Paper Reference
- Title: Agent-centric Debugging Interface (ADI)
- Authors: Jiahong Xiang, Xiaoyang Xu, Xiaopan Chu, et al.
- Year: 2026
- URL: https://arxiv.org/abs/2604.24212
- Venue: FSE 2026 (ACM SIGSOFT Conference on Foundations of Software Engineering)
Paper Summary
Introduces a function-level interaction paradigm for cost-efficient autonomous debugging. Powered by Frame Lifetime Trace — a comprehensive data structure encapsulating a function's stateful execution trace with high-level navigational commands.
Proposed Feature
Implement Frame Lifetime Tracing for granular function-level debugging:
Core Capabilities
- Function-Level Traces: Capture not just agent decisions but individual function calls with their full state
- Frame Inspection: Inspect function arguments, return values, and intermediate state at each call
- Navigation Commands: Jump to next/previous function call, filter by function name, search within frames
- Cost Analysis: Show per-function token cost and latency breakdown
Technical Approach
- Extend the SDK's auto-patch system to capture function-level traces (not just agent-level)
- Implement Frame Lifetime Trace data structure matching the paper's design
- Build frame inspection panel in the frontend with collapsible function calls
- Add cost analysis aggregation per function
Impact
Provides the granularity needed for debugging complex agent code where the issue is inside a specific function, not just at the decision level. The FSE 2026 venue validates this as a top-tier software engineering contribution.
Labels
enhancement, paper-inspired, high-priority
Paper Reference
Paper Summary
Introduces a function-level interaction paradigm for cost-efficient autonomous debugging. Powered by Frame Lifetime Trace — a comprehensive data structure encapsulating a function's stateful execution trace with high-level navigational commands.
Proposed Feature
Implement Frame Lifetime Tracing for granular function-level debugging:
Core Capabilities
Technical Approach
Impact
Provides the granularity needed for debugging complex agent code where the issue is inside a specific function, not just at the decision level. The FSE 2026 venue validates this as a top-tier software engineering contribution.
Labels
enhancement, paper-inspired, high-priority