3 Code Datasets for Training Coding Agents | Dataset Spotlight by Defined.ai EP.10

This episode of Dataset Spotlight covers three datasets for teams building coding agents: git history, linked Jira projects, and agent traces with tool use. When an agent breaks a build or edits the wrong file, someone on the team has to clean it up. Here's what's inside: ✅ Git Repositories with Version History 367 complete git repositories, including the full version history wherever it's available: commits, contributors, branches and more. Built for models that edit and fix code. ✅ Code Repositories with Linked Jira Projects 501 code repositories, each accompanied by its linked Jira project, across domains like backend services, automation, e-commerce and healthcare. Built for agents that need project context. ✅ Code-Agent Traces with Tool Use 1 billion tokens of English code-agent traces: queries sent to LLMs to act on users' codebases, with the tool traces and model actions included. Built for training tool-using coding agents. All three are on the Defined.ai Marketplace. 👉 Get in touch for free samples: https://defined.ai/contact-us?utm_source=youtube&utm_medium=organic-social&utm_campaign=dataset-spotlight&utm_content=video&utm_term=ep-10 New Dataset Spotlight every two weeks. Subscribe to catch the next one. #TrainingData #AIAgents #MachineLearning