Gemini 3.1 Pro thinking model requires thought_signature to be echoed
in functionResponse. SSE stream doesn't reliably include it in individual
chunks. Switch tool-calling rounds to non-streaming generateContent which
always returns the complete response with thought_signature present.
Made-with: Cursor
Thinking models attach a thought_signature to functionCall parts.
Must be echoed back in functionResponse or API returns 400.
Carry it through ToolCall -> ChatMessage -> toGeminiContents().
Made-with: Cursor
URL param 'mark-account' != workspace slug 'mark'. Fetch default token
from /api/workspaces?include_default_token=true which resolves the real
slug server-side.
Made-with: Cursor
- ensureWorkspaceForUser() now calls mintWorkspaceApiKey('default') on first workspace creation
- Legacy workspaces without a default key get one minted on first request
- GET /api/workspaces/[slug]/keys/default reveals (or mints) the default token for session users
- Chat panel fetches the token automatically on mount, caches it in localStorage
- No manual setup needed — tool calling works immediately on first sign-in
Made-with: Cursor
- Right-docked chat panel on all workspace pages ([workspace]/layout.tsx)
- Streaming SSE responses with Gemini 3.1 Pro preview via generativelanguage API
- Full tool-calling loop (up to 6 rounds): deploys apps, lists projects, registers
domains, fetches logs — all via existing MCP dispatcher
- Persistent conversation history: fs_chat_threads + fs_chat_messages tables (Postgres)
- Thread management: create, list, rename (auto-title from first message), delete
- Panel collapses to a tab; open state persisted to localStorage
- Read-only mode hint when no MCP token is present
- Graceful content margin shift when panel is open
Made-with: Cursor
The Coolify UI shows a "Required Port: 3000 — All domains must
include this port number" hint on service templates. That hint is
load-bearing: when the URL passed to `setServiceDomains` includes
:<upstream_port>, Coolify's template engine auto-generates everything
that 2.4.5-2.4.7 were doing by hand:
- traefik.http.services.<svc>.loadbalancer.server.port label
- SERVICE_FQDN_<APP>=<fqdn> (no sslip.io leak)
- SERVICE_URL_<APP>=https://<fqdn>
- SERVICE_FQDN_<APP>_<PORT>=<fqdn>:<port>
- SERVICE_URL_<APP>_<PORT>=https://<fqdn>:<port>
Verified end-to-end with twenty:
setServiceDomains(uuid, [{ name:'twenty', url:'https://crm.mark.vibnai.com:3000' }])
followed by `compose up -d --force-recreate twenty` produced HTTP/2
200 from https://crm.mark.vibnai.com on first hit, with the
loadbalancer label present, .env clean, and zero env-rewriting
required.
Changes:
- apps.create template path now reads template.port from the catalog
and calls setServiceDomains with https://<fqdn>:<port>
- listServiceTemplates now accepts port as either number or numeric
string (Coolify ships both shapes in the catalog)
- applyCoolifyPostDeployFixes simplified from ~200 lines to ~50:
drops env rewrite, label injection, and force-recreate steps;
keeps proxy network attach + (background) proxy restart
- CoolifyPostDeployResult.steps shrinks to { proxyNetwork, proxyRestart }
- Removes the python:3-alpine SSH dependency entirely
- buildPythonRunner helper removed
Made-with: Cursor
The post-deploy step that restarts coolify-proxy was running
synchronously inside the HTTP request handler. coolify-proxy is the
same gateway that's serving the request itself, so the restart
killed our outbound response mid-flight — the agent saw curl exit
16 (HTTP/2 framing error) instead of our nicely-formatted result.
Switch to a fire-and-forget shell:
nohup sh -c '(sleep 3 && docker restart coolify-proxy) ...' &
The SSH command returns within ~50ms, we finish the HTTP response,
and Traefik re-discovers labels 3s later — same end state as before
but without breaking the calling request.
Made-with: Cursor
Two fixes for transient Coolify queue lag observed when smoke-testing
v2.4.5:
1. ensureServiceReachable no longer false-fails on early `exited` status.
Coolify's queue worker can take 60-120s to dequeue a `start` request;
during that window service.applications[*].status returns the stale
`exited` (= "never started") state. Previously the polling loop
treated that as terminal failure after 90s, returning started:false
on stacks that were about to come up healthy.
The new logic requires "evidence of activity" (status starting:* or
running:* seen at least once) before treating subsequent `exited`
reports as terminal. Until activity is observed, the loop just keeps
polling up to the 8-min health timeout.
2. apps.repair (new tool). Re-runs the three post-deploy patches
(env rewrite, traefik port label, coolify-proxy network attach +
force-recreate + proxy restart) against an existing service without
recreating it. Useful when:
- apps.create returned started:false but containers eventually
came up (now the polling fix should make this rare)
- a deploy succeeded mechanically but is serving Traefik 503 or
Mixed Content
- a user rotates a custom domain on an existing app
Params: { uuid, fqdn, publicAppName, port? }
Returns: { reachable, postDeploy: { steps }, probe }
Version bumped to 2.4.6.
Made-with: Cursor
apps.create for service templates now lets Coolify's queue do the
full deploy (compose generation, volumes, internal networking,
healthchecks) and applies three surgical post-deploy fixes that
Coolify's REST API does NOT expose:
1. Rewrites SERVICE_FQDN_* / SERVICE_URL_* in the rendered .env so
frontends that bake their backend URL into the SPA bundle
(Twenty's SERVER_URL, n8n, etc.) point at the real custom domain
instead of the auto-generated sslip.io URL. Without this fix
Twenty's frontend loads on the real HTTPS domain but fires XHRs
at insecure sslip.io, blocking everything as Mixed Content.
2. Injects the missing
traefik.http.services.<svc>.loadbalancer.server.port label.
Coolify generates the routing rules but forgets the port, so
Traefik logs "error: port is missing" and returns 503 forever.
3. Connects coolify-proxy to the project network (Coolify writes a
caddy_ingress_network=<uuid> hint label but never actually runs
docker network connect), then force-recreates ONLY the
public-facing container so the new env+label apply, and
restarts the proxy so Traefik re-discovers.
Polling switches from service.status (which routinely lies as
"starting:unknown" while containers are actually healthy) to the
truthful per-application service.applications[*].status field.
Removes the SSH "docker compose up -d" fallback that v2.4.1-2.4.4
used. That fallback bypassed Coolify's full pipeline, causing
internal services like Postgres/Redis to land on the shared coolify
network where DNS aliases collided with coolify-db/coolify-redis,
producing the "password authentication failed" regression we saw
on Twenty deploys. With v2.4.5 internal services stay on their
isolated project network — only the public app crosses to the
proxy.
Response shape gains: reachable (boolean for HTTPS 2xx/3xx),
appStatus (truthful per-app status from Coolify), postDeploy
(step-by-step diagnostic for each of the three fixes). Existing
started/startDiag fields kept for back-compat.
apps.containers.up / apps.containers.ps remain unchanged for
manual user recovery.
Made-with: Cursor
v2.4.3 attached every stack container to the `coolify` network so
Traefik could reach the public container. But that network also hosts
coolify-db (alias `postgres`) and coolify-redis (alias `redis`).
Docker's embedded DNS resolves unqualified hostnames to the first
container with that name on the network, so once Twenty's
`postgres-<uuid>` joined the coolify network, Twenty's connection
string `postgres://postgres:5432/...` started resolving to coolify-db
and auth-failing in a tight restart loop.
Coolify's own pipeline only attaches the proxied container — filter
by the `traefik.enable=true` label so internal stack members (db,
redis, worker) stay isolated on the project network.
Made-with: Cursor
The Twenty (and any service-template) stack was reachable on its private
project network but invisible to coolify-proxy/Traefik because no
container was joined to the `coolify` network. Public URLs like
crm.mark.vibnai.com returned 503 "no available server" even though the
underlying app was healthy.
Coolify's UI deploy attaches the proxy network as a post-step after the
full stack is up. When a sidecar (e.g. Twenty's worker, which waits ~3
min on twenty's healthcheck) fails its depends_on gate, that post-step
can be skipped and the stack is left isolated.
composeUp now calls attachToCoolifyProxyNetwork() after compose
finishes (best-effort, idempotent), and ensureServiceUp does the same
on the Coolify-queue happy path. Single apps.create call should now
result in a publicly reachable app.
Made-with: Cursor
Coolify's `compose up -d` returns non-zero whenever any sidecar container
hits a `depends_on: condition: service_healthy` timeout. For slow-booting
apps like Twenty (where the worker waits ~3 min for twenty's healthcheck),
this caused apps.create to return started=false even when the primary
stack was running fine.
Now ensureServiceUp probes the host with `docker ps` after a non-zero
compose exit and returns started=true whenever any container is running,
surfacing the compose stderr in startDiag so agents can decide whether
to retry apps.containers.up later.
Made-with: Cursor
Coolify's POST /services/{uuid}/start writes the rendered compose
files but its Laravel queue worker routinely fails to actually
invoke `docker compose up -d`. Until now agents had to SSH to
recover. For an MVP that promises "tell vibn what app you want,
get a URL", that's unacceptable.
- lib/coolify-compose.ts: composeUp/composeDown/composePs over SSH
via a one-shot docker:cli container that bind-mounts the rendered
compose dir (works around vibn-logs being in docker group but not
having read access to /data/coolify/services).
- apps.create (template + composeRaw pathways) now uses
ensureServiceUp which probes whether Coolify's queue actually
spawned containers and falls back to direct docker compose up -d
if not. Result includes startMethod for visibility.
- apps.containers.up / apps.containers.ps exposed as MCP tools for
recovery scenarios and post-env-change recreations.
- Tenant safety: resolveAppOrService validates uuid against the
caller's project before touching anything on the host.
Made-with: Cursor
Adds Coolify one-click template support — 320+ vetted apps deployable
in one MCP call (Twenty, n8n, Supabase, Ghost, etc).
- apps.create gains a 4th pathway: { template: "<slug>", ... }. Auto-
rewrites the Coolify-assigned sslip URL to the workspace FQDN and
applies user envs before starting.
- apps.templates.list / apps.templates.search expose the catalog so
agents can discover slugs. Catalog is fetched from upstream GitHub
and cached in-memory for 1h.
- lib/coolify.ts: + setServiceDomains, updateService, listService-
Templates, searchServiceTemplates. Reuses existing createService.
- next.config.ts: externalize ssh2 + cpu-features from turbopack so
`next build` can complete (native .node binaries can't be ESM-bundled).
Made-with: Cursor
Coolify's /applications/dockercompose creates a Service (not Application)
with its own API surface. Wire it up correctly:
lib/coolify.ts
- createDockerComposeApp returns { uuid, resourceType: 'service' }
- Add startService, stopService, getService, listAllServices helpers
- Add listServiceEnvs, upsertServiceEnv, bulkUpsertServiceEnvs for
the /services/{uuid}/envs endpoint
app/api/mcp/route.ts
- toolAppsList: includes Services (compose stacks) alongside Applications
- toolAppsDeploy: falls back to /services/{uuid}/start for service UUIDs
- toolAppsCreate composeRaw path: uses upsertServiceEnv + startService
instead of Application deploy; notes that domain routing must be
configured post-startup via SERVER_URL env
Made-with: Cursor
Third-party apps (Twenty, Directus, Cal.com, Plane…) should never need
a Gitea repo. This adds two new apps.create pathways:
image: "twentyhq/twenty:1.23.0" → Coolify /applications/dockerimage
composeRaw: "services:\n..." → Coolify /applications/dockercompose
No repo is created, no git clone, no PAT embedding. Agents can fetch
the official docker-compose.yml and pass it inline, or just give an
image name. Pathway 1 (repo) is unchanged.
Also adds volume management tools so agents can self-recover from the
most common compose failure (stale DB volume blocking fresh migrations):
apps.volumes.list { uuid } → list volumes + sizes
apps.volumes.wipe { uuid, volume, confirm } → stop containers,
rm volume, done
Both volume tools go through the same vibn-logs SSH channel. The wipe
tool requires confirm == volume name to prevent accidents and verifies
the volume belongs to the target app (uuid in name).
lib/coolify.ts: createDockerImageApp + createDockerComposeApp helpers,
dockerimage added to CoolifyBuildPack union.
app/api/mcp/route.ts: resolveFqdn + applyEnvsAndDeploy extracted as
shared helpers; toolAppsCreate now dispatches on image/composeRaw/repo.
toolAppsVolumesList + toolAppsVolumesWipe added.
sq() moved to module scope (shared by exec + volumes tools).
Version bumped to 2.3.0.
Made-with: Cursor
Companion to apps.logs. SSH to the Coolify host as vibn-logs, resolve
the target container by app uuid + service, and run the caller's
command through `docker exec ... sh -lc`. No TTY, no stdin — this is
the write-path sibling of apps.logs, purpose-built for migrations,
seeds, CLI invocations, and ad-hoc debugging.
- lib/coolify-containers.ts extracts container enumeration + service
resolution into a shared helper used by both logs and exec.
- lib/coolify-exec.ts wraps docker exec with timeout (60s default,
10-min cap), output byte cap (1 MB default, 5 MB cap), optional
--user / --workdir, and structured audit logging of the command +
target (never the output).
- app/api/mcp/route.ts wires `apps.exec` into the dispatcher and
advertises it in the capabilities manifest.
- app/api/workspaces/[slug]/apps/[uuid]/exec/route.ts exposes the same
tool over REST for session-cookie callers.
Tenant safety: every entrypoint runs getApplicationInProject before
touching SSH, so an agent can only exec in apps belonging to their
workspace.
Made-with: Cursor
Two live-test bugs surfaced while deploying Twenty CRM:
1. apps.domains.set silently 422'd on compose apps
Coolify hard-rejects top-level `domains` for dockercompose build
packs — they must use `docker_compose_domains` (per-service JSON).
setApplicationDomains now detects build_pack (fetched via GET if
not passed) and dispatches correctly. Default service is `server`
(matches Twenty, Plane, Cal.com); override with `service` param.
2. apps.update silently dropped unrecognised fields
Caller got `{ok:true}` even when zero fields persisted. This
created false-positive "bug reports" (e.g. the user-reported
"fqdn returns ok but doesn't persist" — fqdn was never forwarded
at all). apps.update now returns:
- applied: fields that were forwarded to Coolify
- ignored: unknown fields (agent typos, stale field names)
- rerouted: fields that belong to a different tool
(fqdn/domains → apps.domains.set,
git_repository → apps.rewire_git)
400 when nothing applied, 200 with diagnostics otherwise.
Made-with: Cursor
The per-workspace GCS backend (bucket, service account, HMAC keys) was
already provisioned for P5.3 but wasn't reachable through MCP, so
agents using vibn_sk_* tokens couldn't actually use object storage.
Three new tools:
- storage.describe → bucket, region, endpoint, access_key_id.
No secret in response.
- storage.provision → idempotent ensureWorkspaceGcsProvisioned().
- storage.inject_env → writes STORAGE_* (or user-chosen prefix) env
vars into a Coolify app. SECRET_ACCESS_KEY is
tagged is_shown_once so Coolify masks it in
the UI, and it never leaves our backend — the
agent kicks off injection, but the HMAC secret
is read from our DB and pushed directly to
Coolify.
Apps can then hit the bucket with any S3 SDK (aws-sdk, boto3, etc.)
using force_path_style=true and the standard endpoint.
Made-with: Cursor
Coolify was failing all Gitea clones with "Permission denied (publickey)"
because the helper container's SSH hits git.vibnai.com:22 (Ubuntu host
sshd, which doesn't know Gitea keys), while Gitea's builtin SSH is on
host port 22222 (not publicly reachable).
Rather than fight the SSH topology, switch every Vibn-provisioned app
to clone over HTTPS with the workspace bot's PAT embedded in the URL.
The PAT is already stored encrypted per workspace and scoped to that
org, so this gives equivalent isolation with zero SSH dependency.
Changes:
- lib/naming.ts: add giteaHttpsUrl() + redactGiteaHttpsUrl(); mark
giteaSshUrl() as deprecated-for-deploys with a comment.
- lib/coolify.ts: extend CreatePublicAppOpts with install/build/start
commands, base_directory, dockerfile_location, docker_compose_location,
manual_webhook_secret_gitea so it's at parity with the SSH variant.
- app/api/mcp/route.ts:
- apps.create now uses createPublicApp(giteaHttpsUrl(...)) and pulls
the bot PAT via getWorkspaceBotCredentials(). No more private-
deploy-key path for new apps.
- apps.update adds git_commit_sha + docker_compose_location to the
whitelist.
- New apps.rewire_git tool: re-points an app's git_repository at the
canonical HTTPS+PAT URL. Unblocks older apps stuck on SSH URLs
and provides a path for PAT rotation without rebuilding the app.
- lib/gitea.ts: createUser() now issues an immediate PATCH to set
active: true. Gitea's admin-create endpoint creates users as inactive
by default, and inactive users fail permission checks even though
they're org members. GiteaUser gains optional `active` field.
- scripts/activate-workspace-bots.ts: idempotent backfill that flips
active=true for any existing workspace bot that was created before
this fix. Safe to re-run.
- AI_CAPABILITIES.md: document apps.rewire_git; clarify apps.create
uses HTTPS+PAT (no SSH).
Already unblocked in prod for the mark workspace:
- vibn-bot-mark activated.
- twenty-crm's git_repository PATCHed to HTTPS+PAT form; git clone
now succeeds (remaining unrelated error: docker-compose file path).
Made-with: Cursor
Coolify v4's POST/PATCH /applications/{uuid}/envs only accepts key,
value, is_preview, is_literal, is_multiline, is_shown_once. Sending
is_build_time triggers a 422 "This field is not allowed." — it's now
a derived read-only flag (is_buildtime) computed from Dockerfile ARG
usage. Breaks agents trying to upsert env vars.
Three-layer fix so this can't regress:
- lib/coolify.ts: COOLIFY_ENV_WRITE_FIELDS whitelist enforced at the
network boundary, regardless of caller shape
- app/api/workspaces/[slug]/apps/[uuid]/envs: stops forwarding the
field; returns a deprecation warning when callers send it; GET
reads both is_buildtime and is_build_time for version parity
- app/api/mcp/route.ts: same treatment in the MCP dispatcher;
AI_CAPABILITIES.md doc corrected
Also bundles (not related to the above):
- Workspace API keys are now revealable from settings. New
key_encrypted column stores AES-256-GCM(VIBN_SECRETS_KEY, token).
POST /api/workspaces/[slug]/keys/[keyId]/reveal returns plaintext
for session principals only; API-key principals cannot reveal
siblings. Legacy keys stay valid for auth but can't reveal.
- P5.3 Object storage: lib/gcp/storage.ts + lib/workspace-gcs.ts
idempotently provision a per-workspace GCS bucket, service
account, IAM binding and HMAC key. New POST /api/workspaces/
[slug]/storage/buckets endpoint. Migration script + smoke test
included. Proven end-to-end against prod master-ai-484822.
Made-with: Cursor
Adds end-to-end custom apex domain support: workspace-scoped
registration via OpenSRS (Tucows), authoritative DNS via Google
Cloud DNS, and one-call attach that wires registrar nameservers,
DNS records, and Coolify app routing in a single transactional
flow.
Schema (additive, idempotent — run /api/admin/migrate after deploy)
- vibn_workspaces.dns_provider TEXT DEFAULT 'cloud_dns'
Per-workspace DNS backend choice. Future: 'cira_dzone' for
strict CA-only residency on .ca.
- vibn_domains
One row per registered/intended apex. Tracks status
(pending|active|failed|expired), registrar order id, encrypted
registrar manage-user creds (AES-256-GCM, VIBN_SECRETS_KEY),
period, dates, dns_provider/zone_id/nameservers, and a
created_by audit field.
- vibn_domain_events
Append-only lifecycle audit (register.attempt/success/fail,
attach.success, ns.update, lock.toggle, etc).
- vibn_billing_ledger
Workspace-scoped money ledger (CAD by default) with
ref_type/ref_id back to the originating row.
OpenSRS XML client (lib/opensrs.ts)
- Mode-gated host/key (OPENSRS_MODE=test → horizon sandbox,
rejectUnauthorized:false; live → rr-n1-tor, strict TLS).
- MD5 double-hash signature.
- Pure Node https module (no undici dep).
- Verbs: lookupDomain, getDomainPrice, checkDomain, registerDomain,
updateDomainNameservers, setDomainLock, getResellerBalance.
- TLD policy: minPeriodFor() bumps .ai to 2y; CPR/legalType
plumbed through for .ca; registrations default to UNLOCKED so
immediate NS updates succeed without a lock toggle.
DNS provider abstraction (lib/dns/{provider,cloud-dns}.ts)
- DnsProvider interface (createZone/getZone/setRecords/deleteZone)
so the workspace residency knob can swap backends later.
- cloudDnsProvider implementation against Google Cloud DNS using
the existing vibn-workspace-provisioner SA (roles/dns.admin).
- Idempotent zone creation, additions+deletions diff for rrsets.
Shared GCP auth (lib/gcp-auth.ts)
- Single getGcpAccessToken() helper used by Cloud DNS today and
future GCP integrations. Prefers GOOGLE_SERVICE_ACCOUNT_KEY_B64,
falls back to ADC.
Workspace-scoped helpers (lib/domains.ts)
- listDomainsForWorkspace, getDomainForWorkspace, createDomainIntent,
markDomainRegistered, markDomainFailed, markDomainAttached,
recordDomainEvent, recordLedgerEntry.
Attach orchestrator (lib/domain-attach.ts)
Single function attachDomain() reused by REST + MCP. For one
apex it:
1. Resolves target → Coolify app uuid OR raw IP OR CNAME.
2. Ensures Cloud DNS managed zone exists.
3. Writes A / CNAME records (apex + requested subdomains).
4. Updates registrar nameservers, with auto unlock-retry-relock
fallback for TLDs that reject NS changes while locked.
5. PATCHes the Coolify application's domain list so Traefik
routes the new hostname.
6. Persists dns_provider/zone_id/nameservers and emits an
attach.success domain_event.
AttachError carries a stable .tag + http status so the caller
can map registrar/dns/coolify failures cleanly.
REST endpoints
- POST /api/workspaces/[slug]/domains/search
- GET /api/workspaces/[slug]/domains
- POST /api/workspaces/[slug]/domains
- GET /api/workspaces/[slug]/domains/[domain]
- POST /api/workspaces/[slug]/domains/[domain]/attach
All routes go through requireWorkspacePrincipal (session OR
Authorization: Bearer vibn_sk_...). Register is idempotent:
re-issuing for an existing intent re-attempts at OpenSRS without
duplicating the row or charging twice.
MCP bridge (app/api/mcp/route.ts → version 2.2.0)
Adds five tools backed by the same library code:
- domains.search (batch availability + pricing)
- domains.list (workspace-owned)
- domains.get (single + recent events)
- domains.register (idempotent OpenSRS register)
- domains.attach (full Cloud DNS + registrar + Coolify)
Sandbox smoke tests (scripts/smoke-opensrs-*.ts)
Standalone Node scripts validating each new opensrs.ts call against
horizon.opensrs.net: balance + lookup + check, TLD policy
(.ca/.ai/.io/.com), full register flow, NS update with systemdns
nameservers, and the lock/unlock toggle that backs the attach
fallback path.
Post-deploy checklist
1. POST https://vibnai.com/api/admin/migrate
-H "x-admin-secret: $ADMIN_MIGRATE_SECRET"
2. Set OPENSRS_* env vars on the vibn-frontend Coolify app
(RESELLER_USERNAME, API_KEY_LIVE, API_KEY_TEST, HOST_LIVE,
HOST_TEST, PORT, MODE). Without them, only domains.list/get
work; search/register/attach return 500.
3. GCP_PROJECT_ID is read from env or defaults to master-ai-484822.
4. Live attach end-to-end against a real apex is queued as a
follow-up — sandbox path is fully proven.
Not in this commit (deliberate)
- The 100+ unrelated in-flight files (mvp-setup wizard, justine
homepage rework, BuildLivePlanPanel, etc) — kept local to keep
blast radius minimal.
Made-with: Cursor
Coolify exposes the template slug on `service_type`; the list endpoint
returns only summaries, so the auth list handler now fetches each
service individually to classify it reliably. Users can name auth
services anything (e.g. "my-login") and they still show up as auth
providers.
Made-with: Cursor
Workspace-owned deploy infra so AI agents can create and destroy
Coolify resources without ever touching the root admin token.
vibn_workspaces
+ coolify_server_uuid, coolify_destination_uuid
+ coolify_environment_name (default "production")
+ coolify_private_key_uuid, gitea_bot_ssh_key_id
ensureWorkspaceProvisioned
+ generates an ed25519 keypair per workspace
+ pushes pubkey to the Gitea bot user (read/write scoped by team)
+ registers privkey in Coolify as a reusable deploy key
New endpoints under /api/workspaces/[slug]/
apps/ POST (private-deploy-key from Gitea repo)
apps/[uuid] PATCH, DELETE?confirm=<name>
apps/[uuid]/domains GET, PATCH (policy: *.{ws}.vibnai.com only)
databases/ GET, POST (8 types incl. postgres, clickhouse, dragonfly)
databases/[uuid] GET, PATCH, DELETE?confirm=<name>
auth/ GET, POST (Pocketbase, Authentik, Keycloak, Pocket-ID, Logto, Supertokens)
auth/[uuid] DELETE?confirm=<name>
MCP (/api/mcp) gains 15 new tools that mirror the REST surface and
enforce the same workspace tenancy + delete-confirm guard.
Safety: destructive ops require ?confirm=<exact-resource-name>; volumes
are kept by default (pass delete_volumes=true to drop).
Made-with: Cursor
Without this the bot PAT 403s on POST /orgs/{org}/repos, which is
the single most important operation — creating new project repos
inside the workspace's Gitea org.
Made-with: Cursor
Ship Phases 1–3 of the multi-tenant AI access plan so an AI agent can
act on a Vibn workspace with one bearer token and zero admin reach.
Phase 1 — Gitea bot per workspace
- Add gitea_bot_username / gitea_bot_user_id / gitea_bot_token_encrypted
columns to vibn_workspaces (migrate route).
- New lib/auth/secret-box.ts (AES-256-GCM, VIBN_SECRETS_KEY) for PAT at rest.
- Extend lib/gitea.ts with createUser, createAccessTokenFor (Sudo PAT),
createOrgTeam, addOrgTeamMember, ensureOrgTeamMembership.
- ensureWorkspaceProvisioned now mints a vibn-bot-<slug> user, adds it to
a Writers team (write perms only) on the workspace's org, and stores
its PAT encrypted.
- GET /api/workspaces/[slug]/gitea-credentials returns a workspace-scoped
bot PAT + clone URL template; session or vibn_sk_ bearer auth.
Phase 2 — Tenant-safe Coolify proxy + real MCP
- lib/coolify.ts: projectUuidOf, listApplicationsInProject,
getApplicationInProject, TenantError, env CRUD, deployments list.
- Workspace-scoped REST endpoints (all filtered by coolify_project_uuid):
GET/POST /api/workspaces/[slug]/apps/[uuid](/deploy|/envs|/deployments),
GET /api/workspaces/[slug]/deployments/[deploymentUuid]/logs.
- Full rewrite of /api/mcp off legacy Firebase onto Postgres vibn_sk_
keys, exposing workspace.describe, gitea.credentials, projects.*,
apps.* (list/get/deploy/deployments, envs.list/upsert/delete).
Phase 3 — Settings UI AI bundle
- GET /api/workspaces/[slug]/bootstrap.sh: curl|sh installer that writes
.cursor/rules, .cursor/mcp.json and appends VIBN_* to .env.local.
Embeds the caller's vibn_sk_ token when invoked with bearer auth.
- WorkspaceKeysPanel: single AiAccessBundleCard with system-prompt block,
one-line bootstrap, Reveal-bot-PAT button, collapsible manual-setup
fallback. Minted-key modal also shows the bootstrap one-liner.
Ops prerequisites:
- Set VIBN_SECRETS_KEY (>=16 chars) on the frontend.
- Run /api/admin/migrate to add the three bot columns.
- GITEA_API_TOKEN must be a site-admin token (needed for admin/users
+ Sudo PAT mint); otherwise provision_status lands on 'partial'.
Made-with: Cursor
The settings page imported `auth` from `@/lib/firebase/config` and called
`auth.currentUser` inside an unguarded `useEffect`. Since the app runs on
PostgreSQL + NextAuth (Firebase isn't configured), `auth` was `undefined`
and the uncaught TypeError crashed React's commit, leaving the page blank
behind the Next.js dev error overlay. The WorkspaceKeysPanel never got a
chance to mount even though `/api/workspaces` was returning fine.
Swap to `useSession()` from `next-auth/react` to read display name + email
from the existing NextAuth session. Drop the dead fetch to
`/api/workspace/{slug}/settings`, which was never implemented.
Made-with: Cursor
The panel was fetching /api/workspaces/{urlSlug} where {urlSlug}
is whatever is in the `[workspace]` dynamic segment (e.g.
"mark-account"). That slug has nothing to do with vibn_workspaces.slug,
which is derived from the user's email — so the fetch 404'd, the
component showed "Loading workspace…" forever, and minting/revoking
would target a non-existent workspace.
Now:
- GET /api/workspaces lazy-creates a workspace row if the signed-in
user has none (migration path for accounts created before the
signIn hook was added).
- WorkspaceKeysPanel discovers the user's actual workspace from that
list and uses *its* slug for all subsequent calls (details, keys,
provisioning, revocation).
- Empty / error states render a proper card with a retry button
instead of a bare "Workspace not found." line.
Made-with: Cursor
Adds logical multi-tenancy on top of Coolify + Gitea so every Vibn
account gets its own isolated tenant boundary, and exposes that
boundary to AI agents (Cursor, Claude Code, scripts) through
per-workspace bearer tokens.
Schema (additive, idempotent — run /api/admin/migrate once after deploy)
- vibn_workspaces: slug, name, owner, coolify_project_uuid,
coolify_team_id (reserved for when Coolify ships POST /teams),
gitea_org, provision_status
- vibn_workspace_members: room for multi-user workspaces later
- vibn_workspace_api_keys: sha256-hashed bearer tokens
- fs_projects.vibn_workspace_id: nullable FK linking projects
to their workspace
Provisioning
- On first sign-in, ensureWorkspaceForUser() inserts the row
(no network calls — keeps signin fast).
- On first project create, ensureWorkspaceProvisioned() lazily
creates a Coolify Project (vibn-ws-{slug}) and a Gitea org
(vibn-{slug}). Failures are recorded on the row, not thrown,
and POST /api/workspaces/{slug}/provision retries.
Auth surface
- lib/auth/workspace-auth.ts: requireWorkspacePrincipal() accepts
either a NextAuth session or "Authorization: Bearer vibn_sk_...".
The bearer key is hard-pinned to one workspace — it cannot
reach any other tenant.
- mintWorkspaceApiKey / listWorkspaceApiKeys / revokeWorkspaceApiKey
Routes
- GET /api/workspaces list
- GET /api/workspaces/[slug] details
- POST /api/workspaces/[slug]/provision retry provisioning
- GET /api/workspaces/[slug]/keys list keys
- POST /api/workspaces/[slug]/keys mint key (token shown once)
- DELETE /api/workspaces/[slug]/keys/[keyId] revoke
UI
- components/workspace/WorkspaceKeysPanel.tsx: identity card,
keys CRUD with one-time secret reveal, and a "Connect Cursor"
block with copy/download for:
.cursor/rules/vibn-workspace.mdc — rule telling the agent
about the API + workspace IDs + house rules
~/.cursor/mcp.json — MCP server registration with key
embedded (server URL is /api/mcp; HTTP MCP route lands next)
.env.local — VIBN_API_KEY + smoke-test curl
- Slotted into existing /[workspace]/settings between Workspace
and Notifications cards (no other layout changes).
projects/create
- Resolves the user's workspace (creating + provisioning lazily).
- Repos go under workspace.gitea_org (falls back to GITEA_ADMIN_USER
for backwards compat).
- Coolify services are created inside workspace.coolify_project_uuid
(renamed {slug}-{appName} to stay unique within the namespace) —
no more per-Vibn-project Coolify Project sprawl.
- Stamps vibn_workspace_id on fs_projects.
lib/gitea
- createOrg, getOrg, addOrgOwner, getUser
- createRepo now routes /orgs/{owner}/repos when owner != admin
Also includes prior-turn auth hardening that was already in
authOptions.ts (CredentialsProvider for dev-local, isLocalNextAuth
cookie config) bundled in to keep the auth layer in one consistent
state.
.env.example
- Documents GITEA_API_URL / GITEA_API_TOKEN / GITEA_ADMIN_USER /
GITEA_WEBHOOK_SECRET and COOLIFY_URL / COOLIFY_API_TOKEN /
COOLIFY_SERVER_UUID, with the canonical hostnames
(git.vibnai.com, coolify.vibnai.com).
Post-deploy
- Run once: curl -X POST https://vibnai.com/api/admin/migrate \\
-H "x-admin-secret: \$ADMIN_MIGRATE_SECRET"
- Existing users get a workspace row on next sign-in.
- Existing fs_projects keep working (legacy gitea owner + their
own per-project Coolify Projects); new projects use the
workspace-scoped path.
Not in this commit (follow-ups)
- Wiring requireWorkspacePrincipal into the rest of /api/projects/*
so API keys can drive existing routes
- HTTP MCP server at /api/mcp (the mcp.json snippet already
points at the right URL — no client re-setup when it lands)
- Backfill script to assign legacy fs_projects to a workspace
Made-with: Cursor
- Add app/styles/justine/01-homepage.css: rules from 01_homepage.html scoped to [data-justine]
- Replace app/(marketing) with app/(justine): layout wraps data-justine + Plus Jakarta
- JustineHomePage/Nav/Footer: original class names (btn-ink, hero-grid, …) + inline styles from HTML
- Remove app/justine-marketing.css; move /features /pricing /privacy /terms under (justine)
Made-with: Cursor
- Replace blue/purple gradients with ink gradient text and cream/parch CTA surface
- Step badges and transformation icons use primary (ink) fills
- /features page icons unified to text-primary; Lora section titles
- Tree view status colors use semantic tokens instead of blue/green
Made-with: Cursor
- Map Justine tokens to shadcn CSS variables (--vibn-* aliases)
- Switch fonts to Inter + Lora via next/font (IBM Plex Mono for code)
- Base typography: body Inter, h1–h3 Lora; marketing hero + wordmark serif
- Project shell and global chrome use semantic colors
- Replace Outfit/Newsreader references across TSX inline styles
Made-with: Cursor
New users were being sent to /marks-account/projects. Now derives
workspace from the signed-in user's email so everyone lands on
their own workspace after Google OAuth.
Made-with: Cursor
- PRD page now has a tabbed view: PRD | Architecture
Architecture tab renders apps, packages, infrastructure, integrations,
and risk notes as structured cards. Only shown when arch doc exists.
- Advisor route now includes the architecture summary and key fields
in the COO's knowledge context so the orchestrator knows what's
been planned technically
Made-with: Cursor
The advisor route now proxies to /orchestrator/chat on agents.vibnai.com
instead of calling Gemini directly. The Orchestrator (Claude Sonnet 4.6)
has full tool access — Gitea, Coolify, web search, memory, agent spawning.
- Build project knowledge_context from DB (name, vision, repo, PRD,
phases, apps, recent sessions) and inject as COO persona + data
- Convert frontend history format (model→assistant) for the orchestrator
- Return orchestrator reply as streaming text response
- Session scoped per project for in-memory context persistence
Made-with: Cursor
