Reasonable is a high-performance, reasoning-centered AI agent service built on the modern LlamaIndex Workflows architecture.

Unlike traditional LLM wrappers that simply "get an answer," Reasonable is designed with a "Reasoning-First" philosophy. Here’s a breakdown of what makes it unique:

Core Architecture

Workflows-Native: It uses the v0.14+ LlamaIndex event-driven architecture, which allows for complex, multi-step agentic behavior (like the ReAct loop) that is far more robust than legacy sequential agents.
Chain-of-Thought (CoT) Extraction: It is specifically engineered to handle "Thinking" tokens. It identifies and streams internal reasoning processes separately from final answers, providing total transparency into how the AI arrived at a conclusion.
Ollama Integration: It is locally-optimized, currently leveraging models like Qwen 3.5 (9B) to perform heavy reasoning tasks without relying on external cloud APIs for the "brain" of the operation.

The "Socratic" Philosophy

Following our latest architectural pivot, Reasonable has transitioned from a simple "solver" into a Socratic Tutor.

The Heuristic Ladder: Instead of blurting out answers, the agent uses diagnostic tools to analyze the "shape" of a problem, then guides the user through identification, strategy, and execution.
Scaffolded Interaction: It treats tools as "Math Diagnostic Engines" that provide facts to the agent, which then uses those facts to challenge and guide the user.

Technical Stack

Backend: FastAPI with full CORS support and SSE (Server-Sent Events) streaming.
Protocol: A specialized Event Schema pulse, reasoning, call, result, message) that allows frontends (like the one you're running via npm run dev) to render complex agentic thoughts and tool interactions in real-time.

In short, Reasonable is about making AI interaction more transparent, educational, and—as the name implies—driven by clear, observable reasoning.

kdotdev