feat(api): POST /deploy/new redeploy=true — replace-in-place by name (DOG-30)

[mastermanas805]

Summary

• Adds an optional redeploy=true multipart form field to POST /deploy/new. When set with a matching name, the handler looks up the team's most-recent non-terminal deployment for (team_id, env, env_vars._name) and routes through the same compute path as POST /deploy/:id/redeploy — same app_id, same URL, same provider_id. Closes the duplicate-URL incident (2026-05-30) where three identical calls for "truehomie-web" minted three distinct subdomains because there was no upsert-by-name path.
• Response gains a redeployed: bool field (always present — false on the fresh-deploy path, true on the in-place path). Stable shape across both branches so agents can branch on a single key.
• Extracts the redeploy compute goroutine into runRedeployAsync so the vault-ref resolution, internal-env-key stripping, compute.Redeploy, status update, audit emit, failure-autopsy capture, and build-log fetch all live in ONE place (called from both POST /deploy/:id/redeploy and the new in-place branch).

Error surfaces

code	status	when
redeploy_requires_name	400	redeploy=true + empty name
no_existing_deployment_to_redeploy	404	redeploy=true but no live row matches (team-scoped — cross-team requests resolve here too, never leaking existence)
not_ready	409	match found but the existing row has no provider_id yet (initial build still running)

Coverage block (rule 17)

Symptom:        POST /deploy/new with same name fans out to N app_ids/URLs
Enumeration:    rg -F 'generateAppID' / rg -F '/deploy/new' / rg -F 'CreateDeploymentWithCap'
Sites found:    1 in api (handlers/deploy.go:New)
                + MCP `redeploy` separately broken (see cross-repo TODOs)
Sites touched:  1 — handlers/deploy.go:New gains the in-place branch
Coverage tests: TestDeployNew_RedeployTrue_MatchFound_ReusesAppID
                TestDeployNew_RedeployTrue_NoMatch_Returns404
                TestDeployNew_RedeployTrue_NoName_Returns400
                TestDeployNew_RedeployFalse_CreatesNewAsBefore
                TestDeployNew_RedeployTrue_WrongTeam_Returns404
                TestDeployNew_Response_AlwaysIncludesRedeployedField
Live verified:  awaiting user verification after merge + deploy
                (rule 14: curl https://api.instanode.dev/healthz → commit_id match
                 then POST /deploy/new redeploy=true against staging app)

Rule 22 surface checklist

• api/internal/handlers/openapi.go — redeploy boolean on DeployRequest, redeployed boolean on DeployResponse + DeployResponse.item.
• cross-repo TODOs (call-out for the docs/MCP/infra agents):
  • content/llms.txt + content/api-reference.md — document redeploy=true on POST /deploy/new so agents discover the path
  • mcp/src/index.ts — wire redeploy arg into create_deploy and/or fix the bodyless POST in the existing redeploy tool (mcp/src/index.ts:1444 — server requires multipart tarball, current MCP tool sends no body and gets 400). This PR unblocks the API side but the MCP wedge needs the matching client change.
  • infra/k8s/prometheus-rules.yaml + infra/newrelic/alerts/*.json + infra/newrelic/dashboards/instanode-reliability.json — add alert + dashboard tile for instant_deploy_redeploy_total{outcome="not_found"} (rising rate = agents guessing names; pair with list_deployments discovery), plus the success tile (rule 25 — every new metric ships with its alert + tile).
  • infra/observability/METRICS-CATALOG.md — row for instant_deploy_redeploy_total.

Files

• internal/handlers/deploy.go — shouldRedeployInPlace(form) helper; in-place branch in New; runRedeployAsync extracted from Redeploy.
• internal/models/deployment.go — FindActiveDeploymentByTeamEnvName (status filter: building/deploying/healthy/failed; excludes deleted/expired/stopped).
• internal/models/audit_kinds.go — AuditKindDeployRedeployRequested (deploy.redeploy.requested), emitted from both POST /deploy/:id/redeploy and POST /deploy/new in-place; source field distinguishes the two.
• internal/metrics/metrics.go — instant_deploy_redeploy_total{outcome="success|not_found|wrong_team|missing_name"}.
• internal/handlers/openapi.go — schema updates.
• internal/handlers/deploy_redeploy_inplace_test.go — 6 new tests.

Test plan

• go test -run 'TestDeployNew_Redeploy|TestDeployNew_Response_AlwaysIncludesRedeployedField' -short -count=1 ./internal/handlers/ — all 6 PASS locally
• go vet ./... clean
• make gate — single unrelated flake in TestLinkGitHubID (test-DB pollution from prior runs, users_github_id_key duplicate; passes in isolation: go test -run TestLinkGitHubID -short -count=1 ./internal/models/ → ok)
• Live verification: after merge + auto-deploy, POST https://api.instanode.dev/deploy/new with redeploy=true against a real deploy, confirm same app_id + URL come back; verify instant_deploy_redeploy_total{outcome="success"} ticks in /metrics

🤖 Generated with Claude Code