From One-Shot LLM to Multi-Turn Agent: How I Rebuilt My Text-to-Diagram Tool

The problem: one LLM call isn't enough

I built an AI tool that turns natural language into editable diagrams. You describe a flow, an LLM generates Mermaid code, and the result renders in real time.

The v1 was a single LLM call: user input → LLM → diagram. After a few days in production, two failure modes kept showing up:

Syntax errors with no recovery. The generated Mermaid code looks correct, but Mermaid throws a Parse error and the user sees a red stack trace. They leave.

Silent semantic drift. The user asks for "a login flow" — the LLM helpfully adds a "send verification code" actor. It doesn't ask. It doesn't tell you it added anything.

From One-Shot LLM to Multi-Turn Agent: How I Rebuilt My Text-to-Diagram Tool

Related reading

Let your LLM take real-world actions — without giving it the last word

Lessons from Embedding an LLM Inside a Drag-and-Drop Editor

How to Orchestrate Autonomous Sub-Agents Without Blowing Your LLM Context Window

Designing tools so an LLM actually calls them correctly: 5 patterns from the…

Write Loops, Not Prompts: Why AI Agents Work Better When They Iterate

Your AI Agent doesn't need more tools. It needs better orchestration.