Public API
Read-only HTTP endpoints for campaigns, submissions, rounds, the leader, and the weight vector.
The API is read-only and needs no key. The base URL is
https://api.pareton.ai. An OpenAPI schema is served at
https://api.pareton.ai/openapi.json.
Use it to track a submission without watching the dashboard.
Campaigns
| Endpoint | Returns |
|---|---|
GET /v1/campaigns | Every campaign |
GET /v1/campaigns/{campaign_id} | One campaign and what it pins |
GET /v1/stats | Subnet-wide counters |
Submissions
| Endpoint | Returns |
|---|---|
GET /v1/campaigns/{campaign_id}/submissions | Submissions in one campaign |
GET /v1/campaigns/{campaign_id}/submissions/{patch_hash} | One submission, with its events |
GET /v1/campaigns/{campaign_id}/submissions/{patch_hash}/build-log | Sanitized build log tail |
Prefer the campaign-scoped path. A submission is identified by campaign plus
patch hash together, so the short forms GET /v1/submissions/{patch_hash} and
its build-log path return 409 when the same bytes were submitted to more
than one campaign.
A submission detail carries a round object once it is seated: round id,
ordinal, entry status, score, and disqualify reason. The entry status is the
outcome. There is no separate verdict field.
Rounds and the leader
| Endpoint | Returns |
|---|---|
GET /v1/campaigns/{campaign_id}/rounds | Rounds, newest ordinal first, paginated |
GET /v1/rounds/{round_id} | One round, every entry, and the live phase |
GET /v1/rounds/{id}/entries/{entry_id}/report | One entry's per-prompt score breakdown |
GET /v1/campaigns/{campaign_id}/leader | The seated leader |
GET /v1/campaigns/{campaign_id}/score-progress | One point per ordinal, for charting |
GET /v1/campaigns/{campaign_id}/leader returns 404 when the crown is
vacant. A vacant crown has no row, so this is the expected answer, not an error.
The report endpoint is how you check a score rather than take it on trust. It
returns each prompt's speedup, the token count both engines were compared at,
both absolute end-to-end times, and the gate reason for any prompt that scored
nothing. prompt_summary counts how many prompts scored and how many the
tolerance gate zeroed. The baseline entry stores its SLA replay rather than a
comparison against itself, so it returns prompts: [] with sla populated.
An entry that never reached scoring returns prompts: [] and its reason.
Weights
GET /v1/weights returns the newest stored weight vector and a per-campaign
breakdown. See Validators and emissions.
Caching
A response that contains a pending or running round is sent no-store.
complete and void rounds are terminal and cacheable. Poll live rounds; cache
settled ones.
What the API never returns
Evidence stays behind its gate: evidence_s3_url is never selected, on any
endpoint. The per-entry report is served, but only through the report endpoint
above, so the round detail stays small enough to poll.
Identity is the hotkey. List responses truncate it to 16 characters. Detail responses carry it in full.