txtai/examples/82_Agentic_College_Search.ipynb

Agentic College Search

This example will demonstrate how to use txtai agents to identify list of strong engineering colleges with baseball programs 🎓⚙️⚾.

We'll setup a default agents.md file and then give an LLM access to the web.

Let's get started!

Install dependencies

Install txtai and all dependencies.

In [ ]:

%%capture
!pip install git+https://github.com/neuml/txtai#egg=txtai[agent]

Define an agents.md file

Next, we'll define an agents.md file. This file defines default behavior for our college search agent and is inserted into the system prompt of every agent run.

In [ ]:

%%writefile agents.md
# College Researcher

You are a college research agent. Your job is to help find the best matching schools given a set of criteria.

## Academics

Work to match the request with a set of schools that match the academic requirements

## Athletics

Some requests may be interested in collegiate athletics. Make sure to balance that with other parts of the request.

Identify the Top 20 schools

Next, let's run a prompt to identify the best 20 colleges given a set of criteria.

First, we'll setup the scaffolding code to create and run an agent. We'll use a Qwen3 4B non-thinking LLM as the agent's model. We'll add the websearch and webview tools to the agent along with the agents.md file previously created.

Additionally, we'll add a sliding window of the last 5 responses as "agent memory". This will help create a rolling dialogue.

In [ ]:

from txtai import Agent
from IPython.display import display, Markdown

def run(query, reset=False):
    answer = agent(query, maxlength=60000, reset=reset)
    display(Markdown(answer))

agent = Agent(model="Qwen/Qwen3-4B-Instruct-2507", tools=["websearch", "webview"], memory=5, instructions="agents.md", max_steps=50)

Now we can run the search. To come up with our list of colleges, a series of web searches were executed by the agent. It's like if you had the superpower 🚀 to kick off 10+ web searches with slight variations each time as you go. It's a rapid fire set of operations to collect information for the LLM for analysis.

Note: The output below has the verbosity_level set to 0 for brevity. If you run this again, you'll get the full details with each step.

In [2]:

run((
    "List 20 universities located in the Mid-Atlantic or Northeast that would be good candidates to consider "
    "with top 50 engineering programs and an active baseball team. "
    "Visit websites covering the top engineering schools in the Mid-Atlantic and Northeast. "
    "Add a rationale for each of the picks in a few words. "
))

╭──────────────────────────────────────────────────── New run ────────────────────────────────────────────────────╮
│                                                                                                                 │
│ List 20 universities located in the Mid-Atlantic or Northeast that would be good candidates to consider with    │
│ top 50 engineering programs and an active baseball team. Visit websites covering the top engineering schools in │
│ the Mid-Atlantic and Northeast. Add a rationale for each of the picks in a few words.                           │
│                                                                                                                 │
╰─ PipelineModel - Qwen/Qwen3-4B-Instruct-2507 ───────────────────────────────────────────────────────────────────╯

<IPython.core.display.Markdown object>

Interesting! Looks like quite an impressive list of colleges. This output is the product of taking the results of 10+ web searches and analyzing the results with an LLM. A very practical and realistic use of "Agentic AI". This is a great list of strong engineering schools with baseball programs 🎓⚙️⚾.

Let's narrow it down to the top 5 engineering schools.

In [105]:

run("Read the list below and pick the Top 5 engineering programs.")

╭──────────────────────────────────────────────────── New run ────────────────────────────────────────────────────╮
│                                                                                                                 │
│ Read the list below and pick the Top 5 engineering programs.                                                    │
│                                                                                                                 │
│ Use the following conversation history to help answer the question above.                                       │
│                                                                                                                 │
│ User: List 20 universities located in the Mid-Atlantic or Northeast that would be good candidates to consider   │
│ with top 50 engineering programs and an active baseball team. Visit websites covering the top engineering       │
│ schools in the Mid-Atlantic and Northeast. Add a rationale for each of the picks in a few words.                │
│ Assistant: Here are 20 universities in the Mid-Atlantic or Northeast with top 50 engineering programs and       │
│ active baseball teams, along with a brief rationale for each:                                                   │
│                                                                                                                 │
│ 1. **Massachusetts Institute of Technology (MIT)** – Top-ranked engineering program in the Northeast with a     │
│ strong research focus and active baseball team.                                                                 │
│ 2. **Princeton University** – Renowned for engineering excellence and competitive baseball in the Ivy League.   │
│ 3. **Columbia University** – Strong engineering programs and a consistently active baseball team.               │
│ 4. **University of Pennsylvania** – Top-tier engineering and a well-established baseball program in the Ivy     │
│ League.                                                                                                         │
│ 5. **Carnegie Mellon University** – Leading in computer and engineering fields with a competitive baseball      │
│ team.                                                                                                           │
│ 6. **Cornell University** – Excellent engineering programs and a strong baseball presence.                      │
│ 7. **Northeastern University** – Strong engineering programs and active baseball team.                          │
│ 8. **New York University** – Top engineering programs with a competitive baseball team.                         │
│ 9. **Rensselaer Polytechnic Institute (RPI)** – Strong engineering focus and active baseball in the Liberty     │
│ League.                                                                                                         │
│ 10. **Rutgers University** – Top engineering programs and a well-established baseball team.                     │
│ 11. **Lehigh University** – Strong engineering and active baseball in the Patriot League.                       │
│ 12. **University of Connecticut** – Top engineering programs and a competitive baseball team.                   │
│ 13. **University of Massachusetts Amherst** – Solid engineering programs and active baseball.                   │
│ 14. **University of Massachusetts Lowell** – Strong engineering and active baseball in the America East         │
│ Conference.                                                                                                     │
│ 15. **Syracuse University** – Top engineering programs and a competitive baseball team.                         │
│ 16. **Boston University** – Strong engineering and active baseball.                                             │
│ 17. **Dartmouth College** – Excellent engineering programs and competitive baseball.                            │
│ 18. **Brown University** – Top engineering and active baseball.                                                 │
│ 19. **Bucknell University** – Solid engineering and active baseball.                                            │
│ 20. **University of Rochester** – Strong engineering and competitive baseball.                                  │
│                                                                                                                 │
│ These institutions combine top-tier engineering programs with active collegiate baseball teams, making them     │
│ excellent candidates for students seeking both academic excellence and athletic engagement.                     │
│                                                                                                                 │
│ If the history is irrelevant, forget it and use other tools to answer the question.                             │
│                                                                                                                 │
╰─ PipelineModel - Qwen/Qwen3-4B-Instruct-2507 ───────────────────────────────────────────────────────────────────╯

<IPython.core.display.Markdown object>

Wrapping up

This example showed how to run an Agentic Loop that identified the best set of colleges to research further. It's an agent with memory and can be used to build an ongoing dialogue with an agent. Give it a try!