Convoship gives you two complementary test surfaces: utterance testing for intent matching, and the session simulator for multi-turn flows.
Utterance test tab
Type a phrase and see matched intent, confidence score, and extracted parameters. Use this while tuning training phrases and entity bindings.
Floating simulator
- Available on every agent page via the chat icon (bottom-right).
- Shows user and bot bubbles; debug trace lists nodes executed per turn.
- Node names from the inspector appear in the trace (not raw IDs).
- Sessions restart automatically after completion when you send a new message.
Reading the trace
Each bot turn may include trace pills for intent match, entity capture, tool calls, and Python scripts. Open JSON in History for the full turn payload including debug metadata.
Workflow simulator
On the workflow Flow tab, WorkflowSimulator exercises draft graphs via POST /v1/runtime/workflows/{id}/sessions and /turns. Supports flow-agent hops, instruction_agent LLM loops, knowledge retrieval, router/decision branching, workspace tool nodes, and nested workflow nodes. Requires Editor (runtime.simulate). Subject to workspace rpm_cap_per_agent per workflow id.
Workflow chains are tested with POST /v1/workflow-chains/{slug}/run from the chain editor. Inbound webhooks use ingress URLs documented on the Workflows overview (HMAC X-Chain-Signature). History tab lists paginated sessions; Publish tab runs validate-publish before release.