This is the code for the Talk "Beyond Chatbots: Financial Innovation with Agentic LLMs". This repo draws inspiration from the paper AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation by Wu et al., and from the examples from LangGraph.
Language Agent Tree Search (LATS), by Zhou, et. al, is a general LLM agent search algorithm that combines reflection/evaluation and search (specifically monte-carlo trees search) to get achieve better overall task performance compared to similar techniques like ReACT, Reflexion, or Tree of Thoughts:
You will find two notebooks which you can directly open in Google Colab:
LATS.ipynb: This is the notebook shown during the talk for creating LATS.LangGraph_multi_agents_investment_analysis.ipynb: This is the notebook to create a multi-agent with LangGraph.
For the multi-agents, you will use the following tools:
- Exa, after account login, get your API key here. To find the exact content you're looking for on the web using embeddings-based search.
- SerpApi here, after account login, get your API key to do look for existing patents.
- Python REPL, please note that Python REPL can execute arbitrary code on the host machine (e.g., delete files, make network requests). Use with caution.
- Tools to access and write to a
.txtfile and create a plot of historical prices. - How to define utilities to help create the graph.
- How to create a team supervisor and the team of agents.
The interaction of the multi-agents looks like this:
