ProjAgent: Procedural Similarity Retrieval for Repository-Level Code Generation

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: QiHong Chen, Aaron Imani, Iftekhar Ahmed
• Venue: arXiv: cs.SE
• Published: 2026-07-09
• Collected: 2026-07-10
• URL: http://arxiv.org/abs/2607.08691v1
• PDF: https://arxiv.org/pdf/2607.08691v1
• DOI: UNKNOWN

Plain-Language Brief

The paper introduces ProjAgent, a system for generating code at the repository level that considers procedural similarity as a new method for retrieving appropriate functions. ProjAgent deconstructs target functions into intermediate steps, uses an agentic workflow to retrieve similar functions from the repository, and integrates these with conventional semantic retrieval for richer context. It also incorporates a feedback loop using compiler and static analysis for code repair.

What the Document Says

• ProjAgent is a system for repository-level code generation that introduces procedural similarity as an explicit retrieval signal.
• ProjAgent decomposes target functions into intermediate reasoning steps.
• ProjAgent employs an agentic workflow to retrieve repository functions with similar procedural behavior at each step.
• The retrieved procedural context is integrated with conventional semantic retrieval for code generation.
• ProjAgent incorporates a conservative static-analysis feedback loop for iterative code repair.
• Evaluation on REPOCOD shows ProjAgent achieves 41.14% Pass@1, outperforming existing retrieval-based baselines.

Key Concepts Mentioned

• ProjAgent
• repository-level code generation
• procedural similarity
• intermediate reasoning steps
• agentic workflow
• conventional semantic retrieval
• compiler feedback
• static analysis feedback
• REPOCOD

Claims Made by the Document

• Procedural similarity is an effective and previously unexplored retrieval dimension for repository-level code generation.

Evidence Used

• Evaluation on REPOCOD shows ProjAgent achieves 41.14% Pass@1, outperforming existing retrieval-based baselines.

Methods / Approach

• Decomposition of target functions into intermediate reasoning steps
• Agentic workflow for repository function retrieval
• Integration of procedural and conventional semantic retrieval
• Conservative static-analysis feedback loop for code repair

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

The paper does not discuss potential limitations or challenges beyond the scope of its methodology.

Open Questions

• What are other potential dimensions for repository-level code retrieval?
• How scalable is ProjAgent in large repositories?

Notes for Future Q&A

May include interpretation. Labeled where it does.

• The paper does not discuss how ProjAgent handles functions with complex cross-file dependencies mentioned in the abstract.
• It is unclear whether ProjAgent can adapt to project-specific conventions other than through procedural similarity.

Raw Abstract / Excerpt

Repository-level code generation requires implementing target functions while accounting for complex cross-file dependencies and project-specific conventions. Existing retrieval methods predominantly rely on lexical, structural, or semantic similarity, often overlooking repository functions that implement similar procedural logic despite differing in identifiers or application domains. We propose ProjAgent, a repository-level code generation system that introduces procedural similarity as an explicit retrieval signal. ProjAgent decomposes the target function into intermediate reasoning steps and employs an agentic workflow to retrieve repository functions that exhibit similar procedural behavior at each step. The retrieved procedural context is integrated with conventional semantic retrieval to construct a richer repository context for code generation. ProjAgent further incorporates a conservative static-analysis feedback loop that iteratively repairs generated code using compiler and static-analysis feedback. Evaluated on REPOCOD, ProjAgent achieves 41.14% Pass@1, outperforming existing retrieval-based baselines. These results demonstrate that procedural similarity is an effective and previously unexplored retrieval dimension for repository-level code generation.