langroid 0.1.39__tar.gz → 0.1.41__tar.gz

{langroid-0.1.39 → langroid-0.1.41}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: langroid
-Version: 0.1.39
+Version: 0.1.41
 Summary: Harness LLMs with Multi-Agent Programming
 License: MIT
 Author: Prasad Chalasani
@@ -39,6 +39,7 @@ Requires-Dist: openai (>=0.27.5,<0.28.0)
 Requires-Dist: pandas (>=2.0.3,<3.0.0)
 Requires-Dist: pre-commit (>=3.3.2,<4.0.0)
 Requires-Dist: prettytable (>=3.8.0,<4.0.0)
+Requires-Dist: psycopg2 (>=2.9.7,<3.0.0)
 Requires-Dist: pydantic (==1.10.11)
 Requires-Dist: pygithub (>=1.58.1,<2.0.0)
 Requires-Dist: pygments (>=2.15.1,<3.0.0)
@@ -77,6 +78,8 @@ Description-Content-Type: text/markdown
 [![Static Badge](https://img.shields.io/badge/Discord-orange?logoColor=orange&link=https%3A%2F%2Fdiscord.gg%2FZU36McDgDs)](https://discord.gg/ZU36McDgDs)
 [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb)
 
+[![Share on Twitter](https://img.shields.io/twitter/url?style=social&url=https://github.com/langroid/langroid)](https://twitter.com/intent/tweet?text=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.%20https://github.com/langroid/langroid)
+[![Share on LinkedIn](https://img.shields.io/badge/Share%20on-LinkedIn-blue)](https://www.linkedin.com/shareArticle?mini=true&url=https://github.com/langroid/langroid&title=Langroid:%20A%20Powerful,%20Elegant%20Framework&summary=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.)
 
 </div>
 
@@ -114,7 +117,8 @@ for ideas on what to contribute.
 <summary> <b>:fire: Updates/Releases</b></summary>
 
 - **Aug 2023:**
-  - **0.1.38:** [`GoogleSearchTool`](langroid/agent/stateless_tools/google_search_tool.py) to enable Agents (their LLM) to do Google searches via function-calling/tools.
+  - **Example:** [Answer questions](examples/docqa/chat-search.py) using Google Search + vecdb-retrieval from URL contents. 
+  - **0.1.39:** [`GoogleSearchTool`](langroid/agent/stateless_tools/google_search_tool.py) to enable Agents (their LLM) to do Google searches via function-calling/tools.
     See [this chat example](examples/basic/chat-search.py) for how easy it is to add this tool to an agent.
   - **Colab notebook** to try the quick-start examples: [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb) 
   - **0.1.37:** Added [`SQLChatAgent`](langroid/agent/special/sql_chat_agent.py) -- thanks to our latest contributor [Rithwik Babu](https://github.com/rithwikbabu)!
@@ -232,7 +236,8 @@ export OPENAI_API_KEY=your-key-here-without-quotes
 <details>
 <summary><b>Optional Setup Instructions (click to expand) </b></summary>
 
-All of the below are optional and not strictly needed to run any of the examples.
+All of the following environment variable settings are optional, and some are only needed 
+to use specific features (as noted below).
 
 - **Qdrant** Vector Store API Key, URL. This is only required if you want to use Qdrant cloud.
   You can sign up for a free 1GB account at [Qdrant cloud](https://cloud.qdrant.io).
@@ -357,8 +362,6 @@ task.run() # ... a loop seeking response from LLM or User at each turn
 <details>
 <summary><b> Three communicating agents </b></summary>
 
-```python
-
 A toy numbers game, where when given a number `n`:
 - `repeater_agent`'s LLM simply returns `n`,
 - `even_agent`'s LLM returns `n/2` if `n` is even, else says "DO-NOT-KNOW"
@@ -663,13 +666,24 @@ script in the `langroid-examples` repo.
 
 # :heart: Thank you to our [supporters](https://github.com/langroid/langroid/stargazers)
 
-If you like this repo, don't forget to leave a star :star: !
+If you like this project, please give it a star ⭐ and 📢 spread the word in your network:
+
+[![Share on Twitter](https://img.shields.io/twitter/url?style=social&url=https://github.com/langroid/langroid)](https://twitter.com/intent/tweet?text=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.%20https://github.com/langroid/langroid)
+[![Share on LinkedIn](https://img.shields.io/badge/Share%20on-LinkedIn-blue)](https://www.linkedin.com/shareArticle?mini=true&url=https://github.com/langroid/langroid&title=Langroid:%20A%20Powerful,%20Elegant%20Framework&summary=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.)
+
+
+
+
+Your support will help build Langroid's momentum and community.
+
+
+
+
+# Langroid Co-Founders
+
+- [Prasad Chalasani](https://www.linkedin.com/in/pchalasani/) (IIT BTech/CS, CMU PhD/ML; Independent ML Consultant)
+- [Somesh Jha](https://www.linkedin.com/in/somesh-jha-80208015/) (IIT BTech/CS, CMU PhD/CS; Professor of CS, U Wisc at Madison)
 
-# Contributors
 
-- Prasad Chalasani (IIT BTech/CS, CMU PhD/ML; Independent ML Consultant)
-- Somesh Jha (IIT BTech/CS, CMU PhD/CS; Professor of CS, U Wisc at Madison)
-- Mohannad Alhanahnah (Research Associate, U Wisc at Madison)
-- Ashish Hooda (IIT BTech/CS; PhD Candidate, U Wisc at Madison)
 
 

{langroid-0.1.39 → langroid-0.1.41}/README.md

@@ -13,6 +13,8 @@
 [![Static Badge](https://img.shields.io/badge/Discord-orange?logoColor=orange&link=https%3A%2F%2Fdiscord.gg%2FZU36McDgDs)](https://discord.gg/ZU36McDgDs)
 [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb)
 
+[![Share on Twitter](https://img.shields.io/twitter/url?style=social&url=https://github.com/langroid/langroid)](https://twitter.com/intent/tweet?text=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.%20https://github.com/langroid/langroid)
+[![Share on LinkedIn](https://img.shields.io/badge/Share%20on-LinkedIn-blue)](https://www.linkedin.com/shareArticle?mini=true&url=https://github.com/langroid/langroid&title=Langroid:%20A%20Powerful,%20Elegant%20Framework&summary=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.)
 
 </div>
 
@@ -50,7 +52,8 @@ for ideas on what to contribute.
 <summary> <b>:fire: Updates/Releases</b></summary>
 
 - **Aug 2023:**
-  - **0.1.38:** [`GoogleSearchTool`](langroid/agent/stateless_tools/google_search_tool.py) to enable Agents (their LLM) to do Google searches via function-calling/tools.
+  - **Example:** [Answer questions](examples/docqa/chat-search.py) using Google Search + vecdb-retrieval from URL contents. 
+  - **0.1.39:** [`GoogleSearchTool`](langroid/agent/stateless_tools/google_search_tool.py) to enable Agents (their LLM) to do Google searches via function-calling/tools.
     See [this chat example](examples/basic/chat-search.py) for how easy it is to add this tool to an agent.
   - **Colab notebook** to try the quick-start examples: [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb) 
   - **0.1.37:** Added [`SQLChatAgent`](langroid/agent/special/sql_chat_agent.py) -- thanks to our latest contributor [Rithwik Babu](https://github.com/rithwikbabu)!
@@ -168,7 +171,8 @@ export OPENAI_API_KEY=your-key-here-without-quotes
 <details>
 <summary><b>Optional Setup Instructions (click to expand) </b></summary>
 
-All of the below are optional and not strictly needed to run any of the examples.
+All of the following environment variable settings are optional, and some are only needed 
+to use specific features (as noted below).
 
 - **Qdrant** Vector Store API Key, URL. This is only required if you want to use Qdrant cloud.
   You can sign up for a free 1GB account at [Qdrant cloud](https://cloud.qdrant.io).
@@ -293,8 +297,6 @@ task.run() # ... a loop seeking response from LLM or User at each turn
 <details>
 <summary><b> Three communicating agents </b></summary>
 
-```python
-
 A toy numbers game, where when given a number `n`:
 - `repeater_agent`'s LLM simply returns `n`,
 - `even_agent`'s LLM returns `n/2` if `n` is even, else says "DO-NOT-KNOW"
@@ -599,12 +601,23 @@ script in the `langroid-examples` repo.
 
 # :heart: Thank you to our [supporters](https://github.com/langroid/langroid/stargazers)
 
-If you like this repo, don't forget to leave a star :star: !
+If you like this project, please give it a star ⭐ and 📢 spread the word in your network:
+
+[![Share on Twitter](https://img.shields.io/twitter/url?style=social&url=https://github.com/langroid/langroid)](https://twitter.com/intent/tweet?text=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.%20https://github.com/langroid/langroid)
+[![Share on LinkedIn](https://img.shields.io/badge/Share%20on-LinkedIn-blue)](https://www.linkedin.com/shareArticle?mini=true&url=https://github.com/langroid/langroid&title=Langroid:%20A%20Powerful,%20Elegant%20Framework&summary=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.)
+
+
+
+
+Your support will help build Langroid's momentum and community.
+
+
+
+
+# Langroid Co-Founders
+
+- [Prasad Chalasani](https://www.linkedin.com/in/pchalasani/) (IIT BTech/CS, CMU PhD/ML; Independent ML Consultant)
+- [Somesh Jha](https://www.linkedin.com/in/somesh-jha-80208015/) (IIT BTech/CS, CMU PhD/CS; Professor of CS, U Wisc at Madison)
 
-# Contributors
 
-- Prasad Chalasani (IIT BTech/CS, CMU PhD/ML; Independent ML Consultant)
-- Somesh Jha (IIT BTech/CS, CMU PhD/CS; Professor of CS, U Wisc at Madison)
-- Mohannad Alhanahnah (Research Associate, U Wisc at Madison)
-- Ashish Hooda (IIT BTech/CS; PhD Candidate, U Wisc at Madison)
 

{langroid-0.1.39 → langroid-0.1.41}/langroid/agent/base.py

@@ -128,9 +128,10 @@ class Agent(ABC):
                 raise ValueError("message_class must be a subclass of ToolMessage")
             tool = message_class.default_value("request")
             self.llm_tools_map[tool] = message_class
-            if hasattr(message_class, "handle"):
+            if hasattr(message_class, "handle") and not hasattr(self, tool):
                 """
                 If the message class has a `handle` method,
+                and does NOT have a method with the same name as the tool,
                 then we create a method for the agent whose name
                 is the value of `tool`, and whose body is the `handle` method.
                 This removes a separate step of having to define this method

{langroid-0.1.39 → langroid-0.1.41}/langroid/agent/special/doc_chat_agent.py

@@ -304,14 +304,21 @@ class DocChatAgent(ChatAgent):
         )
 
     @no_type_check
-    def answer_from_docs(self, query: str) -> Document:
-        """Answer query based on docs in vecdb, and conv history"""
-        response = Document(
-            content=NO_ANSWER,
-            metadata=DocMetaData(
-                source="None",
-            ),
-        )
+    def get_relevant_extracts(self, query: str) -> List[Document]:
+        """
+        Get list of docs or extracts relevant to a query. These could be:
+        - the original docs, if they exist and are not too long, or
+        - a list of doc-chunks retrieved from the VecDB
+            that are "relevant" to the query, if these are not too long, or
+        - a list of relevant extracts from these doc-chunks
+
+        Args:
+            query (str): query to search for
+
+        Returns:
+            List[Document]: list of relevant docs
+
+        """
         if len(self.dialog) > 0 and not self.config.conversation_mode:
             # In conversation mode, we let self.message_history accumulate
             # and do not need to convert to standalone query
@@ -324,7 +331,7 @@ class DocChatAgent(ChatAgent):
 
         passages = self.original_docs
 
-        # if original docs too long, no need to look for relevant parts.
+        # if original docs not too long, no need to look for relevant parts.
         if (
             passages is None
             or self.original_docs_length > self.config.max_context_tokens
@@ -335,7 +342,7 @@ class DocChatAgent(ChatAgent):
                     k=self.config.parsing.n_similar_docs,
                 )
             if len(docs_and_scores) == 0:
-                return response
+                return []
             passages = [
                 Document(content=d.content, metadata=d.metadata)
                 for (d, _) in docs_and_scores
@@ -347,6 +354,29 @@ class DocChatAgent(ChatAgent):
             with console.status("[cyan]LLM Extracting verbatim passages..."):
                 with StreamingIfAllowed(self.llm, False):
                     extracts = self.llm.get_verbatim_extracts(query, passages)
+
+        return extracts
+
+    @no_type_check
+    def answer_from_docs(self, query: str) -> Document:
+        """
+        Answer query based on relevant docs from the VecDB
+
+        Args:
+            query (str): query to answer
+
+        Returns:
+            Document: answer
+        """
+        response = Document(
+            content=NO_ANSWER,
+            metadata=DocMetaData(
+                source="None",
+            ),
+        )
+        extracts = self.get_relevant_extracts(query)
+        if len(extracts) == 0:
+            return response
         with ExitStack() as stack:
             # conditionally use Streaming or rich console context
             cm = (

{langroid-0.1.39 → langroid-0.1.41}/langroid/agent/special/retriever_agent.py

@@ -97,7 +97,7 @@ class RetrieverAgent(DocChatAgent, ABC):
             query_str = query.content
         else:
             query_str = query
-        docs = self.get_relevant_docs(query_str)
+        docs = self.get_relevant_extracts(query_str)
         if len(docs) == 0:
             return None
         content = "\n\n".join([d.content for d in docs])
@@ -135,7 +135,7 @@ class RetrieverAgent(DocChatAgent, ABC):
         ]
         return docs
 
-    def get_relevant_docs(self, query: str) -> List[Document]:
+    def get_relevant_extracts(self, query: str) -> List[Document]:
         """
         Given a query, get the records/docs whose contents are most relevant to the
             query. First get nearest docs from vector store, then select the best

{langroid-0.1.39 → langroid-0.1.41}/langroid/agent/special/sql_chat_agent.py

@@ -28,9 +28,9 @@ console = Console()
 
 DEFAULT_SQL_CHAT_SYSTEM_MESSAGE = """
 You are a savvy data scientist/database administrator, with expertise in 
-answering questions by querying a SQL database.
+answering questions by querying a {dialect} database.
 You do not have access to the database 'db' directly, so you will need to use the 
-`run_query` tool/function-call to answer the question.
+`run_query` tool/function-call to answer questions.
 
 The below JSON schema maps the SQL database structure. It outlines tables, each 
 with a description and columns. Each table is identified by a key, 
@@ -146,7 +146,7 @@ class SQLChatAgent(ChatAgent):
 
         # Update the system message with the table information
         self.config.system_message = self.config.system_message.format(
-            schema_dict=schema_dict
+            schema_dict=schema_dict, dialect=self.engine.dialect.name
         )
 
         super().__init__(config)

{langroid-0.1.39 → langroid-0.1.41}/langroid/agent/special/table_chat_agent.py

@@ -10,7 +10,9 @@ the code and returns the result as a string.
 import io
 import logging
 import sys
+from typing import no_type_check
 
+import numpy as np
 import pandas as pd
 from rich.console import Console
 
@@ -26,16 +28,72 @@ logger = logging.getLogger(__name__)
 console = Console()
 
 DEFAULT_TABLE_CHAT_SYSTEM_MESSAGE = """
-You are a savvy data scientist, with expertise in analyzing tabular dataset,
+You are a savvy data scientist, with expertise in analyzing tabular datasets,
 using Python and the Pandas library for dataframe manipulation.
 Since you do not have access to the dataframe 'df', you
 will need to use the `run_code` tool/function-call to answer the question.
-The columns in the dataframe are:
-{columns}
+Here is a summary of the dataframe:
+{summary}
 Do not assume any columns other than those shown.
+In the code you submit to the `run_code` tool/function, 
+do not forget to include any necessary imports, such as `import pandas as pd`.
+Sometimes you may not be able to answer the question in a single call to `run_code`,
+so you can use a series of calls to `run_code` to build up the answer. 
+For example you may first want to know something about the possible values in a column.
+
+If you receive a null or other unexpected result, see if you have made an assumption
+in your code, and try another way, or use `run_code` to explore the dataframe 
+before submitting your final code. 
+
+Start by asking me what I want to know about the data.
 """
 
 
+@no_type_check
+def dataframe_summary(df: pd.DataFrame) -> str:
+    """
+    Generate a structured summary for a pandas DataFrame containing numerical
+    and categorical values.
+
+    Args:
+        df (pd.DataFrame): The input DataFrame to summarize.
+
+    Returns:
+        str: A nicely structured and formatted summary string.
+    """
+
+    # Column names display
+    col_names_str = (
+        "COLUMN NAMES:\n" + " ".join([f"'{col}'" for col in df.columns]) + "\n\n"
+    )
+
+    # Numerical data summary
+    num_summary = df.describe().applymap(lambda x: "{:.2f}".format(x))
+    num_str = "Numerical Column Summary:\n" + num_summary.to_string() + "\n\n"
+
+    # Categorical data summary
+    cat_columns = df.select_dtypes(include=[np.object_]).columns
+    cat_summary_list = []
+
+    for col in cat_columns:
+        unique_values = df[col].unique()
+        if len(unique_values) < 10:
+            cat_summary_list.append(f"'{col}': {', '.join(map(str, unique_values))}")
+        else:
+            cat_summary_list.append(f"'{col}': {df[col].nunique()} unique values")
+
+    cat_str = "Categorical Column Summary:\n" + "\n".join(cat_summary_list) + "\n\n"
+
+    # Missing values summary
+    nan_summary = df.isnull().sum().rename("missing_values").to_frame()
+    nan_str = "Missing Values Column Summary:\n" + nan_summary.to_string() + "\n"
+
+    # Combine the summaries into one structured string
+    summary_str = col_names_str + num_str + cat_str + nan_str
+
+    return summary_str
+
+
 class TableChatAgentConfig(ChatAgentConfig):
     system_message: str = DEFAULT_TABLE_CHAT_SYSTEM_MESSAGE
     user_message: None | str = None
@@ -77,10 +135,8 @@ class TableChatAgent(ChatAgent):
             df = read_tabular_data(config.data, config.separator)
 
         self.df = df
-        columns_with_quotes = [f"'{c}'" for c in df.columns]
-        config.system_message = config.system_message.format(
-            columns=", ".join(columns_with_quotes)
-        )
+        summary = dataframe_summary(df)
+        config.system_message = config.system_message.format(summary=summary)
 
         super().__init__(config)
         self.config: TableChatAgentConfig = config

langroid-0.1.41/langroid/agent/stateless_tools/google_search_tool.py

@@ -0,0 +1,28 @@
+"""
+A tool to trigger a Google search for a given query, and return the top results with
+their titles, links, summaries. Since the tool is stateless (i.e. does not need
+access to agent state), it can be enabled for any agent, without having to define a
+special method inside the agent: `agent.enable_message(GoogleSearchTool)`
+
+NOTE: Using this tool requires setting the GOOGLE_API_KEY and GOOGLE_CSE_ID
+environment variables in your `.env` file, as explained in the
+[README](https://github.com/langroid/langroid#gear-installation-and-setup).
+"""
+
+from langroid.agent.tool_message import ToolMessage
+from langroid.parsing.web_search import google_search
+
+
+class GoogleSearchTool(ToolMessage):
+    request: str = "web_search"
+    purpose: str = """
+            To search the web and return up to <num_results> links relevant to 
+            the given <query>. 
+            """
+    query: str
+    num_results: int
+
+    def handle(self) -> str:
+        search_results = google_search(self.query, self.num_results)
+        # return Title, Link, Summary of each result, separated by two newlines
+        return "\n\n".join(str(result) for result in search_results)

{langroid-0.1.39 → langroid-0.1.41}/langroid/language_models/openai_gpt.py

@@ -4,7 +4,7 @@ import logging
 import os
 import sys
 from enum import Enum
-from typing import Any, Dict, List, Optional, Tuple, Union, cast
+from typing import Any, Dict, List, Optional, Tuple, Union
 
 import openai
 from dotenv import load_dotenv
@@ -361,13 +361,13 @@ class OpenAIGPT(LanguageModel):
             LLMResponse object
         """
         openai.api_key = self.api_key
-        if type(messages) == str:
+        if isinstance(messages, str):
             llm_messages = [
                 LLMMessage(role=Role.SYSTEM, content="You are a helpful assistant."),
                 LLMMessage(role=Role.USER, content=messages),
             ]
         else:
-            llm_messages = cast(List[LLMMessage], messages)
+            llm_messages = messages
 
         @retry_with_exponential_backoff
         def completions_with_backoff(**kwargs):  # type: ignore

langroid-0.1.39/langroid/agent/stateless_tools/google_search_tool.py → langroid-0.1.41/langroid/parsing/web_search.py

@@ -1,3 +1,11 @@
+"""
+Utilities for web search.
+
+NOTE: Using Google Search requires setting the GOOGLE_API_KEY and GOOGLE_CSE_ID
+environment variables in your `.env` file, as explained in the
+[README](https://github.com/langroid/langroid#gear-installation-and-setup).
+"""
+
 import os
 from typing import Dict, List
 
@@ -7,12 +15,10 @@ from dotenv import load_dotenv
 from googleapiclient.discovery import Resource, build
 from requests.models import Response
 
-from langroid.agent.tool_message import ToolMessage
 
-
-class GoogleSearchResult:
+class WebSearchResult:
     """
-    Class representing a Google Search result, containing the title, link,
+    Class representing a Web Search result, containing the title, link,
     summary and full content of the result.
     """
 
@@ -58,29 +64,16 @@ class GoogleSearchResult:
         }
 
 
-class GoogleSearchTool(ToolMessage):
-    request: str = "web_search"
-    purpose: str = """
-            To search the web and return up to <num_results> links relevant to 
-            the given <query>. 
-            """
-    query: str
-    num_results: int
-
-    def handle(self) -> str:
-        load_dotenv()
-        api_key = os.getenv("GOOGLE_API_KEY")
-        cse_id = os.getenv("GOOGLE_CSE_ID")
-        service: Resource = build("customsearch", "v1", developerKey=api_key)
-        raw_results = (
-            service.cse()
-            .list(q=self.query, cx=cse_id, num=self.num_results)
-            .execute()["items"]
-        )
+def google_search(query: str, num_results: int = 5) -> List[WebSearchResult]:
+    load_dotenv()
+    api_key = os.getenv("GOOGLE_API_KEY")
+    cse_id = os.getenv("GOOGLE_CSE_ID")
+    service: Resource = build("customsearch", "v1", developerKey=api_key)
+    raw_results = (
+        service.cse().list(q=query, cx=cse_id, num=num_results).execute()["items"]
+    )
 
-        search_results: List[GoogleSearchResult] = [
-            GoogleSearchResult(result["title"], result["link"], 3500, 300)
-            for result in raw_results
-        ]
-        # return Title and Link of each result, separated by two newlines
-        return "\n\n".join(str(result) for result in search_results)
+    return [
+        WebSearchResult(result["title"], result["link"], 3500, 300)
+        for result in raw_results
+    ]

{langroid-0.1.39 → langroid-0.1.41}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langroid"
-version = "0.1.39"
+version = "0.1.41"
 description = "Harness LLMs with Multi-Agent Programming"
 authors = ["Prasad Chalasani <pchalasani@gmail.com>"]
 readme = "README.md"
@@ -59,6 +59,7 @@ sqlalchemy = "^2.0.19"
 prettytable = "^3.8.0"
 google-api-python-client = "^2.95.0"
 lxml = "^4.9.3"
+psycopg2 = "^2.9.7"
 
 [tool.poetry.group.dev.dependencies]
 pytest = "^7.3.1"

langroid-0.1.41/setup.py

@@ -0,0 +1,94 @@
+# -*- coding: utf-8 -*-
+from setuptools import setup
+
+packages = \
+['langroid',
+ 'langroid.agent',
+ 'langroid.agent.special',
+ 'langroid.agent.stateless_tools',
+ 'langroid.cachedb',
+ 'langroid.embedding_models',
+ 'langroid.language_models',
+ 'langroid.parsing',
+ 'langroid.prompts',
+ 'langroid.scripts',
+ 'langroid.utils',
+ 'langroid.utils.llms',
+ 'langroid.utils.output',
+ 'langroid.utils.web',
+ 'langroid.vector_store']
+
+package_data = \
+{'': ['*']}
+
+install_requires = \
+['autopep8>=2.0.2,<3.0.0',
+ 'black[jupyter]>=23.3.0,<24.0.0',
+ 'bs4>=0.0.1,<0.0.2',
+ 'chromadb>=0.3.21,<0.4.0',
+ 'colorlog>=6.7.0,<7.0.0',
+ 'faker>=18.9.0,<19.0.0',
+ 'fakeredis>=2.12.1,<3.0.0',
+ 'fire>=0.5.0,<0.6.0',
+ 'flake8>=6.0.0,<7.0.0',
+ 'google-api-python-client>=2.95.0,<3.0.0',
+ 'halo>=0.0.31,<0.0.32',
+ 'lxml>=4.9.3,<5.0.0',
+ 'mkdocs-awesome-pages-plugin>=2.8.0,<3.0.0',
+ 'mkdocs-gen-files>=0.4.0,<0.5.0',
+ 'mkdocs-jupyter>=0.24.1,<0.25.0',
+ 'mkdocs-literate-nav>=0.6.0,<0.7.0',
+ 'mkdocs-material>=9.1.5,<10.0.0',
+ 'mkdocs-section-index>=0.3.5,<0.4.0',
+ 'mkdocs>=1.4.2,<2.0.0',
+ 'mkdocstrings[python]>=0.21.2,<0.22.0',
+ 'momento>=1.7.0,<2.0.0',
+ 'mypy>=1.2.0,<2.0.0',
+ 'nltk>=3.8.1,<4.0.0',
+ 'openai>=0.27.5,<0.28.0',
+ 'pandas>=2.0.3,<3.0.0',
+ 'pre-commit>=3.3.2,<4.0.0',
+ 'prettytable>=3.8.0,<4.0.0',
+ 'psycopg2>=2.9.7,<3.0.0',
+ 'pydantic==1.10.11',
+ 'pygithub>=1.58.1,<2.0.0',
+ 'pygments>=2.15.1,<3.0.0',
+ 'pyparsing>=3.0.9,<4.0.0',
+ 'pypdf>=3.12.2,<4.0.0',
+ 'python-dotenv>=1.0.0,<2.0.0',
+ 'qdrant-client>=1.3.1,<2.0.0',
+ 'redis>=4.5.5,<5.0.0',
+ 'requests-oauthlib>=1.3.1,<2.0.0',
+ 'requests>=2.31.0,<3.0.0',
+ 'rich>=13.3.4,<14.0.0',
+ 'ruff>=0.0.270,<0.0.271',
+ 'sqlalchemy>=2.0.19,<3.0.0',
+ 'tiktoken>=0.3.3,<0.4.0',
+ 'trafilatura>=1.5.0,<2.0.0',
+ 'typer>=0.7.0,<0.8.0',
+ 'types-redis>=4.5.5.2,<5.0.0.0',
+ 'types-requests>=2.31.0.1,<3.0.0.0',
+ 'wget>=3.2,<4.0']
+
+extras_require = \
+{'hf-embeddings': ['sentence-transformers==2.2.2', 'torch==2.0.0']}
+
+setup_kwargs = {
+    'name': 'langroid',
+    'version': '0.1.41',
+    'description': 'Harness LLMs with Multi-Agent Programming',
+    'long_description': '<div align="center">\n  <img src="docs/assets/langroid-card-lambda-ossem-rust-1200-630.png" alt="Logo" \n        width="400" align="center">\n</div>\n\n<div align="center">\n\n[![Pytest](https://github.com/langroid/langroid/actions/workflows/pytest.yml/badge.svg)](https://github.com/langroid/langroid/actions/workflows/pytest.yml)\n[![codecov](https://codecov.io/gh/langroid/langroid/branch/main/graph/badge.svg?token=H94BX5F0TE)](https://codecov.io/gh/langroid/langroid)\n[![Lint](https://github.com/langroid/langroid/actions/workflows/validate.yml/badge.svg)](https://github.com/langroid/langroid/actions/workflows/validate.yml)\n[![Docs](https://github.com/langroid/langroid/actions/workflows/mkdocs-deploy.yml/badge.svg)](https://github.com/langroid/langroid/actions/workflows/mkdocs-deploy.yml)\n[![Static Badge](https://img.shields.io/badge/Documentation-blue?link=https%3A%2F%2Flangroid.github.io%2Flangroid%2F&link=https%3A%2F%2Flangroid.github.io%2Flangroid%2F)](https://langroid.github.io/langroid)\n[![Static Badge](https://img.shields.io/badge/Discord-orange?logoColor=orange&link=https%3A%2F%2Fdiscord.gg%2FZU36McDgDs)](https://discord.gg/ZU36McDgDs)\n[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb)\n\n[![Share on Twitter](https://img.shields.io/twitter/url?style=social&url=https://github.com/langroid/langroid)](https://twitter.com/intent/tweet?text=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.%20https://github.com/langroid/langroid)\n[![Share on LinkedIn](https://img.shields.io/badge/Share%20on-LinkedIn-blue)](https://www.linkedin.com/shareArticle?mini=true&url=https://github.com/langroid/langroid&title=Langroid:%20A%20Powerful,%20Elegant%20Framework&summary=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.)\n\n</div>\n\n<h3 align="center">\n  <a target="_blank" \n    href="https://langroid.github.io/langroid/" rel="dofollow">\n      <strong>Documentation</strong></a>\n  &middot;\n  <a target="_blank" href="https://github.com/langroid/langroid-examples" rel="dofollow">\n      <strong>Examples Repo</strong></a>\n  &middot;\n  <a target="_blank" href="https://discord.gg/ZU36McDgDs" rel="dofollow">\n      <strong>Discord</strong></a>\n  &middot;\n  <a target="_blank" href="./CONTRIBUTING.md" rel="dofollow">\n      <strong>Contributing</strong></a>\n\n  <br />\n</h3>\n\n`Langroid` is an intuitive, lightweight, extensible and principled\nPython framework to easily build LLM-powered applications. \nYou set up Agents, equip them with optional components (LLM, \nvector-store and methods), assign them tasks, and have them \ncollaboratively solve a problem by exchanging messages. \nThis Multi-Agent paradigm is inspired by the\n[Actor Framework](https://en.wikipedia.org/wiki/Actor_model)\n(but you do not need to know anything about this!).\nWe welcome contributions -- See the [contributions](./CONTRIBUTING.md) document\nfor ideas on what to contribute.\n\n**Questions, Feedback, Ideas? Join us on [Discord](https://discord.gg/ZU36McDgDs)!**\n\n<details>\n<summary> <b>:fire: Updates/Releases</b></summary>\n\n- **Aug 2023:**\n  - **Example:** [Answer questions](examples/docqa/chat-search.py) using Google Search + vecdb-retrieval from URL contents. \n  - **0.1.39:** [`GoogleSearchTool`](langroid/agent/stateless_tools/google_search_tool.py) to enable Agents (their LLM) to do Google searches via function-calling/tools.\n    See [this chat example](examples/basic/chat-search.py) for how easy it is to add this tool to an agent.\n  - **Colab notebook** to try the quick-start examples: [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb) \n  - **0.1.37:** Added [`SQLChatAgent`](langroid/agent/special/sql_chat_agent.py) -- thanks to our latest contributor [Rithwik Babu](https://github.com/rithwikbabu)!\n  - Multi-agent Example: [Autocorrect chat](examples/basic/autocorrect.py)\n- **July 2023:** \n  - **0.1.30:** Added [`TableChatAgent`](langroid/agent/special/table_chat_agent.py) to \n    [chat](examples/data-qa/table_chat.py) with tabular datasets (dataframes, files, URLs): LLM generates Pandas code,\n    and code is executed using Langroid\'s tool/function-call mechanism. \n  - **Demo:** 3-agent system for Audience [Targeting](https://langroid.github.io/langroid/demos/targeting/audience-targeting/).\n  - **0.1.27**: Added [support](langroid/cachedb/momento_cachedb.py) \n    for [Momento Serverless Cache](https://www.gomomento.com/) as an alternative to Redis.\n  - **0.1.24**: [`DocChatAgent`](langroid/agent/special/doc_chat_agent.py) \n    now [accepts](langroid/parsing/pdf_parser.py) PDF files or URLs.\n\n</details>\n\n# :rocket: Demo\nSuppose you want to extract structured information about the key terms \nof a commercial lease document. You can easily do this with Langroid using a two-agent system,\nas we show in the [langroid-examples](https://github.com/langroid/langroid-examples/blob/main/examples/docqa/chat_multi_extract.py) repo.\nThe demo showcases just a few of the many features of Langroid, such as:\n- Multi-agent collaboration: `LeaseExtractor` is in charge of the task, and its LLM (GPT4) generates questions \nto be answered by the `DocAgent`.\n- Retrieval augmented question-answering, with **source-citation**: `DocAgent` LLM (GPT4) uses retrieval from a vector-store to \nanswer the `LeaseExtractor`\'s questions, cites the specific excerpt supporting the answer. \n- Function-calling (also known as tool/plugin): When it has all the information it \nneeds, the `LeaseExtractor` LLM presents the information in a structured \nformat using a Function-call. \n\nHere is what it looks like in action:\n\n![Demo](docs/assets/demos/lease-extractor-demo.gif)\n\n\n# :zap: Highlights\n\n- **Agents as first-class citizens:** The [Agent](https://langroid.github.io/langroid/reference/agent/base/#langroid.agent.base.Agent) class encapsulates LLM conversation state,\n  and optionally a vector-store and tools. Agents are a core abstraction in Langroid;\n  Agents act as _message transformers_, and by default provide 3 _responder_ methods, one corresponding to each entity: LLM, Agent, User.\n- **Tasks:** A [Task](https://langroid.github.io/langroid/reference/agent/task/) class wraps an Agent, and gives the agent instructions (or roles, or goals), \n  manages iteration over an Agent\'s responder methods, \n  and orchestrates multi-agent interactions via hierarchical, recursive\n  task-delegation. The `Task.run()` method has the same \n  type-signature as an Agent\'s responder\'s methods, and this is key to how \n  a task of an agent can delegate to other sub-tasks: from the point of view of a Task,\n  sub-tasks are simply additional responders, to be used in a round-robin fashion \n  after the agent\'s own responders.\n- **Modularity, Reusabilily, Loose coupling:** The `Agent` and `Task` abstractions allow users to design\n  Agents with specific skills, wrap them in Tasks, and combine tasks in a flexible way.\n- **LLM Support**: Langroid supports OpenAI LLMs including GPT-3.5-Turbo,\n  GPT-4-0613\n- **Caching of LLM responses:** Langroid supports [Redis](https://redis.com/try-free/) and \n  [Momento](https://www.gomomento.com/) to cache LLM responses.\n- **Vector-stores**: [Qdrant](https://qdrant.tech/) and [Chroma](https://www.trychroma.com/) are currently supported.\n  Vector stores allow for Retrieval-Augmented-Generation (RAG).\n- **Grounding and source-citation:** Access to external documents via vector-stores \n   allows for grounding and source-citation.\n- **Observability, Logging, Lineage:** Langroid generates detailed logs of multi-agent interactions and\n  maintains provenance/lineage of messages, so that you can trace back\n  the origin of a message.\n- **Tools/Plugins/Function-calling**: Langroid supports OpenAI\'s recently\n  released [function calling](https://platform.openai.com/docs/guides/gpt/function-calling)\n  feature. In addition, Langroid has its own native equivalent, which we\n  call **tools** (also known as "plugins" in other contexts). Function\n  calling and tools have the same developer-facing interface, implemented\n  using [Pydantic](https://docs.pydantic.dev/latest/),\n  which makes it very easy to define tools/functions and enable agents\n  to use them. Benefits of using Pydantic are that you never have to write\n  complex JSON specs for function calling, and when the LLM\n  hallucinates malformed JSON, the Pydantic error message is sent back to\n  the LLM so it can fix it!\n\n--- \n\n# :gear: Installation and Setup\n\n### Install `langroid`\nLangroid requires Python 3.11+. We recommend using a virtual environment.\nUse `pip` to install `langroid` (from PyPi) to your virtual environment:\n```bash\npip install langroid\n```\nThe core Langroid package lets you use OpenAI Embeddings models via their API. \nIf you instead want to use the `all-MiniLM-L6-v2` embeddings model\nfrom from HuggingFace, install Langroid like this:\n```bash\npip install langroid[hf-embeddings]\n```\nNote that this will install `torch` and `sentence-transformers` libraries.\n\n### Set up environment variables (API keys, etc)\n\nTo get started, all you need is an OpenAI API Key.\nIf you don\'t have one, see [this OpenAI Page](https://help.openai.com/en/collections/3675940-getting-started-with-openai-api).\nCurrently only OpenAI models are supported. Others will be added later\n(Pull Requests welcome!).\n\nIn the root of the repo, copy the `.env-template` file to a new file `.env`: \n```bash\ncp .env-template .env\n```\nThen insert your OpenAI API Key. \nYour `.env` file should look like this:\n```bash\nOPENAI_API_KEY=your-key-here-without-quotes\n````\n\nAlternatively, you can set this as an environment variable in your shell\n(you will need to do this every time you open a new shell):\n```bash\nexport OPENAI_API_KEY=your-key-here-without-quotes\n```\n\n\n<details>\n<summary><b>Optional Setup Instructions (click to expand) </b></summary>\n\nAll of the following environment variable settings are optional, and some are only needed \nto use specific features (as noted below).\n\n- **Qdrant** Vector Store API Key, URL. This is only required if you want to use Qdrant cloud.\n  You can sign up for a free 1GB account at [Qdrant cloud](https://cloud.qdrant.io).\n  If you skip setting up these, Langroid will use Qdrant in local-storage mode.\n  Alternatively [Chroma](https://docs.trychroma.com/) is also currently supported. \n  We use the local-storage version of Chroma, so there is no need for an API key.\n  Langroid uses Qdrant by default.\n- **Redis** Password, host, port: This is optional, and only needed to cache LLM API responses\n  using Redis Cloud. Redis [offers](https://redis.com/try-free/) a free 30MB Redis account\n  which is more than sufficient to try out Langroid and even beyond.\n  If you don\'t set up these, Langroid will use a pure-python \n  Redis in-memory cache via the [Fakeredis](https://fakeredis.readthedocs.io/en/latest/) library.\n- **Momento** Serverless Caching of LLM API responses (as an alternative to Redis). \n   To use Momento instead of Redis:\n  - enter your Momento Token in the `.env` file, as the value of `MOMENTO_AUTH_TOKEN` (see example file below),\n  - in the `.env` file set `CACHE_TYPE=momento` (instead of `CACHE_TYPE=redis` which is the default).\n- **GitHub** Personal Access Token (required for apps that need to analyze git\n  repos; token-based API calls are less rate-limited). See this\n  [GitHub page](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens).\n- **Google Custom Search API Credentials:** Only needed to enable an Agent to use the `GoogleSearchTool`.\n  To use Google Search as an LLM Tool/Plugin/function-call, \n  you\'ll need to set up \n  [a Google API key](https://developers.google.com/custom-search/v1/introduction#identify_your_application_to_google_with_api_key),\n  then [setup a Google Custom Search Engine (CSE) and get the CSE ID](https://developers.google.com/custom-search/docs/tutorial/creatingcse).\n  (Documentation for these can be challenging, we suggest asking GPT4 for a step-by-step guide.)\n  After obtaining these credentials, store them as values of \n  `GOOGLE_API_KEY` and `GOOGLE_CSE_ID` in your `.env` file. \n  Full documentation on using this (and other such "stateless" tools) is coming soon, but \n  in the meantime take a peek at this [chat example](examples/basic/chat-search.py), which \n  shows how you can easily equip an Agent with a `GoogleSearchtool`.\n  \n\n\nIf you add all of these optional variables, your `.env` file should look like this:\n```bash\nOPENAI_API_KEY=your-key-here-without-quotes\nGITHUB_ACCESS_TOKEN=your-personal-access-token-no-quotes\nCACHE_TYPE=redis # or momento\nREDIS_PASSWORD=your-redis-password-no-quotes\nREDIS_HOST=your-redis-hostname-no-quotes\nREDIS_PORT=your-redis-port-no-quotes\nMOMENTO_AUTH_TOKEN=your-momento-token-no-quotes # instead of REDIS* variables\nQDRANT_API_KEY=your-key\nQDRANT_API_URL=https://your.url.here:6333 # note port number must be included\nGOOGLE_API_KEY=your-key\nGOOGLE_CSE_ID=your-cse-id\n```\n</details>\n\n---\n\n# :tada: Usage Examples\n\nThese are quick teasers to give a glimpse of what you can do with Langroid\nand how your code would look. \n\n:warning: The code snippets below are intended to give a flavor of the code\nand they are **not** complete runnable examples! For that we encourage you to \nconsult the [`langroid-examples`](https://github.com/langroid/langroid-examples) \nrepository.\n\n:information_source: The various LLM prompts and instructions in Langroid\nhave been tested to work well with GPT4.\nSwitching to GPT3.5-Turbo is easy via a config flag\n(e.g., `cfg = OpenAIGPTConfig(chat_model=OpenAIChatModel.GPT3_5_TURBO)`),\nand may suffice for some applications, but in general you may see inferior results.\n\n\n:book: Also see the\n[`Getting Started Guide`](https://langroid.github.io/langroid/quick-start/)\nfor a detailed tutorial.\n\nClick to expand any of the code examples below.\nAll of these can be run in a Colab notebook:\n[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb)\n\n<details>\n<summary> <b> Direct interaction with OpenAI LLM </b> </summary>\n\n```python\nfrom langroid.language_models.openai_gpt import ( \n        OpenAIGPTConfig, OpenAIChatModel, OpenAIGPT,\n)\nfrom langroid.language_models.base import LLMMessage, Role\n\ncfg = OpenAIGPTConfig(chat_model=OpenAIChatModel.GPT4)\n\nmdl = OpenAIGPT(cfg)\n\nmessages = [\n  LLMMessage(content="You are a helpful assistant",  role=Role.SYSTEM), \n  LLMMessage(content="What is the capital of Ontario?",  role=Role.USER),\n]\nresponse = mdl.chat(messages, max_tokens=200)\nprint(response.message)\n```\n</details>\n\n<details>\n<summary> <b> Define an agent, set up a task, and run it </b> </summary>\n\n```python\nfrom langroid.agent.chat_agent import ChatAgent, ChatAgentConfig\nfrom langroid.agent.task import Task\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\n\nconfig = ChatAgentConfig(\n    llm = OpenAIGPTConfig(\n        chat_model=OpenAIChatModel.GPT4,\n    ),\n    vecdb=None, # no vector store\n)\nagent = ChatAgent(config)\n# get response from agent\'s LLM, and put this in an interactive loop...\n# answer = agent.llm_response("What is the capital of Ontario?")\n  # ... OR instead, set up a task (which has a built-in loop) and run it\ntask = Task(agent, name="Bot") \ntask.run() # ... a loop seeking response from LLM or User at each turn\n```\n</details>\n\n<details>\n<summary><b> Three communicating agents </b></summary>\n\nA toy numbers game, where when given a number `n`:\n- `repeater_agent`\'s LLM simply returns `n`,\n- `even_agent`\'s LLM returns `n/2` if `n` is even, else says "DO-NOT-KNOW"\n- `odd_agent`\'s LLM returns `3*n+1` if `n` is odd, else says "DO-NOT-KNOW"\n\nFirst define the 3 agents, and set up their tasks with instructions:\n\n```python\nfrom langroid.utils.constants import NO_ANSWER\nfrom langroid.agent.chat_agent import ChatAgent, ChatAgentConfig\nfrom langroid.agent.task import Task\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\nconfig = ChatAgentConfig(\n    llm = OpenAIGPTConfig(\n        chat_model=OpenAIChatModel.GPT4,\n    ),\n    vecdb = None,\n)\nrepeater_agent = ChatAgent(config)\nrepeater_task = Task(\n    repeater_agent,\n    name = "Repeater",\n    system_message="""\n    Your job is to repeat whatever number you receive.\n    """,\n    llm_delegate=True, # LLM takes charge of task\n    single_round=False, \n)\neven_agent = ChatAgent(config)\neven_task = Task(\n    even_agent,\n    name = "EvenHandler",\n    system_message=f"""\n    You will be given a number. \n    If it is even, divide by 2 and say the result, nothing else.\n    If it is odd, say {NO_ANSWER}\n    """,\n    single_round=True,  # task done after 1 step() with valid response\n)\n\nodd_agent = ChatAgent(config)\nodd_task = Task(\n    odd_agent,\n    name = "OddHandler",\n    system_message=f"""\n    You will be given a number n. \n    If it is odd, return (n*3+1), say nothing else. \n    If it is even, say {NO_ANSWER}\n    """,\n    single_round=True,  # task done after 1 step() with valid response\n)\n```\nThen add the `even_task` and `odd_task` as sub-tasks of `repeater_task`, \nand run the `repeater_task`, kicking it off with a number as input:\n```python\nrepeater_task.add_sub_task([even_task, odd_task])\nrepeater_task.run("3")\n```\n\n</details>\n\n<details>\n<summary><b> Simple Tool/Function-calling example </b></summary>\n\nLangroid leverages Pydantic to support OpenAI\'s\n[Function-calling API](https://platform.openai.com/docs/guides/gpt/function-calling)\nas well as its own native tools. The benefits are that you don\'t have to write\nany JSON to specify the schema, and also if the LLM hallucinates a malformed\ntool syntax, Langroid sends the Pydantic validation error (suitiably sanitized) \nto the LLM so it can fix it!\n\nSimple example: Say the agent has a secret list of numbers, \nand we want the LLM to find the smallest number in the list. \nWe want to give the LLM a `probe` tool/function which takes a\nsingle number `n` as argument. The tool handler method in the agent\nreturns how many numbers in its list are at most `n`.\n\nFirst define the tool using Langroid\'s `ToolMessage` class:\n\n\n```python\nfrom langroid.agent.tool_message import ToolMessage\nclass ProbeTool(ToolMessage):\n  request: str = "probe" # specifies which agent method handles this tool\n  purpose: str = """\n        To find how many numbers in my list are less than or equal to  \n        the <number> you specify.\n        """ # description used to instruct the LLM on when/how to use the tool\n  number: int  # required argument to the tool\n```\n\nThen define a `SpyGameAgent` as a subclass of `ChatAgent`, \nwith a method `probe` that handles this tool:\n\n```python\nfrom langroid.agent.chat_agent import ChatAgent, ChatAgentConfig\nclass SpyGameAgent(ChatAgent):\n  def __init__(self, config: ChatAgentConfig):\n    super().__init__(config)\n    self.numbers = [3, 4, 8, 11, 15, 25, 40, 80, 90]\n\n  def probe(self, msg: ProbeTool) -> str:\n    # return how many numbers in self.numbers are less or equal to msg.number\n    return str(len([n for n in self.numbers if n <= msg.number]))\n```\n\nWe then instantiate the agent and enable it to use and respond to the tool:\n\n```python\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\nspy_game_agent = SpyGameAgent(\n    ChatAgentConfig(\n        name="Spy",\n        llm = OpenAIGPTConfig(\n            chat_model=OpenAIChatModel.GPT4,\n        ),\n        vecdb=None,\n        use_tools=False, #  don\'t use Langroid native tool\n        use_functions_api=True, # use OpenAI function-call API\n    )\n)\nspy_game_agent.enable_message(ProbeTool)\n```\n\nFor a full working example see the\n[chat-agent-tool.py](https://github.com/langroid/langroid-examples/blob/main/examples/quick-start/chat-agent-tool.py)\nscript in the `langroid-examples` repo.\n</details>\n\n<details>\n<summary> <b>Tool/Function-calling to extract structured information from text </b> </summary>\n\nSuppose you want an agent to extract \nthe key terms of a lease, from a lease document, as a nested JSON structure.\nFirst define the desired structure via Pydantic models:\n\n```python\nfrom pydantic import BaseModel\nclass LeasePeriod(BaseModel):\n    start_date: str\n    end_date: str\n\n\nclass LeaseFinancials(BaseModel):\n    monthly_rent: str\n    deposit: str\n\nclass Lease(BaseModel):\n    period: LeasePeriod\n    financials: LeaseFinancials\n    address: str\n```\n\nThen define the `LeaseMessage` tool as a subclass of Langroid\'s `ToolMessage`.\nNote the tool has a required argument `terms` of type `Lease`:\n\n```python\nclass LeaseMessage(ToolMessage):\n    request: str = "lease_info"\n    purpose: str = """\n        Collect information about a Commercial Lease.\n        """\n    terms: Lease\n```\n\nThen define a `LeaseExtractorAgent` with a method `lease_info` that handles this tool,\ninstantiate the agent, and enable it to use and respond to this tool:\n\n```python\nclass LeaseExtractorAgent(ChatAgent):\n    def lease_info(self, message: LeaseMessage) -> str:\n        print(\n            f"""\n        DONE! Successfully extracted Lease Info:\n        {message.terms}\n        """\n        )\n        return json.dumps(message.terms.dict())\n    \nlease_extractor_agent = LeaseExtractorAgent(\n  ChatAgentConfig(\n    llm=OpenAIGPTConfig(),\n    use_functions_api=False,\n    use_tools=True,\n  )\n)\nlease_extractor_agent.enable_message(LeaseMessage)\n```\n\nSee the [`chat_multi_extract.py`](https://github.com/langroid/langroid-examples/blob/main/examples/docqa/chat_multi_extract.py)\nscript in the `langroid-examples` repo for a full working example.\n</details>\n\n<details>\n<summary><b> Chat with documents (file paths, URLs, etc) </b></summary>\n\nLangroid provides a specialized agent class `DocChatAgent` for this purpose.\nIt incorporates document sharding, embedding, storage in a vector-DB, \nand retrieval-augmented query-answer generation.\nUsing this class to chat with a collection of documents is easy.\nFirst create a `DocChatAgentConfig` instance, with a \n`doc_paths` field that specifies the documents to chat with.\n\n```python\nfrom langroid.agent.doc_chat_agent import DocChatAgentConfig\nconfig = DocChatAgentConfig(\n  doc_paths = [\n    "https://en.wikipedia.org/wiki/Language_model",\n    "https://en.wikipedia.org/wiki/N-gram_language_model",\n    "/path/to/my/notes-on-language-models.txt",\n  ]\n  llm = OpenAIGPTConfig(\n    chat_model=OpenAIChatModel.GPT4,\n  ),\n  vecdb=VectorStoreConfig(\n    type="qdrant",\n  ),\n)\n```\n\nThen instantiate the `DocChatAgent` (this ingests the docs into the vector-store):\n\n```python\nagent = DocChatAgent(config)\n```\nThen we can either ask the agent one-off questions,\n```python\nagent.chat("What is a language model?")\n```\nor wrap it in a `Task` and run an interactive loop with the user:\n```python\nfrom langroid.task import Task\ntask = Task(agent)\ntask.run()\n```\n\nSee full working scripts in the \n[`docqa`](https://github.com/langroid/langroid-examples/tree/main/examples/docqa)\nfolder of the `langroid-examples` repo.\n</details>\n\n<details>\n<summary><b> :fire: Chat with tabular data (file paths, URLs, dataframes) </b></summary>\n\nUsing Langroid you can set up a `TableChatAgent` with a dataset (file path, URL or dataframe),\nand query it. The Agent\'s LLM generates Pandas code to answer the query, \nvia function-calling (or tool/plugin), and the Agent\'s function-handling method\nexecutes the code and returns the answer.\n\nHere is how you can do this:\n\n```python\nfrom langroid.agent.special.table_chat_agent import TableChatAgent, TableChatAgentConfig\nfrom langroid.agent.task import Task\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\n```\n\nSet up a `TableChatAgent` for a data file, URL or dataframe\n(Ensure the data table has a header row; the delimiter/separator is auto-detected):\n```python\ndataset =  "https://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-red.csv"\n# or dataset = "/path/to/my/data.csv"\n# or dataset = pd.read_csv("/path/to/my/data.csv")\nagent = TableChatAgent(\n    config=TableChatAgentConfig(\n        data=dataset,  \n        llm=OpenAIGPTConfig(\n            chat_model=OpenAIChatModel.GPT4,\n        ),\n    )\n)\n```\nSet up a task, and ask one-off questions like this: \n\n```python\ntask = Task(\n  agent, \n  name = "DataAssistant",\n  default_human_response="", # to avoid waiting for user input\n)\nresult = task.run(\n  "What is the average alcohol content of wines with a quality rating above 7?",\n  turns=2 # return after user question, LLM fun-call/tool response, Agent code-exec result\n) \nprint(result.content)\n```\nOr alternatively, set up a task and run it in an interactive loop with the user:\n\n```python\ntask = Task(agent, name="DataAssistant")\ntask.run()\n``` \n\nFor a full working example see the \n[`table_chat.py`](https://github.com/langroid/langroid-examples/tree/main/examples/data-qa/table_chat.py)\nscript in the `langroid-examples` repo.\n\n\n</details>\n\n---\n\n# :heart: Thank you to our [supporters](https://github.com/langroid/langroid/stargazers)\n\nIf you like this project, please give it a star ⭐ and 📢 spread the word in your network:\n\n[![Share on Twitter](https://img.shields.io/twitter/url?style=social&url=https://github.com/langroid/langroid)](https://twitter.com/intent/tweet?text=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.%20https://github.com/langroid/langroid)\n[![Share on LinkedIn](https://img.shields.io/badge/Share%20on-LinkedIn-blue)](https://www.linkedin.com/shareArticle?mini=true&url=https://github.com/langroid/langroid&title=Langroid:%20A%20Powerful,%20Elegant%20Framework&summary=Langroid%20is%20a%20powerful,%20elegant%20new%20framework%20to%20easily%20build%20%23LLM%20applications.%20You%20set%20up%20LLM-powered%20Agents%20with%20vector-stores,%20assign%20tasks,%20and%20have%20them%20collaboratively%20solve%20problems%20via%20message-transformations.)\n\n\n\n\nYour support will help build Langroid\'s momentum and community.\n\n\n\n\n# Langroid Co-Founders\n\n- [Prasad Chalasani](https://www.linkedin.com/in/pchalasani/) (IIT BTech/CS, CMU PhD/ML; Independent ML Consultant)\n- [Somesh Jha](https://www.linkedin.com/in/somesh-jha-80208015/) (IIT BTech/CS, CMU PhD/CS; Professor of CS, U Wisc at Madison)\n\n\n\n',
+    'author': 'Prasad Chalasani',
+    'author_email': 'pchalasani@gmail.com',
+    'maintainer': 'None',
+    'maintainer_email': 'None',
+    'url': 'None',
+    'packages': packages,
+    'package_data': package_data,
+    'install_requires': install_requires,
+    'extras_require': extras_require,
+    'python_requires': '>=3.8.1,<3.12',
+}
+
+
+setup(**setup_kwargs)

langroid-0.1.39/setup.py

@@ -1,93 +0,0 @@
-# -*- coding: utf-8 -*-
-from setuptools import setup
-
-packages = \
-['langroid',
- 'langroid.agent',
- 'langroid.agent.special',
- 'langroid.agent.stateless_tools',
- 'langroid.cachedb',
- 'langroid.embedding_models',
- 'langroid.language_models',
- 'langroid.parsing',
- 'langroid.prompts',
- 'langroid.scripts',
- 'langroid.utils',
- 'langroid.utils.llms',
- 'langroid.utils.output',
- 'langroid.utils.web',
- 'langroid.vector_store']
-
-package_data = \
-{'': ['*']}
-
-install_requires = \
-['autopep8>=2.0.2,<3.0.0',
- 'black[jupyter]>=23.3.0,<24.0.0',
- 'bs4>=0.0.1,<0.0.2',
- 'chromadb>=0.3.21,<0.4.0',
- 'colorlog>=6.7.0,<7.0.0',
- 'faker>=18.9.0,<19.0.0',
- 'fakeredis>=2.12.1,<3.0.0',
- 'fire>=0.5.0,<0.6.0',
- 'flake8>=6.0.0,<7.0.0',
- 'google-api-python-client>=2.95.0,<3.0.0',
- 'halo>=0.0.31,<0.0.32',
- 'lxml>=4.9.3,<5.0.0',
- 'mkdocs-awesome-pages-plugin>=2.8.0,<3.0.0',
- 'mkdocs-gen-files>=0.4.0,<0.5.0',
- 'mkdocs-jupyter>=0.24.1,<0.25.0',
- 'mkdocs-literate-nav>=0.6.0,<0.7.0',
- 'mkdocs-material>=9.1.5,<10.0.0',
- 'mkdocs-section-index>=0.3.5,<0.4.0',
- 'mkdocs>=1.4.2,<2.0.0',
- 'mkdocstrings[python]>=0.21.2,<0.22.0',
- 'momento>=1.7.0,<2.0.0',
- 'mypy>=1.2.0,<2.0.0',
- 'nltk>=3.8.1,<4.0.0',
- 'openai>=0.27.5,<0.28.0',
- 'pandas>=2.0.3,<3.0.0',
- 'pre-commit>=3.3.2,<4.0.0',
- 'prettytable>=3.8.0,<4.0.0',
- 'pydantic==1.10.11',
- 'pygithub>=1.58.1,<2.0.0',
- 'pygments>=2.15.1,<3.0.0',
- 'pyparsing>=3.0.9,<4.0.0',
- 'pypdf>=3.12.2,<4.0.0',
- 'python-dotenv>=1.0.0,<2.0.0',
- 'qdrant-client>=1.3.1,<2.0.0',
- 'redis>=4.5.5,<5.0.0',
- 'requests-oauthlib>=1.3.1,<2.0.0',
- 'requests>=2.31.0,<3.0.0',
- 'rich>=13.3.4,<14.0.0',
- 'ruff>=0.0.270,<0.0.271',
- 'sqlalchemy>=2.0.19,<3.0.0',
- 'tiktoken>=0.3.3,<0.4.0',
- 'trafilatura>=1.5.0,<2.0.0',
- 'typer>=0.7.0,<0.8.0',
- 'types-redis>=4.5.5.2,<5.0.0.0',
- 'types-requests>=2.31.0.1,<3.0.0.0',
- 'wget>=3.2,<4.0']
-
-extras_require = \
-{'hf-embeddings': ['sentence-transformers==2.2.2', 'torch==2.0.0']}
-
-setup_kwargs = {
-    'name': 'langroid',
-    'version': '0.1.39',
-    'description': 'Harness LLMs with Multi-Agent Programming',
-    'long_description': '<div align="center">\n  <img src="docs/assets/langroid-card-lambda-ossem-rust-1200-630.png" alt="Logo" \n        width="400" align="center">\n</div>\n\n<div align="center">\n\n[![Pytest](https://github.com/langroid/langroid/actions/workflows/pytest.yml/badge.svg)](https://github.com/langroid/langroid/actions/workflows/pytest.yml)\n[![codecov](https://codecov.io/gh/langroid/langroid/branch/main/graph/badge.svg?token=H94BX5F0TE)](https://codecov.io/gh/langroid/langroid)\n[![Lint](https://github.com/langroid/langroid/actions/workflows/validate.yml/badge.svg)](https://github.com/langroid/langroid/actions/workflows/validate.yml)\n[![Docs](https://github.com/langroid/langroid/actions/workflows/mkdocs-deploy.yml/badge.svg)](https://github.com/langroid/langroid/actions/workflows/mkdocs-deploy.yml)\n[![Static Badge](https://img.shields.io/badge/Documentation-blue?link=https%3A%2F%2Flangroid.github.io%2Flangroid%2F&link=https%3A%2F%2Flangroid.github.io%2Flangroid%2F)](https://langroid.github.io/langroid)\n[![Static Badge](https://img.shields.io/badge/Discord-orange?logoColor=orange&link=https%3A%2F%2Fdiscord.gg%2FZU36McDgDs)](https://discord.gg/ZU36McDgDs)\n[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb)\n\n\n</div>\n\n<h3 align="center">\n  <a target="_blank" \n    href="https://langroid.github.io/langroid/" rel="dofollow">\n      <strong>Documentation</strong></a>\n  &middot;\n  <a target="_blank" href="https://github.com/langroid/langroid-examples" rel="dofollow">\n      <strong>Examples Repo</strong></a>\n  &middot;\n  <a target="_blank" href="https://discord.gg/ZU36McDgDs" rel="dofollow">\n      <strong>Discord</strong></a>\n  &middot;\n  <a target="_blank" href="./CONTRIBUTING.md" rel="dofollow">\n      <strong>Contributing</strong></a>\n\n  <br />\n</h3>\n\n`Langroid` is an intuitive, lightweight, extensible and principled\nPython framework to easily build LLM-powered applications. \nYou set up Agents, equip them with optional components (LLM, \nvector-store and methods), assign them tasks, and have them \ncollaboratively solve a problem by exchanging messages. \nThis Multi-Agent paradigm is inspired by the\n[Actor Framework](https://en.wikipedia.org/wiki/Actor_model)\n(but you do not need to know anything about this!).\nWe welcome contributions -- See the [contributions](./CONTRIBUTING.md) document\nfor ideas on what to contribute.\n\n**Questions, Feedback, Ideas? Join us on [Discord](https://discord.gg/ZU36McDgDs)!**\n\n<details>\n<summary> <b>:fire: Updates/Releases</b></summary>\n\n- **Aug 2023:**\n  - **0.1.38:** [`GoogleSearchTool`](langroid/agent/stateless_tools/google_search_tool.py) to enable Agents (their LLM) to do Google searches via function-calling/tools.\n    See [this chat example](examples/basic/chat-search.py) for how easy it is to add this tool to an agent.\n  - **Colab notebook** to try the quick-start examples: [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb) \n  - **0.1.37:** Added [`SQLChatAgent`](langroid/agent/special/sql_chat_agent.py) -- thanks to our latest contributor [Rithwik Babu](https://github.com/rithwikbabu)!\n  - Multi-agent Example: [Autocorrect chat](examples/basic/autocorrect.py)\n- **July 2023:** \n  - **0.1.30:** Added [`TableChatAgent`](langroid/agent/special/table_chat_agent.py) to \n    [chat](examples/data-qa/table_chat.py) with tabular datasets (dataframes, files, URLs): LLM generates Pandas code,\n    and code is executed using Langroid\'s tool/function-call mechanism. \n  - **Demo:** 3-agent system for Audience [Targeting](https://langroid.github.io/langroid/demos/targeting/audience-targeting/).\n  - **0.1.27**: Added [support](langroid/cachedb/momento_cachedb.py) \n    for [Momento Serverless Cache](https://www.gomomento.com/) as an alternative to Redis.\n  - **0.1.24**: [`DocChatAgent`](langroid/agent/special/doc_chat_agent.py) \n    now [accepts](langroid/parsing/pdf_parser.py) PDF files or URLs.\n\n</details>\n\n# :rocket: Demo\nSuppose you want to extract structured information about the key terms \nof a commercial lease document. You can easily do this with Langroid using a two-agent system,\nas we show in the [langroid-examples](https://github.com/langroid/langroid-examples/blob/main/examples/docqa/chat_multi_extract.py) repo.\nThe demo showcases just a few of the many features of Langroid, such as:\n- Multi-agent collaboration: `LeaseExtractor` is in charge of the task, and its LLM (GPT4) generates questions \nto be answered by the `DocAgent`.\n- Retrieval augmented question-answering, with **source-citation**: `DocAgent` LLM (GPT4) uses retrieval from a vector-store to \nanswer the `LeaseExtractor`\'s questions, cites the specific excerpt supporting the answer. \n- Function-calling (also known as tool/plugin): When it has all the information it \nneeds, the `LeaseExtractor` LLM presents the information in a structured \nformat using a Function-call. \n\nHere is what it looks like in action:\n\n![Demo](docs/assets/demos/lease-extractor-demo.gif)\n\n\n# :zap: Highlights\n\n- **Agents as first-class citizens:** The [Agent](https://langroid.github.io/langroid/reference/agent/base/#langroid.agent.base.Agent) class encapsulates LLM conversation state,\n  and optionally a vector-store and tools. Agents are a core abstraction in Langroid;\n  Agents act as _message transformers_, and by default provide 3 _responder_ methods, one corresponding to each entity: LLM, Agent, User.\n- **Tasks:** A [Task](https://langroid.github.io/langroid/reference/agent/task/) class wraps an Agent, and gives the agent instructions (or roles, or goals), \n  manages iteration over an Agent\'s responder methods, \n  and orchestrates multi-agent interactions via hierarchical, recursive\n  task-delegation. The `Task.run()` method has the same \n  type-signature as an Agent\'s responder\'s methods, and this is key to how \n  a task of an agent can delegate to other sub-tasks: from the point of view of a Task,\n  sub-tasks are simply additional responders, to be used in a round-robin fashion \n  after the agent\'s own responders.\n- **Modularity, Reusabilily, Loose coupling:** The `Agent` and `Task` abstractions allow users to design\n  Agents with specific skills, wrap them in Tasks, and combine tasks in a flexible way.\n- **LLM Support**: Langroid supports OpenAI LLMs including GPT-3.5-Turbo,\n  GPT-4-0613\n- **Caching of LLM responses:** Langroid supports [Redis](https://redis.com/try-free/) and \n  [Momento](https://www.gomomento.com/) to cache LLM responses.\n- **Vector-stores**: [Qdrant](https://qdrant.tech/) and [Chroma](https://www.trychroma.com/) are currently supported.\n  Vector stores allow for Retrieval-Augmented-Generation (RAG).\n- **Grounding and source-citation:** Access to external documents via vector-stores \n   allows for grounding and source-citation.\n- **Observability, Logging, Lineage:** Langroid generates detailed logs of multi-agent interactions and\n  maintains provenance/lineage of messages, so that you can trace back\n  the origin of a message.\n- **Tools/Plugins/Function-calling**: Langroid supports OpenAI\'s recently\n  released [function calling](https://platform.openai.com/docs/guides/gpt/function-calling)\n  feature. In addition, Langroid has its own native equivalent, which we\n  call **tools** (also known as "plugins" in other contexts). Function\n  calling and tools have the same developer-facing interface, implemented\n  using [Pydantic](https://docs.pydantic.dev/latest/),\n  which makes it very easy to define tools/functions and enable agents\n  to use them. Benefits of using Pydantic are that you never have to write\n  complex JSON specs for function calling, and when the LLM\n  hallucinates malformed JSON, the Pydantic error message is sent back to\n  the LLM so it can fix it!\n\n--- \n\n# :gear: Installation and Setup\n\n### Install `langroid`\nLangroid requires Python 3.11+. We recommend using a virtual environment.\nUse `pip` to install `langroid` (from PyPi) to your virtual environment:\n```bash\npip install langroid\n```\nThe core Langroid package lets you use OpenAI Embeddings models via their API. \nIf you instead want to use the `all-MiniLM-L6-v2` embeddings model\nfrom from HuggingFace, install Langroid like this:\n```bash\npip install langroid[hf-embeddings]\n```\nNote that this will install `torch` and `sentence-transformers` libraries.\n\n### Set up environment variables (API keys, etc)\n\nTo get started, all you need is an OpenAI API Key.\nIf you don\'t have one, see [this OpenAI Page](https://help.openai.com/en/collections/3675940-getting-started-with-openai-api).\nCurrently only OpenAI models are supported. Others will be added later\n(Pull Requests welcome!).\n\nIn the root of the repo, copy the `.env-template` file to a new file `.env`: \n```bash\ncp .env-template .env\n```\nThen insert your OpenAI API Key. \nYour `.env` file should look like this:\n```bash\nOPENAI_API_KEY=your-key-here-without-quotes\n````\n\nAlternatively, you can set this as an environment variable in your shell\n(you will need to do this every time you open a new shell):\n```bash\nexport OPENAI_API_KEY=your-key-here-without-quotes\n```\n\n\n<details>\n<summary><b>Optional Setup Instructions (click to expand) </b></summary>\n\nAll of the below are optional and not strictly needed to run any of the examples.\n\n- **Qdrant** Vector Store API Key, URL. This is only required if you want to use Qdrant cloud.\n  You can sign up for a free 1GB account at [Qdrant cloud](https://cloud.qdrant.io).\n  If you skip setting up these, Langroid will use Qdrant in local-storage mode.\n  Alternatively [Chroma](https://docs.trychroma.com/) is also currently supported. \n  We use the local-storage version of Chroma, so there is no need for an API key.\n  Langroid uses Qdrant by default.\n- **Redis** Password, host, port: This is optional, and only needed to cache LLM API responses\n  using Redis Cloud. Redis [offers](https://redis.com/try-free/) a free 30MB Redis account\n  which is more than sufficient to try out Langroid and even beyond.\n  If you don\'t set up these, Langroid will use a pure-python \n  Redis in-memory cache via the [Fakeredis](https://fakeredis.readthedocs.io/en/latest/) library.\n- **Momento** Serverless Caching of LLM API responses (as an alternative to Redis). \n   To use Momento instead of Redis:\n  - enter your Momento Token in the `.env` file, as the value of `MOMENTO_AUTH_TOKEN` (see example file below),\n  - in the `.env` file set `CACHE_TYPE=momento` (instead of `CACHE_TYPE=redis` which is the default).\n- **GitHub** Personal Access Token (required for apps that need to analyze git\n  repos; token-based API calls are less rate-limited). See this\n  [GitHub page](https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/managing-your-personal-access-tokens).\n- **Google Custom Search API Credentials:** Only needed to enable an Agent to use the `GoogleSearchTool`.\n  To use Google Search as an LLM Tool/Plugin/function-call, \n  you\'ll need to set up \n  [a Google API key](https://developers.google.com/custom-search/v1/introduction#identify_your_application_to_google_with_api_key),\n  then [setup a Google Custom Search Engine (CSE) and get the CSE ID](https://developers.google.com/custom-search/docs/tutorial/creatingcse).\n  (Documentation for these can be challenging, we suggest asking GPT4 for a step-by-step guide.)\n  After obtaining these credentials, store them as values of \n  `GOOGLE_API_KEY` and `GOOGLE_CSE_ID` in your `.env` file. \n  Full documentation on using this (and other such "stateless" tools) is coming soon, but \n  in the meantime take a peek at this [chat example](examples/basic/chat-search.py), which \n  shows how you can easily equip an Agent with a `GoogleSearchtool`.\n  \n\n\nIf you add all of these optional variables, your `.env` file should look like this:\n```bash\nOPENAI_API_KEY=your-key-here-without-quotes\nGITHUB_ACCESS_TOKEN=your-personal-access-token-no-quotes\nCACHE_TYPE=redis # or momento\nREDIS_PASSWORD=your-redis-password-no-quotes\nREDIS_HOST=your-redis-hostname-no-quotes\nREDIS_PORT=your-redis-port-no-quotes\nMOMENTO_AUTH_TOKEN=your-momento-token-no-quotes # instead of REDIS* variables\nQDRANT_API_KEY=your-key\nQDRANT_API_URL=https://your.url.here:6333 # note port number must be included\nGOOGLE_API_KEY=your-key\nGOOGLE_CSE_ID=your-cse-id\n```\n</details>\n\n---\n\n# :tada: Usage Examples\n\nThese are quick teasers to give a glimpse of what you can do with Langroid\nand how your code would look. \n\n:warning: The code snippets below are intended to give a flavor of the code\nand they are **not** complete runnable examples! For that we encourage you to \nconsult the [`langroid-examples`](https://github.com/langroid/langroid-examples) \nrepository.\n\n:information_source: The various LLM prompts and instructions in Langroid\nhave been tested to work well with GPT4.\nSwitching to GPT3.5-Turbo is easy via a config flag\n(e.g., `cfg = OpenAIGPTConfig(chat_model=OpenAIChatModel.GPT3_5_TURBO)`),\nand may suffice for some applications, but in general you may see inferior results.\n\n\n:book: Also see the\n[`Getting Started Guide`](https://langroid.github.io/langroid/quick-start/)\nfor a detailed tutorial.\n\nClick to expand any of the code examples below.\nAll of these can be run in a Colab notebook:\n[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/langroid/langroid/blob/main/examples/langroid_quick_examples.ipynb)\n\n<details>\n<summary> <b> Direct interaction with OpenAI LLM </b> </summary>\n\n```python\nfrom langroid.language_models.openai_gpt import ( \n        OpenAIGPTConfig, OpenAIChatModel, OpenAIGPT,\n)\nfrom langroid.language_models.base import LLMMessage, Role\n\ncfg = OpenAIGPTConfig(chat_model=OpenAIChatModel.GPT4)\n\nmdl = OpenAIGPT(cfg)\n\nmessages = [\n  LLMMessage(content="You are a helpful assistant",  role=Role.SYSTEM), \n  LLMMessage(content="What is the capital of Ontario?",  role=Role.USER),\n]\nresponse = mdl.chat(messages, max_tokens=200)\nprint(response.message)\n```\n</details>\n\n<details>\n<summary> <b> Define an agent, set up a task, and run it </b> </summary>\n\n```python\nfrom langroid.agent.chat_agent import ChatAgent, ChatAgentConfig\nfrom langroid.agent.task import Task\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\n\nconfig = ChatAgentConfig(\n    llm = OpenAIGPTConfig(\n        chat_model=OpenAIChatModel.GPT4,\n    ),\n    vecdb=None, # no vector store\n)\nagent = ChatAgent(config)\n# get response from agent\'s LLM, and put this in an interactive loop...\n# answer = agent.llm_response("What is the capital of Ontario?")\n  # ... OR instead, set up a task (which has a built-in loop) and run it\ntask = Task(agent, name="Bot") \ntask.run() # ... a loop seeking response from LLM or User at each turn\n```\n</details>\n\n<details>\n<summary><b> Three communicating agents </b></summary>\n\n```python\n\nA toy numbers game, where when given a number `n`:\n- `repeater_agent`\'s LLM simply returns `n`,\n- `even_agent`\'s LLM returns `n/2` if `n` is even, else says "DO-NOT-KNOW"\n- `odd_agent`\'s LLM returns `3*n+1` if `n` is odd, else says "DO-NOT-KNOW"\n\nFirst define the 3 agents, and set up their tasks with instructions:\n\n```python\nfrom langroid.utils.constants import NO_ANSWER\nfrom langroid.agent.chat_agent import ChatAgent, ChatAgentConfig\nfrom langroid.agent.task import Task\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\nconfig = ChatAgentConfig(\n    llm = OpenAIGPTConfig(\n        chat_model=OpenAIChatModel.GPT4,\n    ),\n    vecdb = None,\n)\nrepeater_agent = ChatAgent(config)\nrepeater_task = Task(\n    repeater_agent,\n    name = "Repeater",\n    system_message="""\n    Your job is to repeat whatever number you receive.\n    """,\n    llm_delegate=True, # LLM takes charge of task\n    single_round=False, \n)\neven_agent = ChatAgent(config)\neven_task = Task(\n    even_agent,\n    name = "EvenHandler",\n    system_message=f"""\n    You will be given a number. \n    If it is even, divide by 2 and say the result, nothing else.\n    If it is odd, say {NO_ANSWER}\n    """,\n    single_round=True,  # task done after 1 step() with valid response\n)\n\nodd_agent = ChatAgent(config)\nodd_task = Task(\n    odd_agent,\n    name = "OddHandler",\n    system_message=f"""\n    You will be given a number n. \n    If it is odd, return (n*3+1), say nothing else. \n    If it is even, say {NO_ANSWER}\n    """,\n    single_round=True,  # task done after 1 step() with valid response\n)\n```\nThen add the `even_task` and `odd_task` as sub-tasks of `repeater_task`, \nand run the `repeater_task`, kicking it off with a number as input:\n```python\nrepeater_task.add_sub_task([even_task, odd_task])\nrepeater_task.run("3")\n```\n\n</details>\n\n<details>\n<summary><b> Simple Tool/Function-calling example </b></summary>\n\nLangroid leverages Pydantic to support OpenAI\'s\n[Function-calling API](https://platform.openai.com/docs/guides/gpt/function-calling)\nas well as its own native tools. The benefits are that you don\'t have to write\nany JSON to specify the schema, and also if the LLM hallucinates a malformed\ntool syntax, Langroid sends the Pydantic validation error (suitiably sanitized) \nto the LLM so it can fix it!\n\nSimple example: Say the agent has a secret list of numbers, \nand we want the LLM to find the smallest number in the list. \nWe want to give the LLM a `probe` tool/function which takes a\nsingle number `n` as argument. The tool handler method in the agent\nreturns how many numbers in its list are at most `n`.\n\nFirst define the tool using Langroid\'s `ToolMessage` class:\n\n\n```python\nfrom langroid.agent.tool_message import ToolMessage\nclass ProbeTool(ToolMessage):\n  request: str = "probe" # specifies which agent method handles this tool\n  purpose: str = """\n        To find how many numbers in my list are less than or equal to  \n        the <number> you specify.\n        """ # description used to instruct the LLM on when/how to use the tool\n  number: int  # required argument to the tool\n```\n\nThen define a `SpyGameAgent` as a subclass of `ChatAgent`, \nwith a method `probe` that handles this tool:\n\n```python\nfrom langroid.agent.chat_agent import ChatAgent, ChatAgentConfig\nclass SpyGameAgent(ChatAgent):\n  def __init__(self, config: ChatAgentConfig):\n    super().__init__(config)\n    self.numbers = [3, 4, 8, 11, 15, 25, 40, 80, 90]\n\n  def probe(self, msg: ProbeTool) -> str:\n    # return how many numbers in self.numbers are less or equal to msg.number\n    return str(len([n for n in self.numbers if n <= msg.number]))\n```\n\nWe then instantiate the agent and enable it to use and respond to the tool:\n\n```python\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\nspy_game_agent = SpyGameAgent(\n    ChatAgentConfig(\n        name="Spy",\n        llm = OpenAIGPTConfig(\n            chat_model=OpenAIChatModel.GPT4,\n        ),\n        vecdb=None,\n        use_tools=False, #  don\'t use Langroid native tool\n        use_functions_api=True, # use OpenAI function-call API\n    )\n)\nspy_game_agent.enable_message(ProbeTool)\n```\n\nFor a full working example see the\n[chat-agent-tool.py](https://github.com/langroid/langroid-examples/blob/main/examples/quick-start/chat-agent-tool.py)\nscript in the `langroid-examples` repo.\n</details>\n\n<details>\n<summary> <b>Tool/Function-calling to extract structured information from text </b> </summary>\n\nSuppose you want an agent to extract \nthe key terms of a lease, from a lease document, as a nested JSON structure.\nFirst define the desired structure via Pydantic models:\n\n```python\nfrom pydantic import BaseModel\nclass LeasePeriod(BaseModel):\n    start_date: str\n    end_date: str\n\n\nclass LeaseFinancials(BaseModel):\n    monthly_rent: str\n    deposit: str\n\nclass Lease(BaseModel):\n    period: LeasePeriod\n    financials: LeaseFinancials\n    address: str\n```\n\nThen define the `LeaseMessage` tool as a subclass of Langroid\'s `ToolMessage`.\nNote the tool has a required argument `terms` of type `Lease`:\n\n```python\nclass LeaseMessage(ToolMessage):\n    request: str = "lease_info"\n    purpose: str = """\n        Collect information about a Commercial Lease.\n        """\n    terms: Lease\n```\n\nThen define a `LeaseExtractorAgent` with a method `lease_info` that handles this tool,\ninstantiate the agent, and enable it to use and respond to this tool:\n\n```python\nclass LeaseExtractorAgent(ChatAgent):\n    def lease_info(self, message: LeaseMessage) -> str:\n        print(\n            f"""\n        DONE! Successfully extracted Lease Info:\n        {message.terms}\n        """\n        )\n        return json.dumps(message.terms.dict())\n    \nlease_extractor_agent = LeaseExtractorAgent(\n  ChatAgentConfig(\n    llm=OpenAIGPTConfig(),\n    use_functions_api=False,\n    use_tools=True,\n  )\n)\nlease_extractor_agent.enable_message(LeaseMessage)\n```\n\nSee the [`chat_multi_extract.py`](https://github.com/langroid/langroid-examples/blob/main/examples/docqa/chat_multi_extract.py)\nscript in the `langroid-examples` repo for a full working example.\n</details>\n\n<details>\n<summary><b> Chat with documents (file paths, URLs, etc) </b></summary>\n\nLangroid provides a specialized agent class `DocChatAgent` for this purpose.\nIt incorporates document sharding, embedding, storage in a vector-DB, \nand retrieval-augmented query-answer generation.\nUsing this class to chat with a collection of documents is easy.\nFirst create a `DocChatAgentConfig` instance, with a \n`doc_paths` field that specifies the documents to chat with.\n\n```python\nfrom langroid.agent.doc_chat_agent import DocChatAgentConfig\nconfig = DocChatAgentConfig(\n  doc_paths = [\n    "https://en.wikipedia.org/wiki/Language_model",\n    "https://en.wikipedia.org/wiki/N-gram_language_model",\n    "/path/to/my/notes-on-language-models.txt",\n  ]\n  llm = OpenAIGPTConfig(\n    chat_model=OpenAIChatModel.GPT4,\n  ),\n  vecdb=VectorStoreConfig(\n    type="qdrant",\n  ),\n)\n```\n\nThen instantiate the `DocChatAgent` (this ingests the docs into the vector-store):\n\n```python\nagent = DocChatAgent(config)\n```\nThen we can either ask the agent one-off questions,\n```python\nagent.chat("What is a language model?")\n```\nor wrap it in a `Task` and run an interactive loop with the user:\n```python\nfrom langroid.task import Task\ntask = Task(agent)\ntask.run()\n```\n\nSee full working scripts in the \n[`docqa`](https://github.com/langroid/langroid-examples/tree/main/examples/docqa)\nfolder of the `langroid-examples` repo.\n</details>\n\n<details>\n<summary><b> :fire: Chat with tabular data (file paths, URLs, dataframes) </b></summary>\n\nUsing Langroid you can set up a `TableChatAgent` with a dataset (file path, URL or dataframe),\nand query it. The Agent\'s LLM generates Pandas code to answer the query, \nvia function-calling (or tool/plugin), and the Agent\'s function-handling method\nexecutes the code and returns the answer.\n\nHere is how you can do this:\n\n```python\nfrom langroid.agent.special.table_chat_agent import TableChatAgent, TableChatAgentConfig\nfrom langroid.agent.task import Task\nfrom langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig\n```\n\nSet up a `TableChatAgent` for a data file, URL or dataframe\n(Ensure the data table has a header row; the delimiter/separator is auto-detected):\n```python\ndataset =  "https://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-red.csv"\n# or dataset = "/path/to/my/data.csv"\n# or dataset = pd.read_csv("/path/to/my/data.csv")\nagent = TableChatAgent(\n    config=TableChatAgentConfig(\n        data=dataset,  \n        llm=OpenAIGPTConfig(\n            chat_model=OpenAIChatModel.GPT4,\n        ),\n    )\n)\n```\nSet up a task, and ask one-off questions like this: \n\n```python\ntask = Task(\n  agent, \n  name = "DataAssistant",\n  default_human_response="", # to avoid waiting for user input\n)\nresult = task.run(\n  "What is the average alcohol content of wines with a quality rating above 7?",\n  turns=2 # return after user question, LLM fun-call/tool response, Agent code-exec result\n) \nprint(result.content)\n```\nOr alternatively, set up a task and run it in an interactive loop with the user:\n\n```python\ntask = Task(agent, name="DataAssistant")\ntask.run()\n``` \n\nFor a full working example see the \n[`table_chat.py`](https://github.com/langroid/langroid-examples/tree/main/examples/data-qa/table_chat.py)\nscript in the `langroid-examples` repo.\n\n\n</details>\n\n---\n\n# :heart: Thank you to our [supporters](https://github.com/langroid/langroid/stargazers)\n\nIf you like this repo, don\'t forget to leave a star :star: !\n\n# Contributors\n\n- Prasad Chalasani (IIT BTech/CS, CMU PhD/ML; Independent ML Consultant)\n- Somesh Jha (IIT BTech/CS, CMU PhD/CS; Professor of CS, U Wisc at Madison)\n- Mohannad Alhanahnah (Research Associate, U Wisc at Madison)\n- Ashish Hooda (IIT BTech/CS; PhD Candidate, U Wisc at Madison)\n\n',
-    'author': 'Prasad Chalasani',
-    'author_email': 'pchalasani@gmail.com',
-    'maintainer': 'None',
-    'maintainer_email': 'None',
-    'url': 'None',
-    'packages': packages,
-    'package_data': package_data,
-    'install_requires': install_requires,
-    'extras_require': extras_require,
-    'python_requires': '>=3.8.1,<3.12',
-}
-
-
-setup(**setup_kwargs)