Schema & field limits
Validate required artifact structure, permitted values and field lengths before a dataset moves forward.
ReasonForge validates structured reasoning artifacts, detects near-duplicates and leakage patterns, splits train/evaluation data and measures format and pass@1-style quality checks. Run the Python library and CLI on data you control.
Validate required artifact structure, permitted values and field lengths before a dataset moves forward.
Use SimHash-based comparisons to surface repeated or substantially similar records for review.
Detect configured overlap, watermarks and prompt-echo patterns. Review flagged records instead of treating a heuristic as proof.
Use deterministic splitting and cross-split leakage checks for structured JSONL datasets.
Produce baseline evaluation metrics and compare new runs against an agreed reference.
ReasonForge checks structured artifacts against explicit rules. It does not train models, generate reasoning records or automatically repair bad data. A flagged leak or duplicate is a review signal, not independent proof of wrongdoing.
One-time prices in USD. Source use is subject to the published ReasonForge Source-Available License and Commercial License Addendum. Repository access is verified and arranged after accepted purchase terms.
Study and modify the complete licensed implementation for personal, non-commercial use. No employer or client work.
Request Personal license ↗Use ReasonForge in commercial workflows, services and products under one legal entity. No resale of the source itself.
Request Commercial license ↗Unlimited developers and projects within one licensed organization. No license transfer or use by another entity.
Request Organization license ↗Review an agreed reasoning dataset, configure validation/deduplication checks, run a baseline and deliver metrics, recommendations and evidence files.
Request sprint scope ↗Integrate agreed quality checks into an existing CI pipeline, document failure thresholds and hand over configuration, tests and a runbook.
Request pilot scope ↗Services require an agreed dataset format, access permissions, acceptance tests and delivery schedule. Model training, hosted inference, labeling and custom model development are outside these two packages.
Tell us whether you want source access, a configured baseline or a CI integration.