Q55: What is the sequence of events when a developer triggers an inline code completion?
- IDE captures local context surrounding cursor.
- Client constructs a prompt request payload.
- Secure proxy transmits request to the LLM backend.
- Model generates candidate completions.
- Post-processing filters check for secrets and duplication.
- Rendered in IDE as ghost text.
Key Takeaways
- Context capture $ o$ Prompt build $ o$ Proxy LLM call $ o$ Post-processing filter $ o$ Ghost text.
Knowledge Check
When does the Duplication Detection Filter run in the Copilot completion pipeline?