moraMora / Dataset standards
Dataset standards
Data arrives in the same shape from every company, so a lab can compare and combine it. Each standard lists the parts a dataset can hold. Few companies have them all: a listing says which parts it holds, and a lab pays for those.
Parts, not one blob
Code, commits, reviews, issues: each part is its own table with fixed field names.
Missing is fine, hidden is not
A solo developer has no Jira and no Slack. The listing says so, and the price follows.
Measured before the offer
The provider's agent measures each part (do the tests run?). Mora checks it on the sample.
Agents speak it
A provider's agent fits data to a standard. A lab's agent asks for parts by name, or sends its own schema.
Software project
Training and testing coding agents on real work: a change, why it was made, and whether it held.
| Part | Needed | Holds | Fields | Measured |
|---|---|---|---|---|
| Code | Required | The repository at each commit | repo_id, path, language, content, commit_sha | Languages, Lines of code, Has tests, Builds |
| Commit history | Required | Every change with its message and diff | commit_sha, author_id, authored_at, message, diff | Commits, Years of history, Contributors |
| Pull requests and reviews | Optional | Proposed changes, review comments, and whether they merged | pr_id, title, description, author_id, review_comments, merged, merged_at | Pull requests, Reviewed |
| Issues | Optional | Bug reports and tasks from Jira, Linear or GitHub, linked to the change that closed them | issue_id, title, description, type, status, created_at, resolved_at, fixed_by_commit | Issues, Linked to a fix |
| Test and build runs | Optional | Each CI run with its result and logs | run_id, commit_sha, status, failed_tests, log_excerpt, duration_s | Runs, Failed runs |
| Team discussion | Optional | Slack or Teams threads about the work | thread_id, author_id, posted_at, text, refers_to | Threads, Linked to work |
Support conversations
Support agents and their evaluation: a customer problem and how it was really solved.
| Part | Needed | Holds | Fields | Measured |
|---|---|---|---|---|
| Tickets | Required | Each conversation from first message to last | ticket_id, customer_id, opened_at, channel, messages, status | Tickets, Messages per ticket, Languages |
| Resolution | Optional | What fixed it, and whether the customer came back | ticket_id, resolution, resolved_at, reopened, escalated_to | Resolved, Reopened |
| Customer rating | Optional | The score the customer gave | ticket_id, score, comment | Rated |
| Help articles | Optional | The articles agents answered from | article_id, title, body, updated_at | Articles |
Field service records
Diagnosis and repair agents: a fault, what the technician found, and whether the fix held.
| Part | Needed | Holds | Fields | Measured |
|---|---|---|---|---|
| Work orders | Required | Each job: complaint, findings, what was done | order_id, site_id, equipment_model, complaint, diagnosis, work_done, visited_at | Work orders, Years covered, With a diagnosis |
| Outcome | Optional | Whether the customer called back for the same fault | order_id, callback_within_30_days, repeat_of | Outcome known |
| Sensor readings | Optional | Equipment readings before the visit | site_id, read_at, sensor, value, unit | Jobs with readings |
| Parts used | Optional | What was replaced | order_id, part, quantity | Jobs with parts |
Speech audio
Speech recognition in hard conditions: noise, jargon, people talking over each other.
| Part | Needed | Holds | Fields | Measured |
|---|---|---|---|---|
| Audio | Required | The recordings | clip_id, audio, duration_s, sample_rate, recorded_at, source_id | Hours, Signal quality |
| Transcripts | Optional | What was said, by whom it was checked | clip_id, text, transcribed_by, reviewed | Transcribed, Reviewed by a person |
| Speakers and timing | Optional | Who speaks when | clip_id, speaker_id, start_s, end_s | With speaker turns |