slopo

slopo command [options]

slopo is a CLI tool that detects non-exact code duplication using embedding models. Unlike copy-paste scanners, it finds semantically similar snippets spread across modules or distant locations in large files — the kind of duplication hardest for humans and AI agents to spot.The workflow is init → index → embed → analyze. Indexing parses supported languages into AST-based code units stored in a local SQLite database (slopo.db). Embedding calls go through LiteLLM-compatible providers (for example Voyage AI). Analysis compares cosine similarity, groups pairs into clusters, and reranks by distance in the codebase.Supported languages include Python, TypeScript, JavaScript, Java, Kotlin, C#, Go, and Rust. Reviewed clusters can be added to slopo.ignore.txt and excluded from future reports.

init

Create a configuration file template.

Parse and index code units from source_dir.

Calculate embeddings for indexed code via a configured provider.

Find similar code clusters and write a report to report_dir.

Validate config and display all parameters.

Override the default configuration file location.

Key settings in the config file:

source_dir — directory to index (required)

source_dir_exclude — .gitignore-style exclude patterns

db_file — SQLite database path (default: slopo.db)

report_dir — output directory for analysis reports

ignore_file — text file listing reviewed cluster hashes to skip

embedding_model — model name in LiteLLM format

embedding_dimensions — vector dimensions for the model

embedding_api_key — provider key (or set SLOPO_EMBEDDING_API_KEY)

similarity_threshold — minimum cosine similarity between pairs

rerank_threshold — minimum score after distance-based reranking

body_node_count_threshold — minimum AST body size for a code unit

Requires an external embedding API key and network access for the embed step. Exact-copy duplicates are reported but handled differently from merely similar code.

rg(1), ast-grep(1)

Source code · Homepage