feat(config): multi-deployment environments + ResolveDatabaseTargets

[Kiran01bm]

What ?

Adds an additive deployments: map under databases.<db>.environments.<env> so one environment can fan out to N Tern deployments, plus a new ResolveDatabaseTargets resolver. No production behavior changes yet — no caller of ResolveDatabaseTargets exists (yet).

Why ?

The resolver shape is the contract between config and orchestration. Landing it before the apply write path is updated means downstream callers can be pure consumers of the new API — no half-migrated state in either direction.

Config shape

databases:
  payments:
    type: mysql
    environments:
      production:
        deployments:
          payments-a:
            target: payments
          payments-b:
            target: payments
          payments-c:
            target: payments

tern_deployments:
  payments-a:
    production: "tern-payments-a:9090"
  payments-b:
    production: "tern-payments-b:9090"
  payments-c:
    production: "tern-payments-c:9090"

The map key IS the Tern deployment name — same key space as tern_deployments, and the same value that will eventually land on apply_operations.deployment. Each entry just carries target:.

Resolver API

// New — returns one element per configured deployment, sorted by deployment key.
func (c *ServerConfig) ResolveDatabaseTargets(database, env string) ([]ResolvedDatabaseTarget, error)

// Existing — now delegates to ResolveDatabaseTargets. Errors with a clear
// "use ResolveDatabaseTargets" message when the env resolves to N != 1
// deployments. Single-deployment callers (plan/apply/progress handlers,
// integration tests) are unaffected.
func (c *ServerConfig) ResolveDatabaseTarget(database, env string) (ResolvedDatabaseTarget, error)

Validation rules

ServerConfig.Validate() now also rejects:

• mixing scalar target / deployment with the deployments: map
• mixing local dsn / dsn_from with the deployments: map
• an empty deployments: map
• a map entry with empty target
• a map key not present in tern_deployments, or present but missing an endpoint for the current environment (same cross-check the scalar shape already does)

Back-compat matrix

Config has	ResolveDatabaseTargets	ResolveDatabaseTarget (legacy)
Scalar target + deployment	1-element slice	unchanged, works
deployments: map (N entries)	N elements sorted by deployment key	errors with "use ResolveDatabaseTargets"
Local dsn / dsn_from	1-element slice (Deployment == Target == <db name>)	unchanged, works
Both scalar and map present	validation error at config load	—
Local DSN AND any remote routing	validation error at config load	—
Neither scalar nor map nor local DSN	validation error at config load	—

Design decision worth flagging

The map key is the Tern deployment name, not a separate "logical deployment label". An earlier draft had { deployment: payments-a, target: payments } inside each entry, which would split deployment identity across two keys for no current gain. If we ever need a logical label decoupled from the Tern endpoint name (e.g., for PR-comment display), we can add it later as an additive field on DeploymentTarget.

Cross-deployment promotion order (a deployment_order: analogous to environment_order:) is deferred to a follow-up that also renames the scheduler to operator. This PR sorts by deployment key for deterministic resolution.

Tests

• TestServerConfig_ResolveDatabaseTargets: local DSN, scalar remote, sorted multi-deployment, unknown db, unknown env, multi-deployment via the legacy resolver, scalar via the legacy resolver.
• TestServerConfig_DeploymentsMapValidation (table): happy path + six rejection cases (mixed scalar+map, empty map, empty target, unknown deployment, deployment without endpoint for env, local DSN + map).

Verification

• go build ./... ✅
• go vet ./... ✅
• go test ./pkg/api/... ./pkg/state/... ./pkg/storage/... ./pkg/schema/... ./pkg/webhook/... ✅

Out of scope

• Wiring ResolveDatabaseTargets into the apply / plan write paths — a follow-up will start writing one apply_operations row per resolved deployment.
• Server-owned cross-deployment promotion order (deployment_order:) — deferred to the follow-up that renames scheduler to operator.
• Per-deployment rows in PR progress comments.
• Per-deployment Check Run rollup.

Rollback

Pure revert. No on-disk artifact, no schema changes, no live callers of the new API.

[Copilot]

Pull request overview

Adds first-class config + resolver support for “multi-deployment environments”, allowing a single (database, environment) to resolve to N Tern deployments deterministically, while keeping legacy single-deployment callers working (and explicitly erroring when they encounter N≠1).

Changes:

• Extend EnvironmentConfig with an additive deployments: map and validation rules to prevent mixing local DSN, scalar routing, and map routing.
• Introduce ServerConfig.ResolveDatabaseTargets() and refactor ResolveDatabaseTarget() to delegate and error when multiple deployments resolve.
• Document the new config shape and add Helm values examples; add unit tests covering resolution and validation.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
pkg/api/config.go	Adds Deployments config shape, validates it, and implements ResolveDatabaseTargets + legacy delegation behavior.
pkg/api/config_test.go	Adds tests for multi-deployment resolution ordering and validation rejection cases.
docs/configuration.md	Documents the multi-deployment environment YAML shape and rules.
charts/schemabot/values.yaml	Adds a commented Helm values example showing multi-deployment configuration.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 63d84fe79c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".