Ananse Ntentan: The AI Storyweaver

Inspiration

In the heart of Akan folklore lives Ananse the Spider—not just a trickster, but the keeper of all stories. For centuries, his webs have captured the wisdom, humor, and spirit of a people, proving that stories are the threads that bind humanity together.

We built Ananse Ntentan ("The Spider's Web") as the digital rebirth of this ancient tradition. Our inspiration was simple: everyone has a story, but not everyone has the voice to tell it. We wanted to build a loom where raw imagination—spoken, sketched, or typed—is woven into vibrant, visual realities by an Agentic AI.

What it does

Ananse Ntentan is an agentic storytelling platform that democratizes creativity. It allows users to tell stories in three ways:

1. The Speaker: When you speak into our Narrative Panel, the AI doesn't just transcribe words; it listens to the tremble in your voice or the joy in your laugh, translating emotion into art.
2. The Artist: When you sketch a rough idea, it sees your vision and renders it into a professional comic strip, preserving your unique style.
3. The Collaborator: Through our Chat Interface, you can integrate comic generation seamlessly into conversations. Ananse lives on as an AI co-author, helping weave complex narratives with friends in real-time.
4. The Community: Stories are shared on the "My Feed", granting everyone the opportunity to see their own creations alongside the stories of others.

How we built it

We built this platform using a modern MERN stack supercharged by Google's Gemini 3.0 Pro. But unlike standard apps, we implemented a true Agentic Workflow:

• Multimodal Sensing: We pass raw audio and image data directly to Gemini, allowing it to "hear" hesitation and "see" style.
• Thought Signatures: To solve the "goldfish memory" problem of LLMs, we implemented a persistence layer where the AI saves its reasoning state (thoughtSignature) to MongoDB. This allows it to pick up a story exactly where it left off, maintaining deep continuity.
• The Autonomous Agent: The AI isn't just a chatbot; it's an agent with tools. It can autonomously decide to saveStoryToDB, searchSimilarStories to avoid clichés, or generateCharacterReference to create consistent visual sheets.

Challenges we ran into

Building an agent with memory wasn't easy.

• The Silent Monologue: Managing the "Thought Signatures" in a real-time chat was tricky. We had to ensure the AI's "inner thoughts" were preserved without leaking them to the user interface, while still using them to drive the next turn of the story.
• Streaming Senses: Passing raw binary streams (audio/images) to the LLM via the new Google GenAI SDK required careful handling to ensure the AI could truly "sense" the user's input.

Accomplishments that we're proud of

• The "Context" Brain: Seeing the AI remember a minor character mentioned 50 turns ago in a chat session because of our Thought Signature implementation feels like magic.
• True Multimodal Input: It feels profound to hum a sad tune and have the AI generate a melancholic story, or shout excitedly and get an action sequence.
• Migration Success: We successfully migrated to the cutting-edge @google/genai SDK to unlock thinking_level: HIGH for deep story reasoning.

What we learned

We learned that Agentic AI is fundamentally different from generative AI. It's not just about prompting; it's about giving the model memory, tools, and the autonomy to use them. We learned how to architect a system where the AI is a partner, not just a tool.

What's next for Ananse Ntentan

• Voice Cloning: Allowing the narrated stories to sound like the user.
• Interactive Comics: Panels that you can click to "step into" and explore via a 360-degree view.
• Mobile App: Bringing the oral tradition of storytelling to the device everyone carries.

As long as humanity is driven, comforted, and connected by stories, our web will remain—proudly inspiring you to pull your tales from depths so deep they know no limits.

Built With

• gemini
• mern