Repository files navigation

An experimental library for translating source code to different programming languages using Ollama.

This library enables translation of source code files from one programming language to another using Ollama AI models capabilities. It's an experimental tool that can be helpful for quick code conversion between different languages.

• Rust installed on your system
• Ollama installed and running locally
• A compatible model (default: qwen2.5-coder)

cargo add aleph_ollama

The library can be used via command line through the provided example:

cargo run --example basic<source_file_path><target_language>

cargo r --example basic test/dataset/syracuse.ml java

Which produces the following output:

importjava.util.ArrayList;importjava.util.List;publicclassSyracuse {publicstaticvoidmain(String[]args) {intn =7;List<Integer>suite =syracuse(n);System.out.print("Syracuse for " +n +" : ");for (intx :suite) {System.out.print(x +" ");        }System.out.println();    }publicstaticList<Integer>syracuse(intn) {if (n <=0) {thrownewIllegalArgumentException("Le nombre doit être strictement positif");        }elseif (n ==1) {List<Integer>list =newArrayList<>();list.add(1);returnlist;        }elseif (n %2 ==0) {List<Integer>list =syracuse(n /2);list.add(0,n);returnlist;        }else {List<Integer>list =syracuse(3 *n +1);list.add(0,n);returnlist;        }    }}

The library can be configured through environment variables:

• OLLAMA_MODEL: The model to use for translation (default: "qwen2.5-coder")

• OLLAMA_URL: Ollama endpoint (default "http://localhost:11434/api/generate")

• Translation quality depends on the model used
• Some language-specific concepts may not have direct equivalents in the target language
• The tool is experimental and may require manual adjustments of generated code

Contributions are welcome! Feel free to open an issue or submit a pull request.