HTTP route reference
Every HTTP route the Runink TIDE console registers. All routes are served by the TIDE operator process on the console’s own origin. There is no separate API host and no bearer-token API for browsers: the web app authenticates with the session cookie.
The per-category references give the request and response detail and the exact authorisation of each route:
How routes are authorised
| Class | Rule | Examples |
|---|---|---|
| Open | no session needed | /healthz, /api/me, /api/plans |
| Session read | a GET needs a signed-in session when sign-in is configured (requireSession); on an open console it is served | /api/dashboard, /api/namespaces, /api/gitops |
| Privileged write | a session and the adminRefusal predicate against an allowlist env var; audited on success, failure and refusal | PUT /api/agent-schedules/{name} (CONSOLE_AGENT_ADMINS), POST /api/connections (TIDE_CONNECTION_ADMINS) |
| Machine ingest | a shared token generated once in tide-system; unset means the door refuses everything | POST /api/agent-health (TIDE_HEALTH_INGEST_TOKEN), POST /api/rules-recon/report (TIDE_RECON_INGEST_TOKEN), POST /api/judgement/submissions (TIDE_JUDGEMENT_INGEST_TOKEN) |
| Agent report | a GitHub App installation token for the console’s org (authorizeReporter) | POST /api/session-runs, POST /api/data-governance-findings, POST /api/judgements, POST /api/playbook-runs/step-report |
An allowlist that is unset means “every identity this console admits”. That is not a wildcard,
because sign-in itself fails closed. CONSOLE_SESSION_ADMINS is the exception: unset means
nobody. GET /api/access lists every allowlist. See
Agents › Access.
Reads that could not reach their source still answer 200, with source: "unavailable" and
complete: false in the body. See Status and provenance.
Sign-in and sessions
| Route | Methods | Purpose |
|---|---|---|
/healthz | GET | liveness: answers ok |
/auth/callback | POST | Google Identity Services credential, with the double-submit g_csrf_token |
/auth/google | POST | the Flutter app’s Google token exchange (JSON) |
/auth/password | POST | username and password sign-in, when CONSOLE_ADMIN_PASSWORD is set |
/auth/logout | any | revokes the server-side session |
/api/me | GET | sign-in state and client config: authed, email, authRequired, passwordAuth, googleAuth, googleClientId, recaptchaSiteKey, voiceTts, voiceStt |
/api/sessions/revoke-all | POST | end every session of a subject |
/api/session-admins | GET, PUT | who may read the audit trail and end other people’s sessions |
Audit and access
| Route | Methods | Purpose |
|---|---|---|
/api/audit | GET | newest-first page of the audit chain and its verdict |
/api/audit/{seq} | GET | one record in full, with its neighbours’ hashes |
/api/audit/verify | GET | walk the persisted chain and name the first break |
/api/access | GET | which allowlists are live, their sizes, and your own membership only |
Reading the audit trail requires a session admin (CONSOLE_SESSION_ADMINS).
Chat and voice
| Route | Methods | Purpose |
|---|---|---|
/chat | POST | the orchestration chat: one tool call per message; mutating calls need confirm <code> |
/api/tts | POST | sovereign text-to-speech through TTSD_URL |
/api/stt | POST | sovereign speech-to-text through VOICE_REMOTE_URL |
Platform and cluster (DevEx)
| Route | Methods | Purpose |
|---|---|---|
/api/dashboard | GET | tracked workloads and client instances |
/api/control-planes | GET | FACE and PULSE control-plane CRDs |
/api/cluster-nodes | GET | Kubernetes Nodes and dedicated-taint isolation |
/api/namespaces | GET | per-namespace workloads, pods, events, quotas and a verdict |
/api/nodes/{name} | GET | one node: conditions, capacity, taints, pods by namespace |
/api/services | GET | serving status: ready/desired, and whether a readiness probe backs it |
/api/routes | GET | edge routes and /forge/, traced to Service, endpoints and workload |
/api/app-consoles | GET | DevEx › Application consoles: one card per tool the viewer may reach |
/api/instances/overview | GET | the Overview’s bird’s-eye: every instance and control plane the viewer may see |
/api/instances/open | GET | an instance card’s Open: a member-only entry ticket to that instance |
/api/ci-runners | GET | Agent runs › CI runners |
/api/agent-models | GET | each agent’s model chain, resolved from what it sends: token, router tier, backend, adapter, card, server |
/api/hunter-runs | GET, POST | the backlog hunter’s runs (session read; the POST is the hunter’s token-authenticated ingest) |
/api/knowledge/search | GET | Business rules › Knowledge: search an instance’s corpus |
/api/knowledge/upload, /api/knowledge/ingest | POST | add uploaded files, or a document connection’s text, to an instance’s corpus (admins) |
/api/gitops | GET | Clusters › GitOps: per workload, the desired version from the tide-cd.yml lineage record against what is running, and the drift |
/api/lineage | GET | per image: merged → agents → built → registry → rolled → running; ?sha= answers “did my commit ship” |
/api/doctor | GET | the ops doctor CronJob: state, runs, tracking issue |
/api/metrics | GET | measured process and workload metrics |
/api/devex | GET | DevEx indicators from the persisted run-outcome series |
/api/security | GET | mesh CA fingerprint and expiry per namespace |
Agents and runs
| Route | Methods | Purpose |
|---|---|---|
/api/agents | GET | the agent roster and the inference tiers |
/api/agent-config | GET | the app agents’ settings and OpenBias rules, read from each repo’s HEAD |
/api/agent-runs | GET | PR-triggered agent runs per repo (GitHub Actions) |
/api/agent-runs/logs | GET | one run’s unzipped log text |
/api/runs | GET | every agent run across kinds, repos, CronJobs and app agents |
/api/reviews | GET | running and past reviews, with findings per review type |
/api/reviews/{type}/{runId}[/detail] | GET | one review run’s findings; /detail adds what the agents posted on GitHub |
/api/agent-schedules | GET | each agent’s cadence and arming, shown in the Schedule & runs section of its page |
/api/agent-schedules/{name} | PUT | arm or disarm an agent (privileged) |
/api/agent-health | POST | the apps’ rolled-up agent health (machine ingest) |
/api/session-runs | POST | tide session run reports (agent report) |
/api/session-commands | POST, GET | the remote-control queue a workstation pulls from: enqueue for yourself, or long-poll a claim |
/api/session-commands/{id}/result | POST | a claimed command’s result |
/api/session-queue | GET, POST | the browser’s view of your queue, and the browser’s enqueue path (cookie) |
/api/governance-runs | GET | run history of the governance-family agents; ?repo= filters |
Models and inference (DataEx)
| Route | Methods | Purpose |
|---|---|---|
/api/inference | GET | the inference plane as deployed: model, quantisation, context, engine, OOM history |
/api/inference/hosts | GET | DataEx › Agents › Models & inference: each host and the tiers it serves, resolved live from the model router, servefwd and the engines; a card only where it describes the loaded model; usage from the router since it started |
/api/models | GET | model cards joined with live tier state, a health verdict, token use and tenant CU |
/api/model-readiness | GET | whether the models the AI features call (general, guard and reranker tiers) are ready, for the banner over every page: the model router’s /health and its ops doctor’s engine health. Never calls or wakes a model |
/api/dataex/tasks | GET | Read by Ask TIDE’s dataex_tasks lookup (the DataEx › Tasks page was removed 2026-11-05): each tier’s measured decode rate and the model router’s per-backend queues. Each part carries its own source, reason and age |
/api/servers | GET | Clusters › Nodes & servers: the k0s nodes and the declared off-cluster servers (ConfigMap tide-system/tide-servers), each read through servefwd, with the router tiers it serves |
/api/training-runs | GET | DataEx › Tasks › Training: every rtrain run on each declared training server (progress, losses, ETA, adapter, base + revision), read from the server’s gate through servefwd |
/api/training-config | GET | DataEx › Tasks › Training › Configuration: adapters.yaml + serve.yaml from the training server’s configRepo at its default-branch head, read-only, with what an edit path needs |
/api/finops | GET | allocated compute attributed by initiative and tenant |
Connections, runners and providers (DataEx)
| Route | Methods | Purpose |
|---|---|---|
/api/connections | GET, POST | list or create tenant data-source connections |
/api/connections/{id} | GET, PUT, DELETE | one connection |
/api/connection-types | GET | the connector catalog: settings and credential keys per type |
/api/runners | GET, POST | data-access runners: list, or enrol (a one-time ticket is shown once) |
/api/runners/{id}/… | POST, PATCH | token, connect, revoke, and address or label changes |
/api/runner-tokens/{id} | DELETE | revoke an unused enrolment token |
/api/connectors/status | GET | the cached connector probe; never dials |
/api/connectors/status/refresh | POST | run the connector probe (the one path here that dials; privileged) |
/api/providers | GET, POST | cloud-provider config and the connectors TIDE publishes |
/api/providers/gcp/status | GET | the live GCP fleet link (opt-in) |
/api/github | GET, POST | the GitHub configuration shown on Admin › Platform settings |
The DataEx Connections and Runners pages now use the shared gRPC-web services; the /api/connections*,
/api/runners* and /api/runner-tokens/* routes stay for one release as deprecated aliases (the
tide connection register CLI uses /api/connections).
Trust (DataEx)
| Route | Methods | Purpose |
|---|---|---|
/api/harness | GET | measures and compliance joined into actionable findings |
/api/harness/{id}/act | POST | act on a finding: admin, audit-before-act, idempotent |
/api/guardrails | GET, PUT | autonomy per action class, and the hard guardrails |
/api/measures | GET | the platform self-assessment: judged categories and one remediation measure per failing one |
Intelligence
| Route | Methods | Purpose |
|---|---|---|
/api/swarm | GET | agent swarm: fleet health, roster, run and latency rollup across apps |
/api/domains | GET | data domains clustered from the connection registry |
/api/governance | GET | estate governance: ownership, credential storage, namespace posture |
/api/data-lineage | GET | declared topology and observed flows |
/api/data-lineage/edges | POST | the apps’ observed-lineage ingest (machine token) |
/api/compliance | GET | per-control findings; never a score |
/api/maturity | GET | five measured dimensions, and a composite only when all five are measured |
/api/rules-recon | GET | business-rule reconciliation; ?tenant= |
/api/rules-recon/inputs | GET | the recon agent’s read door (TIDE_RECON_INGEST_TOKEN) |
/api/rules-recon/report | POST | the recon agent’s report (TIDE_RECON_INGEST_TOKEN) |
/api/data-governance-findings | GET, POST | the datagov findings: session read, agent-report write |
/api/data-governance-estate | GET | the datagov agent’s read door: structure and counts only (agent report token) |
/api/judgement | GET | submitted findings and TIDE’s verdicts on them |
/api/judgement/submissions | POST | an external platform’s findings (TIDE_JUDGEMENT_INGEST_TOKEN); any verdict on the wire is discarded |
/api/judgements | GET, POST | the judge agent’s verdicts (agent report token for POST) |
/api/playbook-runs/step-report | POST | a dispatched agent reports its playbook step |
/api/playbook-events | POST | an agent raises a playbook event |
FORGE
| Route | Methods | Purpose |
|---|---|---|
/api/forge-studio | GET | what FORGE › Studio can do for the caller: GitHub writes (FORGE_OWNER_EMAIL), the GitHub App, the planner |
/api/forge-delete | GET, POST | GET: what deleting a FORGE app removes; POST: delete it after the typed-name confirmation |
Billing
| Route | Methods | Purpose |
|---|---|---|
/api/plans | GET | the public plan catalogue of one product, ?product=face|pulse|core (empty is FACE). See Billing |
/api/signup | POST | public self-serve signup: a pending subscription and a checkout |
/api/checkout | POST | a quote and, when configured, a Stripe checkout for the signed-in person |
/api/billing/seats | GET, POST, DELETE | who holds the seats on the signed-in purchaser’s subscription |
/api/stripe/webhook | POST | Stripe events into the subscription store |
gRPC-web on /
The root handler serves, in order: gRPC-web (dispatched by content type, so it needs no route of
its own), the Flutter bundle from CONSOLE_WEB_DIR, and else the HTML dashboard fallback. The
gRPC-web services run in the same process, on the same origin and session, with no CORS: a foreign
Origin is refused with 403. They are:
- TIDE’s own:
CapexService,GovernanceService,WorkspaceService,PlaybookService,ResolveService(an alias of the estate service) andAskService; - the shared services TIDE mounts:
runink.ui.estate.v1.EstateService,runink.ui.profile.v1.ProfileService, and the datasourcesConnectionsServiceandRunnersService.