Bullbrain

Documentation

Bullbrain Docs

Bullbrain is a SOL-side AI project concept for chat, model routing, node quality, and synthetic data workflows. These docs explain what exists now, what is simulated, and what should be connected next.

1. Overview

Bullbrain presents a practical AI network product: users interact through chat and agent surfaces, requests are routed to the best available model or node, and useful prompts become material for future data refinement.

Brand Bullbrain AI
Position SOL-side AI inference and model workflow layer
Current build Static landing page, project docs, and free in-browser demo bot
Next build Replace demo bot logic with a live model endpoint and add wallet flow

2. Quickstart

Open the project locally with any static server from the project folder.

python3 -m http.server 5173 --bind 127.0.0.1

Then visit:

http://127.0.0.1:5173/

The landing page does not require a build step. The demo bot also runs without an API key because it is currently a browser-side prototype.

3. Chat Demo

The chat on the homepage is a free test bot. It is meant to validate product wording, user intent, and chat layout before the project commits to a real inference provider.

  • Suggested prompts are wired through buttons in the Chat section.
  • Responses are generated locally by matching keywords in app.js.
  • No prompt is sent to a third-party service in the current version.
  • The replacement point for a real model is getDemoBotReply().

4. Network Concept

The Bullbrain Network should route inference work to nodes based on practical performance signals rather than hype metrics. The first public dashboard can stay simple.

01

Register capacity

Node operators publish basic availability, supported models, and latency region.

02

Measure quality

Requests generate uptime, response speed, and failure-rate signals.

03

Route requests

The app chooses a route by model fit, cost, latency, and recent node health.

5. Models

Bullbrain can begin with a small curated model list. The point is not to claim every model; the point is to make routing, testing, and comparison easy.

General chat Default assistant behavior and community Q&A
Agent tasks Tool-oriented requests, summaries, structured outputs
Project support Docs, roadmap, token information, and launch checklist responses

6. Datagen Workflow

Datagen is the feedback loop. Prompts from real users can become reviewed examples, then evaluation sets, then tuning material when the project is ready.

  1. Collect useful prompts from chat and agent sessions.
  2. Remove spam, private information, and unsafe content.
  3. Write or select high-quality reference answers.
  4. Score model responses against those references.
  5. Promote the strongest examples into tuning datasets.

7. Replacing The Demo Bot

When a live endpoint is selected, replace the local response function in app.js with a request to that endpoint. Keep API keys on a server, not in the browser. 

async function getLiveBotReply(message) {
  const response = await fetch("/api/chat", {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({ message })
  });

  const data = await response.json();
  return data.reply;
}

For the current prototype, getDemoBotReply() stays local so the page remains free to run and easy to share.

8. Launch Checklist

  • Lock homepage copy and remove any temporary claims.
  • Choose the real chat endpoint and add a small server-side proxy.
  • Add wallet connect only after the user flow is clear.
  • Define node reputation fields before publishing a dashboard.
  • Publish docs for users, node operators, and model contributors.
  • Keep Bullbrain visually independent from Solana and other chain brands.