A quick guide to how things work:
- Converse receives messages from external messaging channels or directly via its own API.
- It creates a comment against a new request, or adds it to an existing request as required.
- Converse parses the comment, and attempts to map it against an existing active conversation.
- If it maps to an existing conversation, Converse updates that conversation accordingly.
- If not, it attempts to see if it should match another older conversation, or start a brand new one.
- If a conversation input is accepted, it executes any required modules, and proceeds to the next stage.
- It checks if there is any outstanding questions or missing information it needs to ask, and if so, responds accordingly.
- If the conversation is complete, it can fire off a final webhook as required.
- This webhook can be used to send back a customised response, or to add additional questions as required.
- If it fails to understand the comment at all, it can trigger a configurable 'unknown response', which can send a message back to the user and/or trigger a webhook to enable human escalation or fallback.
Why not go and check out how it works by saying ‘hello’ at converse.ai.
Updated about a year ago