GitHub user gupta-sahil01 edited a discussion: Ambient "operator recommender" — predictive next-operator suggestions on the canvas
## The idea in one line As a user builds a workflow, show 1–3 faded "ghost" operators on the output port of the operator they just added — ranked suggestions for what likely comes next, one click to insert. ## How it differs from the existing chatbot agent The current chatbot is on-demand and conversational: the user describes a goal, the agent generates a workflow. That's powerful, but it assumes the user already knows what to ask for. This feature is ambient and predictive: suggestions surface automatically as part of normal canvas building, with no prompt. It's aimed at the moment where a user has some data on the canvas but doesn't know which operator unlocks the next step. Think Copilot ghost-text, but for the dataflow graph rather than a chat box. The two are complementary — I'd want to reuse the chatbot's LLM plumbing rather than build a parallel stack. ## Questions for the maintainers Existing work: Is anyone already working on operator suggestions / autocomplete on the canvas? I'd rather extend than collide. Mergeability: If this turns out well, would the team be open to a PR? What would make it acceptable (config flags, opt-in, evaluation, etc.)? ## Quick visuals: <img width="346" height="365" alt="image" src="https://github.com/user-attachments/assets/3d549c6f-17bc-4388-849a-cb1766fb9991" /> <img width="346" height="365" alt="image" src="https://github.com/user-attachments/assets/ea37ce9d-d4eb-47b9-94f5-0364cba82e42" /> The **left panel** is Texera as it exists today. The user either drags operators by hand from a side panel, or types a description into the chatbot, which constructs a workflow and offers to apply it. Both modes assume the user already knows what to ask for. The **right panel** is my idea. The user has only built two operators (CSV Reader → Filter). The pulsing teal port on Filter's output is where your feature lives — three faded "ghost" operators appear automatically, ranked by how well they fit the data. Hovering shows the rationale. Clicking a ghost solidifies it into a real operator on the canvas. No chat needed. GitHub link: https://github.com/apache/texera/discussions/5240 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
