sync: port module changes from constructive-db (a909a8a)

[pyramation]

Summary

Ports the three modules that had drifted from constructive-db/pgpm-modules at constructive-db main a909a8a (post constructive-db#3785). The other 25 shared modules were byte-identical and are untouched; versions stay at 0.45.0 as in prior sync commits.

• function-resolution: install_mantra, install_route_bindings, resolve_capabilities, validate_capabilities (shared-surface capability resolution, route-binding install)
• metaschema-modules: billing_module / billing_provider_module tables gain the credit-pack and provider-first subscription seam columns (credit_packs_table_id, grant_meter_credits_function, get_billing_subscription_by_*, get_plan_pricing_by_external_price)
• metaschema-schema: is_valid_step_up, embedding_chunks

Bundles regenerated with pgpm package. pnpm run policy:check and pgpm test-packages --full-cycle pass.

Also: CI's object-store service container moves from minio/minio:edge-cicd (tag no longer pullable from Docker Hub — every job on main would fail the same way today) to rustfs/rustfs:1.0.0-rc.5, mirroring constructive-db bd08f6532af.

Not ported: db-utils and infra-utils exist only in constructive-db and are not part of this repo.

Link to Devin session: https://app.devin.ai/sessions/42be922ae2324153b365dbda21e5b85d
Open in Devin Desktop: https://app.devin.ai/desktop/session/42be922ae2324153b365dbda21e5b85d?variant=devin
Requested by: @pyramation

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[tenki-reviewer]

Review complete. 🟠 2 high

💬 Inline comments (1)

• 🟠 Redirect insert binds stray $5 on global tier — install_route_bindings.sql:504

📍 Findings outside the diff (1) — 🟠 1 high — defects on lines GitHub can't attach comments to

🟠 High — Drop removed arg from capability test callers · resolve_capabilities.sql:62–69 · unchanged line

// packages/function-resolution/deploy/schemas/function_resolution/procedures/resolve_capabilities.sql
62	CREATE FUNCTION function_resolution.resolve_capabilities(
63	    database_id uuid,
64	    scope text,
65	    entity_id uuid,
66	    function_definition_id uuid,
67	    definition_scope text,
68	    payload jsonb DEFAULT '{}'::jsonb,
69	    channel text DEFAULT NULL

resolve_capabilities and validate_capabilities both drop the definition_database_id parameter, changing their signatures to (uuid, text, uuid, uuid, text, jsonb, text), but the package's own suite still invokes them with the deleted argument (resolve_capabilities.sql:62; callers in capabilities.test.ts at 511, 533, 551, 566, 579, 588, 596, 602). PostgreSQL rejects the stale arity, so every capability test now fails instead of exercising the new contract.

🧹 Nitpicks (1) — 🟢 1 low

• 🟢 Verify omits new get_active_plan_subscription_function (table.sql:24) — The billing_provider_module verify lists three of the four read-seam columns the deploy adds but omits get_active_plan_subscription_function (verify/.../billing_provider_module/table.sql:25).

---

This PR changes capability resolution and schema across three pgpm packages. The core behavioral shift is resolve_capabilities/validate_capabilities losing the definition_database_id parameter in favor of frame-key lookup, plus new "redirect" route-binding support in the function-resolution procedures and new billing columns/tables in metaschema-modules, all mirrored in regenerated sql/*--0.45.0.sql bundles.

Files	Change
function-resolution deploy/verify/revert procedures	Refactor capability resolution to keyed frame lookup, drop definition_database_id, add redirect-binding install logic
function-resolution sql/ bundle	Regenerate to match the new procedure signatures and logic
metaschema-modules billing tables + verify	Add billing module/provider columns (credit packs, grant function, provider function-name seams) and their verify checks
metaschema-schema step-up + embedding_chunks + bundle	Harden is_valid_step_up, add source_fields, regenerate bundle

Two high-severity issues need attention before merge: test callers still invoke the old 8-argument arity, and the redirect-insert path binds a stray $5 parameter on the global tier; the billing-provider verify omits the newly added function column.

_{Reviewed commit: 84ce602}

[tenki-reviewer]

PR #130 refactors capability resolution to frame-key-based lookup (dropping definition_database_id) and adds redirect route bindings plus billing-module table columns across function-resolution, metaschema-modules, and metaschema-schema, with regenerated sql bundles.

Key findings

• 🟠 Drop removed arg from capability test callers — resolve_capabilities.sql:62
• 🟠 Redirect insert binds stray $5 on global tier — install_route_bindings.sql:504

[tenki-reviewer]

🟠 bug · high

Redirect insert binds stray $5 on global tier

The redirect INSERT always binds five USING parameters including key_value, but on the ownerless/global tier where routes_key is NULL the generated value list is VALUES ($1, $2, $3, $4, false, true) with no $5 placeholder (install_route_bindings.sql:504). plpgsql raises wrong number of parameters for dynamic EXECUTE: expected 4, got 5, so every redirect binding installed onto a global-tier routes plane fails at runtime; the sibling route INSERT balances both branches, marking this one as unintentionally unbalanced.

📋 Prompt for AI Agents

In packages/function-resolution/deploy/schemas/function_resolution/procedures/install_route_bindings.sql at the redirect INSERT (lines 504-515), the EXECUTE always passes 5 USING args (including key_value) but the query references only $1..$4 when routes_key IS NULL, so plpgsql fails with 'wrong number of parameters for dynamic EXECUTE: expected 4, got 5' on the global tier. Restructure the EXECUTE so the USING list is conditional: pass key_value only in the routes_key IS NOT NULL branch (mirroring how the route INSERT balances its key/serving prefixes), keeping the counted parameters equal to the highest referenced placeholder in both branches so redirect installs work on ownerless planes.