try-agents-haystack

Trying the Agents 🕵️ --> new feature introduced in Haystack 1.15.0 to make Large Language Models resolve complex queries and tasks.

As of release 1.15, Haystack implements Agents 🕵️!

Put simply, an Agent is a Large Language Model with a specific prompt.

Based on that prompt, the Agent can answer complex questions, by performing a sequence of steps.

At each step, the Agent (our brain 🧠) can select a Tool from its toolbox 🧰 and use it to accomplish a task.

Some examples of Tools: Web Search, calculator, several pipelines/nodes (available in Haystack)...

Demo

In the video, you see the Agent in action with the task of answering questions about books on my reading list.

• I uploaded a CSV containing a minimal reading list 📚
• I initialized two Tools ⚒️:
  • A Question Answering component to answer questions about my reading list
  • A Search tool 🔎🌐, which can browse the web and find information
• I defined an Agent, based on Davinci model and equipped with the Tools defined above.
• Now I can ask complex questions, such as "Can you provide me with information on the shortest book on my reading list, including author and price on Amazon?"" 🚀

Currently, Agents work great with OpenAI Davinci model. Soon, open-source models will also be supported, so you can have the power of LLM at your disposal without giving up your data!

More information

Agents in Haystack

• Introducing Agents in Haystack: Make LLMs resolve complex tasks by Tuana Çelik
• Tutorial: Answering Multihop Questions with Agents
• Haystack documentation: the Agent

Papers

• ReAct: Synergizing Reasoning and Acting in Language Models
• MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning