Legacy Codebase Repository Dataset - 675 Codebases, 262M Lines of Code, 725K commits, 19 Industries | Codebase Data
Overview
Legacy Codebase Repository Dataset - 675 Codebases, 262M Lines of Code, 725K commits, 19 Industries | Codebase Data is a text dataset. Typical applications include fine-tuning & instruction data, alignment & RLHF and code generation. Relevant sectors: AI & Machine Learning and Education. Pricing is not published and is provided on request.
Primary applications
Dataset Description This dataset is a large-scale collection of legacy software systems and enterprise-grade source code designed to support the development of AI systems focused on software modernization, code migration, refactoring, and legacy infrastructure transformation. The dataset includes historical code repositories, monolithic applications, legacy architectures, enterprise software workflows, and production-grade source code across multiple programming languages and technology stacks, including C, C++, Java, COBOL, SQL, Python, PHP, and more. It captures authentic software engineering patterns, deprecated frameworks, technical debt structures, migration workflows, and real-world maintenance scenarios commonly found in enterprise environments. The dataset is highly valuable for building AI coding assistants, modernization agents, automated refactoring systems, and enterprise software transformation pipelines. It also supports workflows for Supervised Fine-Tuning (SFT), Reinforcement Learning with Human Feedback (RLHF), code reasoning, debugging, and repository-level AI understanding. Key Use Cases -Legacy code modernization -Automated code refactoring -Software migration and transformation -Technical debt analysis -Repository-level code understanding - Code debugging and maintenance - Enterprise software automation -AI coding assistants and copilots - Mainframe and legacy infrastructure migration -Documentation and code summarization Dataset Specification -Type: Lega
Related datasets
More in Code & Software →Agentic Coding Trajectory Data | Verbatim Harness Sessions | Claude Code, OpenClaw, OpenCode, Kilo | JSONL
Most agent datasets get cleaned up before they reach the buyer.
- Language
- —
- Volume
- —
- Domain
- AI & Machine Learning
- Use case
- Fine-tuning & instruction data
More text & language datasets
All 776 text & language datasets →- Legal Consultation Text Dataset for Cybersecurity Analysis (Australia, English)
- Legal Consultation Text Dataset for Healthcare Analysis (Germany, German)
- Legal Consultation Text Dataset for Healthcare Analysis (Singapore, English)
- Legal Consultation Text Dataset for Logistics Analysis (France, French)
- Levantine Arabic SMS Communication Dataset
- List of UK Fund Managers (Hedge Funds, Private Capital & Real Assets)
- Lithuanian Speech-to-Text Transcribed Audio Dataset
- LLM Infrastructure Telemetry Dataset | NVIDIA P100 + Dual Xeon | 10K+ Labeled Rows
- Loan Application Call Text Dataset for IT Analysis (United Arab Emirates, Arabic)