This is a submission for the Algolia Agent Studio Challenge: Consumer-Facing Conversational Experiences
What I Built
I built corn, a private, consumer-facing conversational AI designed to support people who are trying to quit porn and regain control over compulsive habits.
corn is not a general chatbot and not a therapist.
It’s a calm, judgment-free recovery companion that focuses on:
- Managing urges in the moment
- Handling relapse without shame
- Staying motivated during difficult phases
- Following a structured 90-day recovery program
- Anonymous journaling and self-reflection
The core problem corn addresses is isolation. Many people struggle silently with this habit and don’t want lectures, guilt, or explicit discussions. corn provides a safe space where users can simply talk — especially during moments when willpower is weakest.
The conversational experience is intentionally simple:
- Short, supportive responses
- No explicit content
- No medical claims
- Focused on “get through this moment” rather than perfection
Demo
đź”— Live Demo:
👉 https://corn-quitter.vercel.app/
📸 Screenshots:
RATE LIMIT IN GOOGLE GEMINI 2.5 FLASH FREE TIER stops the request for providing response. Overall the app works correctly in free testing process using Algolia Sandbox open AI
Testing Video G-Drive Link:
https://drive.google.com/file/d/17dZtFUAm1q4cPE5yExNzTKs60kbANQee/view?usp=sharing
How I Used Algolia Agent Studio
Algolia Agent Studio is the core engine behind corn’s conversational experience.
Instead of putting everything into one index, I designed the agent using multiple purpose-driven indexes, each with a clear responsibility:
Indexed Data Structure
corn_core_intents
Handles real-time conversations like urges, relapse support, motivation, and fallback handling.
corn_90_day_program
Contains structured recovery logic mapped to days and phases (Day 1–90).
corn_journaling_prompts
Stores anonymous journaling prompts that help users process emotions through writing.
Why this mattered
This separation allowed me to:
Route user queries to the right knowledge source
- Avoid mixing emotional support with structured program data
- Keep responses predictable, safe, and context-aware
- Prompt & Instruction Design
- I used strict system instructions to ensure the agent:
- Never produces explicit or triggering content
- Uses a supportive, non-judgmental tone
- Stays strictly within recovery scope
- Uses emojis sparingly to maintain warmth 🌱
Retrieval from Algolia indexes ensures the agent responds based on indexed intent-specific data, not generic LLM guessing.
Why Fast Retrieval Matters
For this use case, speed and relevance are critical.
When someone types:
“I have an urge right now”
They don’t want:
- A long explanation
- A generic motivational speech
- A delayed response
They need:
- The right response
- Immediately
- In the right emotional tone
- Algolia’s fast, contextual retrieval ensures:
- The correct intent is matched instantly
- The agent responds with focused, calming guidance
- No unnecessary or off-topic content is introduced
This makes the experience feel present and reliable, which is especially important for sensitive, time-critical moments.
DEV Team Member Id : https://dev.to/abbas7120
Top comments (0)