Introducing Conversation Transitions

Most chatbots have one mode. They greet you, answer a question, and that's it. If you want to talk about something different, you start over. If the chatbot needs information it doesn't have, it's stuck.

Today we're shipping conversation transitions — the ability for a Hadron chatbot to fluidly move between conversations, handle detours, and come back to where it left off.

What's new

Topics, Conversations, and Stages

Chatbots now have a three-level hierarchy:

• Topics group related conversations (optional)
• Conversations are coherent flows with goals
• Stages are focused steps with prompts and data extraction

Goal-based routing

Every conversation carries goalDescriptions and goalSignals — natural-language descriptions of when that conversation is the right fit. The routing engine matches user messages to conversations semantically, not just by keyword.

The goal stack

When the chatbot needs to detour (e.g., "we need your business type before we can look up government programs"), it pushes a goal onto the stack and comes back when the detour is done. Users never get lost.

Edges with conditions

Conversation designers can wire up routing logic visually: "if the user's profile is incomplete, go to the profile-building conversation first." Edges evaluate conditions against the user's data and fire automatically.

onTrack — continuous drift detection

Every chatbot response now reports whether the conversation is still on track. When it's not, the system re-routes immediately instead of drifting into irrelevance.

Why this matters

Chatbots fail when they lack structure. They succeed when the structure is invisible to the user but present in the system. Conversation transitions are the infrastructure that makes multi-topic chatbots feel natural instead of rigid.

Read the full documentation at Conversation Routing.