Mentorist.app full source code

Data Product Features

• Full Git history: every commit with its parent commits, date, message and diff, so you can trace how the code evolved from first commit to the final release.
• Four platforms, one product: client, server and serverless code for the same business domain. You can trace a feature end to end, for example from the mobile UI through API endpoints to the database and background jobs.
• Production-grade domain logic:
  • user accounts and authentication;
  • in-app subscriptions (App Store, Google Play, Stripe, RevenueCat);
  • push notifications;
  • offline audio playback;
  • a content CMS;
  • content-processing pipelines;
  • analytics events;
  • localisation;
  • an A/B-tested onboarding funnel.
• Release engineering: release branches and version tags (151 on Android, 82 on iOS), fastlane and Gradle build configuration, CI scripts and database migrations.
• Sanitised for safe use:
  • all credentials, API keys, private keys and tokens removed from every revision of the history;
  • personal and customer data removed;
  • contributors anonymised as "Contributor N".
• Verified integrity: after cleaning, the code still parses and compiles exactly like the originals. We checked this with the Swift and Interface Builder compilers, aapt2, the Python AST, the TypeScript compiler and ESLint.

Distribution

• Data Volume: 4 repositories, about 190,000 lines of code (excluding XML layouts and storyboards), about 15,400 commits.
  • Backend (Django/Python): about 104k LOC, 5,473 commits.
  • iOS (Swift): about 43k LOC, 6,884 commits, 5 branches, 82 tags.
  • Android (Kotlin/Java): about 26k LOC, 2,601 commits, 11 branches, 151 tags.
  • Firebase Functions and web onboarding (TypeScript/React): about 17k LOC, 453 commits.

Usage

• Code LLM pre-training and fine-tuning: high-quality, human-written, multi-language production code from a real commercial product, not tutorials or toy projects.
• Commit-message and code-change modelling: pairs of diffs and commit messages for training models to generate, summarise or review code changes.
• Software-evolution and refactoring research: nine years of history covering framework upgrades (for example Objective-C-era patterns to Swift, Java to Kotlin), Django version migrations and changing architecture.
• Cross-platform and full-stack reasoning: the same features implemented on iOS, Android, web and server, useful for code translation and for understanding systems that span several repositories.
• Benchmarking AI coding agents: realistic repositories for evaluating bug localisation, test generation, dependency upgrades and feature implementation.

Coverage

• Geographic Coverage: Global. The product served an international, mainly English-speaking audience. The code was written by a team based in Europe.
• Time Range: November 2016 to June 2025 (Android 2016–2025, iOS 2017–2025, backend 2016 to June 2025, Firebase Functions 2021 to June 2025).
• Demographics (if applicable): Not applicable. The dataset contains source code only, with no end-user data. Domain: consumer EdTech, micro-learning and subscription mobile apps.

License

AI Training Rights

• Use the Data Product to train, fine-tune, and evaluate machine learning models, including large language models.
• Incorporate Data Product content into models and commercialize resulting model outputs.
• Create derivative works (model weights, embeddings, etc.) for any lawful purpose.

• The Data Product itself may not be sold, redistributed, or shared outside of licensed usage.
• Licensee must comply with all applicable laws, including data protection and privacy regulations.

Who Can Use It

• AI labs and model builders: for pre-training and fine-tuning code models on authentic production code with full history.
• AI coding-tool companies: for building benchmarks and evaluation tasks for coding agents on real-world, multi-repository projects.
• Researchers: for empirical software-engineering studies of code evolution, technical debt, refactoring and mobile development practice.
• Businesses: for internal code-intelligence tools, code search, and code review or migration assistants.

Data Dictionary

| Column Name   | Data Type | Description                                         | Possible Values/Notes |
|---------------|-----------|-----------------------------------------------------|-----------------------|
| repository    | string    | Source repository                                   | mentorist-backend, mentorist-ios, mentorist-android, mentorist-firebase-functions |
| commit_sha    | string    | Unique commit identifier                            | 40-char hex SHA-1 |
| parent_shas   | list      | Parent commit(s)                                    | 2+ entries for merge commits |
| author        | string    | Anonymised commit author                            | "Contributor N", consistent across all repositories |
| author_date   | datetime  | Commit timestamp                                    | ISO 8601, 2016-11 to 2025-06 |
| message       | string    | Original commit message                             | Natural language, mostly English; names and secrets redacted |
| branches_tags | list      | Branches and release tags containing the commit     | e.g. develop, master, release versions |
| file_path     | string    | Path of the changed file                            | Original project structure |
| language      | string    | Programming language of the file                    | Swift, Kotlin, Java, Python, TypeScript, JavaScript, HTML/CSS, XML, other |
| change_type   | string    | Type of change                                      | added, modified, deleted, renamed |
| diff          | string    | Unified diff of the change                          | Standard Git patch format |
| content       | string    | Full file content at this revision                  | Secrets replaced with ***REMOVED*** |

• Sanitisation: the whole history was rewritten to remove secrets and personal data, then independently audited with gitleaks, trufflehog and custom scanners. Public client identifiers that are required to build the code and are not secrets are kept (for example the Firebase client config with keys removed, and the Stripe publishable key).
• Third-party code: all vendored third-party and open-source code (libraries, frameworks, fonts) has been removed from the entire history, so the dataset contains only code owned by Taulab OÜ. Dependencies are referenced by name in package manifests.
• Not included: binaries (APK/IPA/AAB), signing keys, environment files and production data. Secrets in the code are replaced with the placeholder REMOVED, so the code builds but does not connect to live services.
• Sample: a representative code sample and a code-survey report (LOC, commits, languages) are available on request for evaluation before purchase.