Private SaaS codebase with 186K pull requests and review threads

Summary

The full private codebase of one B2B SaaS company: 640 repositories of Go, TypeScript and Python services with every pull request, inline review comment and linked issue since 2014. The reviews show a senior team arguing about correctness, naming, migrations and rollout risk, which open-source data rarely captures. Secrets, keys, internal hostnames, customer identifiers and author identities are removed or replaced with stable placeholders.

Collected to order. We collect it from US companies for your order. Scope, volume and timing are confirmed in writing before you sign.

Pricing on request

Request scope and pricing

At a glance

Planned scope. Volume, sources and timing are confirmed in writing before you sign.

Records: 186Kpull requests with review threads
Size: 10.5 GB
Tokens, estimated: 2.6B
Years: 2014 to 2026
Firms: 1

Owner profile: US B2B SaaS company in logistics software, 150 to 300 staff
Industry: Software
Data type: Code and engineering
Headcount: 150 to 300 staff
Region: United States
Languages: English
Formats: Source tree, JSONL
Updated: 2026-10-02

Sources

System	Kind	Records
GitHub	Code	640 repositories, 412K commits
GitHub pull requests	Code	186K pull requests, 1.1M review comments
Jira	Projects and wiki	238K issues linked to commits

Good for

training code review models on real reviewer objections and the author's follow-up commits
evaluating agents that write pull request descriptions from a diff and its linked issue
testing long-context retrieval across a private multi-service codebase no model has seen
learning safe database migration and feature flag rollout patterns

Fields

Field	Type	Description
pr_id	string	De-identified pull request id
repo	string	Repository name, replaced if it names a customer
opened_at	timestamp	When the pull request was opened
merged_at	timestamp	Merge time, null if closed unmerged
title	string	Pull request title
description	string	Author's description
linked_issue	string	Linked issue key
files_changed	int	Files touched by the final diff
review_comments	array<object>	Reviewer, file path and body for each review comment, in order
review_outcome	enum	approved, changes_requested or closed
merge_commit	string	Commit hash that maps into the source tree

Example record

{
  "pr_id": "PR-0071644",
  "repo": "shipments-api",
  "opened_at": "2023-08-02T16:41:09Z",
  "merged_at": "2023-08-04T13:12:55Z",
  "title": "Add composite index on shipments (carrier_id, created_at)",
  "description": "Carrier dashboard query does a sequential scan over 180M rows. Adds an index and points the dashboard query at it.",
  "linked_issue": "PLAT-9184",
  "files_changed": 4,
  "review_comments": [
    {
      "reviewer": "[PERSON_7]",
      "path": "migrations/0142_add_carrier_idx.sql",
      "body": "Needs CREATE INDEX CONCURRENTLY or it locks shipments for the whole build. It also can't run inside the migration transaction."
    },
    {
      "reviewer": "[PERSON_22]",
      "path": "store/shipments.go",
      "body": "Query still orders by id, so the planner may skip the new index. Order by created_at to match."
    }
  ],
  "review_outcome": "approved",
  "merge_commit": "9f3c1e7a"
}

One record, de-identified. Placeholders such as [PERSON_12] and [COMPANY_4] stay the same across the whole dataset, so threads still read correctly.

Licenses available

Exclusive
One buyer. Nobody else can license it during the exclusive term, usually 2 years.
Non-exclusive
Licensed to more than one buyer.
First look
One buyer gets 2 to 4 weeks to make an offer before anyone else sees the dataset.

What ships with it

Datasheet: what it is, what it is good for, verified size, sources, years, format and limits.
Provenance file: the owner shown by code, staff notice and consent status, exclusions, mapped to the EU AI Act Article 53 training-data summary and California AB 2013.
Manifest with a hash for every file.
QC report and de-identification report.
Your own copy, with unique canary records.

How it is de-identified

Default exclusions removed, files flattened to text, identifiers found in three layers and replaced with consistent placeholders, and the result measured against a 99% bar for each identifier type in each source.

The full method