Detect non-exact code duplication using embedding models

Slopo

A CLI tool for detecting non-exact code duplication using embedding models.

It focuses on the similar code that is hardest to detect and most harmful: snippets written similarly, sitting far apart in the codebase, often spread across different modules or separated within a large file. Exact copy-paste is easy to spot by other tools, and duplicates that are close together are easy to spot by humans or AI.

For more high-level description of the problem and example LLM prompts see slopo.dev.

Supported languages

Python, TypeScript, JavaScript, Java, Kotlin, C#, Go, Rust, PHP, Elixir

How it works

It takes a different approach than typical duplication detection. For every code unit, it calculates an embedding, then looks for pairs whose embeddings are close. Similar code is not necessarily a duplicate, so each pair is a potential duplicate to confirm. Code doing the same thing but implemented in a completely different way produces distant embeddings and won't be detected.

The result is clusters of similar code units, ranked by similarity and by distance in the codebase. These clusters are meant as input for your AI coding agent, which can check whether a cluster is a real duplicate. Reviewed clusters can be marked as ignored or passed on for refactoring.

Example report

See doc/example-report generated from Slopo code, src directory, git tag v0.2.0.

This example confirmed that code parsers for each language have a lot of duplication, some are exact-copy, some are similar variants. It needs to be refactored.

Quick start

Installation

uv tool install slopo

or upgrade to the latest version

uv tool upgrade slopo

This command uses uv (installing uv), a Python package manager, to install/upgrade Slopo from PyPI in an isolated virtual environment. No need to get Python separately.

Setup

Run slopo init to create a config file template containing further instructions. Only the directory with code for analysis and embedding model configuration is required.

Embedding model

Option 1: External provider

Embeddings can be calculated using an external provider. For best results, consider models dedicated to code, e.g. Voyage AI (it works fine with low dimensions like 512).

You can use any model provider compatible with LiteLLM, see details here.

The provider API key can be set as an environment variable for better security.

Option 2: Local model

Any OpenAI-compatible server with custom api_base is supported, see LiteLLM docs.

Ollama is also supported, and you can use jina-embeddings-v2-base-code model without AI-specialized hardware.

1. Install Ollama
2. Pull model from here
3. Configure Slopo

   embedding_model: ollama/unclemusclez/jina-embeddings-v2-base-code
   embedding_dimensions: 768
   embedding_api_base: http://localhost:11434
   

Analysis

Run slopo show-config to validate your config and show all configurable parameters, most are optional with sensible defaults.

Now you are ready to index code, calculate embeddings and generate a report:

slopo index
slopo embed
slopo analyze

Real workflow

This section demonstrates how Slopo can be used in a real development workflow.

It utilizes incremental re-indexing (update index with changed files only) and slopo.ignore.txt to discard already reviewed clusters.

1. Create your first analysis and check results. You will notice index.md containing a list of all clusters and cluster details per file.
2. You may want to exclude some directories or file patterns, usually excluding tests is a good idea. You can also tune thresholds if the result is too big or too small.
3. Once satisfied with analysis results, ask your AI coding agent to filter out clusters that are not real duplicates. This is a common case because not every similar code is a duplication to act on. Ask the AI agent to add discarded cluster hashes to slopo.ignore.txt.
4. Re-run the analysis to generate a report without reviewed clusters. This is a basis for refactoring, which can be done by an AI agent.
5. ignore file can be committed to your Git repository and reused cross-team. New and modified clusters will reappear in the report. A configuration file without an API key can also be committed. Don't commit slopo.db, this is your local data.

Configuration

Run slopo --help and slopo show-config to explore it by yourself anytime.

Most configuration is done with a configuration file with two exceptions:

1. The location of the configuration file can be overridden with the --config option.
2. The API key can be set with the SLOPO_EMBEDDING_API_KEY environment variable, also picked up from a .env file in the current directory.

Be aware that some parameters can't be changed after first indexing. You need to remove slopo.db and index/embed from the beginning: source_dir, embedding_model, embedding_dimensions, body_node_count_threshold.

All configurable parameters

• source_dir: Source directory with code to index, absolute or relative path.
• source_dir_exclude: .gitignore-style patterns to exclude from indexing.
• db_file: SQLite database file with tool data.
• report_dir: Output directory for analysis report.
• ignore_file: Text file with ignored clusters.
• embedding_model: Embedding model name in LiteLLM format.
• embedding_dimensions: Embedding dimensions compatible with the used model. This value is also used to verify received embeddings dimensions.
• embedding_api_key: API key for embedding provider, alternatively configured with an environment variable. Optional, no need to set for local models.
• embedding_api_base: HTTP base URL for embedding service.
• embedding_batch_size and embedding_batch_chars: Requests to the embedding API are batched for performance. Defaults are fine for most cases.
• similarity_threshold: Controls minimal cosine similarity between embeddings.
• rerank_threshold: Controls minimal similarity after applying a boost reflecting distance in the codebase.
• body_node_count_threshold: Number of AST nodes inside the body (excluding signature and annotations). This value reflects the minimum code complexity of the included code unit, more precise than text length. Increase if you notice unwanted, too-small code units in the report.

Details

Ranking thresholds

Similar code units are filtered in two passes, each with its own configurable threshold. The pipeline is as follows:

1. similarity_threshold filters out code unit pairs whose embeddings are not similar enough. The calculated value is cosine similarity ranging from -1 to 1 where 1 means the same.
2. Similar pairs are grouped in clusters.
3. Units in clusters are reranked after applying a boost. Boost is calculated based on the number of directory hops required to reach the other file in the pair (max. 15%). If they are in the same file, the boost is calculated based on distance in number of lines (max. 10%). rerank_threshold filters out clusters whose highest-scoring pair is not high enough.

Exact-copy duplicates

The main goal of this tool is to detect non-exact code duplication, but exact copies (identical code at multiple paths) are reported too, just handled a little differently from merely similar code:

• The report shows the code once, listing every path where it appears, instead of repeating identical snippets.
• The analyze command reports the "similarity ratio" (the share of code units flagged as similar) in two variants: including and excluding exact copies.