API reference

The ComputeSDK Platform REST API surface — every /api/v1 endpoint for Actions, the vault, benchmarks, organizations, and the org feed.

Base URL: https://platform.computesdk.com/api/v1. Authenticate every request with Authorization: Bearer <org API key> (see Authentication). Errors are { "error": "<message>" } with a 400/401/403/404/410/429/500 status.

Identity and organizations

Method	Path	Purpose
GET	/me	Authenticated user + their organizations
GET	/organizations	List organizations
GET/PATCH/DELETE	/organizations/{orgId}	Organization detail, update, delete
GET/POST	/organizations/{orgId}/api-keys	List / create org API keys (owner/admin)
DELETE	/organizations/{orgId}/api-keys/{keyId}	Revoke a key
GET	/organizations/{orgId}/logo	Organization logo
GET	/organizations/{orgId}/benchmark-artifacts/{artifactId}	Org-scoped benchmark artifact download
GET	/feed	The org's activity feed

Actions (CI)

Actions is available to every organization — no product entitlement required. See Actions for semantics.

Method	Path	Purpose
GET	/actions/org	The org's Actions context
GET/POST/PATCH	/actions/repos	Connected repos — list, connect a git remote, enable/disable (GitHub App + generic remotes)
GET	/actions/providers	Registered providers — regions, act/usable status
PUT/DELETE	/actions/providers/{provider}/key	Save / remove the org's provider credential
POST	/actions/providers/{provider}/verify	Probe the stored key with a real sandbox
GET	/actions/workflows?repo=	Workflows on enabled repos (dispatchable: false ones run via manual: true)
POST	/actions/dispatch	{ workflowId, ref, inputs?, manual?, requestId?, provider?, providerRegion? } → { runId, created, headSha }
GET	/actions/history?repo=&workflow=	Recent-run window + per-job/step failure rollups
GET	/actions/run-days?timeZone=&repo=&branch=	Days with runs
GET	/actions/run-states?repo=&branch=	Live states for the current run list
GET	/actions/runs	Run list
GET	/actions/runs/day/{YYYY-MM-DD}	Page of runs for a day (cursor via nextCursor)
GET	/actions/runs/{runId}	Run detail
GET	/actions/runs/{runId}/state	Conclusion, job placement, step states, effective context
GET	/actions/runs/{runId}/summary	Failure digest with redacted log tails
GET	/actions/runs/{runId}/stream	SSE: run state + log bytes (?watch=id:jobId:step:offset to resume)
POST	/actions/runs/{runId}/cancel	Cancel → { cancelled }
POST	/actions/runs/{runId}/rerun	Rerun same commit → { runId, created }
GET	/actions/jobs/{jobId}/logs	CiLogSlice (segments, nextOffset, state, totalBytes); ?follow=1 for SSE
GET	/actions/jobs/{jobId}/logs/download	Whole log as text/plain
GET	/actions/jobs/{jobId}/artifacts	Job's artifacts
GET	/actions/jobs/{jobId}/artifacts/{artifactId}	Redirect to signed download URL (410 when expired)

Vault (secrets and variables)

Owner/admin keys only. Not gated on the Actions entitlement. repo=<owner>/<name> scopes to a repo's overrides; kind is secret (default) or variable.

Method	Path	Purpose
GET	/vault?repo=&kind=	List items — names and metadata, never values (with repo, includes inherited org items)
PUT	/vault?repo=	{ name, kind?, value, description?, revealable?, labels?, expectedVersion? } — value stored exactly as sent; revealable/labels fixed at creation
DELETE	/vault?repo=&kind=&name=	Delete an item
POST	/vault/reveal?repo=	{ name, kind? } → { name, kind, value } — variables and revealable secrets only (403 otherwise, 404 when absent)

Benchmarks

Org keys see their own benchmarks plus subscribed (entitled) ones — never other orgs' private definitions. Foreign-but-entitled benchmarks are addressed via /benchmarks/~/{sourceOrgSlug}/{benchmarkSlug}.

Method	Path	Purpose
GET	/benchmarks	List visible benchmarks (limit/offset, max 500)
GET/PUT/PATCH	/benchmarks/{slug}	Benchmark detail, create/update definition
GET	/benchmarks/{slug}/results	Latest published results
GET	/benchmarks/{slug}/branding	Benchmark branding
GET	/benchmarks/{slug}/model-index	Model index
GET	/benchmarks/{slug}/runs	Run list; /runs/days, /runs/day/{day}, /runs/summary for rollups
GET/POST	/benchmarks/{slug}/runs	List / create runs
GET	/benchmarks/{slug}/runs/{runId}	Run detail
GET	/benchmarks/{slug}/runs/{runId}/summary	Run summary
GET	/benchmarks/{slug}/runs/{runId}/progress	Live progress
GET	/benchmarks/{slug}/runs/{runId}/iterations	Per-iteration data
GET	/benchmarks/{slug}/runs/{runId}/artifacts	Run artifacts
GET/POST	/benchmarks/{slug}/runs/{runId}/results	Aggregated results
GET	.../results/timeline, .../results/tasks, .../results/imports	Result provenance
GET/POST	/benchmarks/{slug}/runs/{runId}/participants	Participants under test
GET	.../participants/{slug} · .../participants/{slug}/logs · .../participants/{slug}/workers · .../workers/claim	Participant state, logs, and worker fleet
GET/POST	.../runs/{runId}/workers/{workerId} · /events · /heartbeat · /complete · /fail · /release · /artifacts	Worker lifecycle — for benchmark executors