Quest to Solvin

Project Overview

Quest to Solvin is an AI-driven interactive storytelling prototype inspired by VRMMORPG-themed manhwas such as Overgeared and Legendary Moonlight Sculptor. These stories often feature advanced AIs capable of creating immersive worlds filled with lifelike NPCs. This project set out to explore how close we are to that vision today.

Core Vision: Create a random NPC with personality traits, generate a portrait, and let users chat with them in a way that feels dynamic and story-driven.

Problem & Opportunity

Interactive storytelling has seen experiments with AI, but the boundaries of what's possible remain underexplored. Most narrative games are pre-scripted, limiting spontaneity. With the rise of LLMs, the opportunity lies in letting AI dynamically generate unique characters, conversations, and quests in real time.

Quest to Solvin takes a nod from fantasy novels and games that shaped my imagination and attempts to push AI into that space.

Scope & Features

Must-haves

NPC generation with unique personality, portrait, and chat functionality.

Nice-to-haves

Quest generation (implemented in early form).

Dropped features

Town exploration and map movement—too complex for Streamlit in this version.

NPC creation, quest objectives, lore, and user interaction were tied together through OpenAI APIs. The system begins with NPC generation, produces a pixel-art portrait, and equips the NPC with a quest aligned with their personality. Lore was loaded from markdown files for easy extensibility.

Technical Approach

Stack & Tools

Streamlit for the interface

OpenAI GPT models for dialogue and quest logic

DALL·E for pixel-art portrait generation

Markdown for lore files

Docker for portability

GCP for hosting

Architecture & Design

Code structured using OOP (npc.py, quest.py, world.py)

NPC personalities injected into prompts for consistency

Chat context tracked via message state lists

Lore loaded dynamically into the system to ground NPCs

Challenges & Trade-offs

Evolving APIs

OpenAI model deprecations required constant updates.

Cost Considerations

Image generation via DALL·E was expensive; mitigated by requiring users to provide their own OPENAI_API_KEY.

Trade-offs

As a prototype, consistency was less important than showcasing freedom and creativity.

Validation Challenges

NPCs and quests worked best when users role-played properly; nonsense input led to nonsense output.

UX & Design

Design Philosophy

The flow was designed to mimic an isekai entry: players awaken in an unknown world and encounter an NPC who gives them direction.

Interface

Streamlit-based chat UI with portraits and quests

Priority

User input freedom over scripted storylines

Limitation

No formal usability testing; design relied on my own playtesting

Outcomes & Learnings

What Worked

NPC generation and chat were surprisingly fun and engaging.

What Fell Short

NPCs sometimes felt pushy in quest-giving, and portraits were costly to generate.

Key Lessons Learned

Prompt engineering is powerful but requires iteration.

API integration with LLMs can create engaging prototypes quickly.

Streamlit is great for MVPs but limiting for complex UIs.

Proudest Achievement

Building my first complete AI-driven storytelling prototype from scratch.

Future Directions

Expand Locations

Add branching narratives and multiple world areas

Progression Systems

Implement character progression for longer sessions

AI Agents

Make NPCs more autonomous and lifelike

Persistence Features

Add user accounts and saved game states

Reflection

This project sharpened my skills in prompt design, LLM integration, UI prototyping, and system deployment. The hardest part was UI organization and balancing scope with technical limits.

Advice for Others

"Start small, make it work, then refine." Even a simple prototype can reveal deep insights into how AI can change storytelling.

Outcome

Timeline

Role

Technologies