Quantix
Iterative Cognitive Reactor & Problem Solver
A formal problem-solving reactor that combines DAG-driven task construction, multi-model mesh execution (GPT-4o & DeepSeek), and strict multi-critic consensus gating with reflective episodic memory.
System Overview & Engineering Purpose
Quantix operates as a stateful, iterative cognitive reactor (`CognitiveReactor`). When given a complex problem, it executes up to 3 structured solving attempts. For each cycle, ProblemParser categorizes the challenge, NeuroPlanner establishes an execution DAG, CognitiveMesh coordinates multi-agent thought generation, and MultiCriticPanel issues an ACCEPT or REJECT verdict. On success, insights are permanently integrated into episodic memory via CognitiveMirror.
Execution Pipeline & Component Nodes
Step-by-Step Runtime Protocol
Perception & Parsing
ProblemParser ingests raw unstructured problem prompts and extracts entities, constraints, and verifiable deliverables.
NeuroPlanning & Task Graphing
NeuroPlanner generates an actionable plan, which TaskConstructor transforms into an executable dependency graph.
Cognitive Mesh Execution
Executes nodes across heterogeneous reasoning agents (AnalyticalThinker on GPT-4o, DeepSeek fallback) collecting thoughts.
Critique Gating & Memory Reflection
MultiCriticPanel audits generated thoughts in batches. If approved, Synthesizer emits the solution and CognitiveMirror records insights.
Implementation Snippet
for (let attempt = 1; attempt <= this.maxAttempts; attempt++) {
memory.startNewAttempt();
memory.set('attemptNumber', attempt);
const percepts = await this.perception.parse(rawProblem);
const plan = await this.planner.plan(percepts, lastCritiqueFeedback);
const taskGraph = await this.taskConstructor.construct(plan);
const thoughts = await this.cognitiveMesh.executeTaskGraph(taskGraph);
const critique = await this.critic.evaluate(thoughts, percepts);
if (critique.isAccepted) {
const solution = await this.synthesizer.synthesize(thoughts, percepts);
await this.mirror.reflect(memory);
return { success: true, solution, attempts: attempt };
}
lastCritiqueFeedback = critique.feedback;
}