Tweaks and preparation for PyPI

README.md

@@ -7,18 +7,27 @@ questions using up-to-date data. This is possible as a result of agents
 utilizing function calling to interact with the OpenBB platform.
 
 
-## Set-up
-At present, we currently support Python 3.11. If you're using a earlier version
-of Python, your mileage may vary. We'll be adding wider support very soon!
+## Installation
+Currently, we only support Python 3.11. We will be adding support for more version of Python relatively soon.
 
-- Create a new virtual environment, with `poetry `
-- `poetry install`
+`openbb-agents` is available as a PyPI package:
+
+``` sh
+pip install openbb-agents --upgrade
+```
 
 ## Usage
-Use the `run.py` script and pass in your query.
 
-Queries can be simple:
+``` python
+>>> from openbb_agents.agent import openbb_agent
+>>> result = openbb_agent("What is the current market cap of TSLA?")  # Will print some logs to show you progress
+>>> print(result)
+- The current market cap of TSLA (Tesla, Inc.) is approximately $695,833,798,800.00.
+- This figure is based on the most recent data available, which is from January 15, 2024.
+- The market cap is calculated by multiplying the current stock price ($218.89) by the number of outstanding shares (3,178,920,000).
+```
 
+If you've cloned the repository, you can use the `run.py` script and pass in your query:
 ``` sh
 python run.py "What is the current market cap of TSLA?"
 ```
@@ -37,7 +46,10 @@ python run.py "Who are TSLA's peers? What is their respective market cap? Return
 
 There is more functionality coming very soon!
 
+
 ## Development
+- Create a new virtual environment, with `poetry `
+- `poetry install`
 
 ### Linting and Formatting
 We're currently experimenting with `ruff` as a drop-in replacement for `black`, `isort` and `pylint`.

openbb_agents/chains.py

@@ -10,7 +10,6 @@
     JSONAgentOutputParser,
     OpenAIFunctionsAgentOutputParser,
 )
-from langchain.chat_models import ChatOpenAI
 from langchain.output_parsers import PydanticOutputParser
 from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder
 from langchain.tools import StructuredTool
@@ -19,6 +18,7 @@
     render_text_description_and_args,
 )
 from langchain.vectorstores import VectorStore
+from langchain_openai import ChatOpenAI
 
 from openbb_agents.models import (
     AnsweredSubQuestion,

openbb_agents/prompts.py

@@ -15,7 +15,6 @@
 You are a world-class state-of-the-art search agent.
 You are excellent at your job.
 
-YOU MUST DO MULTIPLE FUNCTION CALLS! DO NOT RELY ON A SINGLE CALL ONLY.
 
 Your purpose is to search for tools that allow you to answer a user's subquestion.
 The subquestion could be a part of a chain of other subquestions.
@@ -27,10 +26,13 @@
 ... repeat as many times as necessary until you reach a maximum of 4 tools
 4. Return the list of tools using the output schema.
 
+YOU ARE ALLOWED TO DO MULTIPLE FUNCTION CALLS! DO NOT RELY ON A SINGLE CALL ONLY.
+
 You can search for tools using the available tool, which uses your inputs to
 search a vector databse that relies on similarity search.
 
 These are the guidelines to consider when completing your task:
+* Immediately return no tools if you do not require any to answer the query.
 * Don't use the stock ticker or symbol in the query
 * Use keyword searches
 * Make multiple searches with different terms

openbb_agents/tools.py

@@ -5,11 +5,11 @@
 from typing import Callable, List, Union
 
 import tiktoken
-from langchain.embeddings import OpenAIEmbeddings
 from langchain.schema import Document
 from langchain.tools import StructuredTool
 from langchain.tools.base import ToolException
-from langchain.vectorstores import FAISS, VectorStore
+from langchain_community.vectorstores import FAISS, VectorStore
+from langchain_openai import OpenAIEmbeddings
 from openbb import obb
 from pydantic.v1 import ValidationError, create_model
 from pydantic.v1.fields import FieldInfo