This is a submission for the Algolia Agent Studio Challenge: Consumer-Facing Conversational Experiences
What I Built
Hello Everyone!
As an event vendor for 5 years, I know that it can be hard to find events you are looking for. You need to sell at places with your target audience.
It would be really nice to have a natural conversation about what you are looking for, rather than querying hard parameters all the time. It is possible to discover even more opportunities with natural language models. It makes it much easier for the exhausted business owner to say "Find me events to vend at this April near Chicago". Instead of typing, checking, toggling, whatevering-- a bunch of fields.
So let's output some related findings for the very tired Miss Bosswoman CEO!
By the way guys, I've been working on this project here and there through time. This Algolia search intelligent chat bot is a wonderful feature.
Demo
How I Used Algolia Agent Studio
I took this opportunity to learn about Algolia and implement it into one of my projects: GoVend, Connecting vendors to events.
Implementing my first intelligent chat assistant was a challenge for sure. Initially, I set up an Algolia agent with Gemini.
“Carl”, the agent I created in Algolia, would take user queries and find related events in the available data.
You can choose what attributes should be searchable on each index. An index is a group of data on Algolia. For me, I had an “Events” index, which included things like name, date, city, etc. I left most of them on to improve searchability.
I explored Algolia’s MCP Server integration. It was fairly straightforward, and I was getting excited about it. By assigning tools to agents, we can build them to be more accurate and useful.
And then of course...
I ran out of API requests as usual. I pushed the envelope a bit here with an Ollama wildcard.
I first connected local Ollama (model: llama3.2:latest)to my app and got that working with my local data.
Why Fast Retrieval Matters
I integrated Algolia's Search API with llama3.2 for natural language event discovery. Users can ask questions like 'Find events in Chicago' and get intelligently formatted results. Algolia's powerful search combined with Ollama's natural language understanding is a winning scenario.
I realized that maybe having a fallback method for searches was a good idea to handle errors in production. Llama3.2 will be able to find some event data no matter what.
I used enhanced Algolia search capabilities by using:
- Zipcode Detection
- Geo-Location Search
- Geocodes the zipcode to latitude/longitude
- Uses Algolia's aroundLatLng geo search
- Smart radius selection:
- 50km if query includes "near", "nearby", "close", "around"
- 100km default radius
- Distance Display
- Shows distance in miles for each event
- Sorted by proximity automatically
- Combined with Time Filters
- Can search "events near 35801 this spring"
- Combines zipcode + seasonal date filtering
Example Queries:
- "Find events near 35801"
- "What's happening close to 90210 in June?"
Highlights that Align with Algolia’s Standards:
Unique Ollama Integration
Error Handling
Graceful fallback system
Clear logging and debugging
No breaking when services unavailableNatural Language Processing
Extracts locations from queries
Understands user intent
Aesthetically pleasing results formattingReal-World Problem Solving
Faced API limits (common in production)
Built resilient solution
The Future:
This was an interesting journey for me, lots of exciting tiny wins, and a couple long drawn out learning experiences. I like that Algolia allows you to configure multiple agents in the background. One day when I stop running out of API requests, I’d like to do a little experiment and have multiple agents talking to each other and see where that ends up.
Repo: GoVend ai_beta
Top comments (0)