            warnings.warn(
+                "Using environment variables for API keys is deprecated and will be removed in a future version. "
+                "Please use ExaSearch.set_config(api_key='your_key') instead of setting environment variables.",
+                DeprecationWarning,
+                stacklevel=2
+            )
+            self.set_config(api_key=os.environ['EXA_API_KEY'])
+
+
+    def execute(self) -> str:
+        api_key: str = self.get_config('api_key')
+        if not api_key:
+            if 'EXA_API_KEY' in os.environ:
+                api_key = os.environ['EXA_API_KEY']
+                warnings.warn(
+                    "Using environment variables for API keys is deprecated and will be removed in a future version. "
+                    "Please use ExaSearch.set_config(api_key='your_key') instead of setting environment variables.",
+                    DeprecationWarning,
+                    stacklevel=2
+                )
+            else:
+                raise OpenAGIException("API KEY NOT FOUND. Use ExaSearch.set_config(api_key='your_key') to set the API key.")
+
+        exa = Exa(api_key=api_key)
+        results = exa.search_and_contents(
+            self.query,
+            text={"max_characters": 512},
+        )
 
-        exa = Exa(api_key = api_key)
-        results = exa.search_and_contents(self.query, 
-                                    text={"max_characters": 512}, 
-                                )
-        content = ""
-        for idx in results.results:
-            content += idx.text.strip()
+        content_parts = []
+        for result in results.results:
+            content_parts.append(result.text.strip())
 
-        content = content.replace("<|endoftext|>","")
-        content = content.replace("NaN","")
-        return content
+        content = "".join(content_parts)
+        return (
+            content.replace("<|endoftext|>", "")
+                  .replace("NaN", "")
+        )

src/openagi/actions/tools/google_search_tool.py

@@ -0,0 +1,38 @@
+from openagi.actions.base import ConfigurableAction
+from pydantic import Field
+from openagi.exception import OpenAGIException
+import logging
+
+try:
+   from googlesearch import search
+except ImportError:
+  raise OpenAGIException("Install googlesearch-python with cmd `pip install googlesearch-python`")
+
+class GoogleSearchTool(ConfigurableAction):
+    """
+    Google Search is a tool used for scraping the Google search engine. Extract information from Google search results.
+    """
+    query: str = Field(..., description="User query or question ")
+
+    max_results: int = Field(
+        default=10,
+        description="Total results, in int, to be executed from the search. Defaults to 10. The limit should be 10 and not execeed more than 10",
+    )
+
+    lang: str = Field(
+        default="en",
+        description = "specify the langauge for your search results."
+    )
+
+    def execute(self):
+        if self.max_results > 15:
+            logging.info("Over threshold value... Limiting the Max results to 15")
+            self.max_results = 15
+
+        context = ""
+        search_results = search(self.query,num_results=self.max_results,lang=self.lang,advanced=True)
+        for info in search_results:
+            context += f"Title: {info.title}. Description: {info.description}. URL: {info.url}"
+
+        return context
+