feat(native): migrate SwarmPinService onto DbClient (#176, stacked on #209)

[Dewinator]

Summary

First service migrated off the (supabaseUrl, supabaseKey) constructor pattern and onto the shared DbClient surface that PR #209 introduced. Stacked on agent/db-client-factory — that PR must merge first.

Why

PR #209 landed the DbClient factory + MYCELIUM_USE_PGLITE switch as the foundation. The next step is migrating ~17 services off the legacy (url, key) pattern, one at a time, so each migration is reviewable on its own. SwarmPinService is the smallest service still on the legacy shape (~250 LOC, 14 tests, narrow surface) — ideal first target.

Changes

• mcp-server/src/services/swarm-pin.ts:
  • Constructor takes DbClient (PostgrestClient | PGliteAdapter) instead of url + key. Service no longer constructs its own SupabaseClient.
  • defaultDeps() widens db locally because TypeScript can't unify the .from() generics across the two backends. Runtime chain shape is identical (gated by pglite-adapter.contract.test.ts). Widening is local to the closure — does not leak.
  • Dropped @supabase/supabase-js import.
• mcp-server/src/index.ts:
  • Constructs a single PostgrestClient inline for swarm-pin, in the same style as middleware/{absorber,digester,prime-fetcher}.ts. The other ~17 service constructors still take (url, key) — they will migrate one-by-one.
  • Comment marker explains why MYCELIUM_USE_PGLITE mode-switch isn't yet wired here: it lights up when index.ts boot goes async (await createDbClient()), and the async refactor is worth doing once enough services consume DbClient to make the blast radius pay off.
• mcp-server/src/__tests__/swarm-pin.test.ts:
  • Constructor calls now pass a typed-but-empty STUB_DB cast. Orchestration tests inject SwarmPinDeps so the underlying client is never touched — the cast purely satisfies the new signature.

Validation

rm -rf dist && npm run build                  → clean
node --test dist/__tests__/swarm-pin.test.js  → 14/14 passed (buildPinReason 5, pinLesson 6, swarmPinLesson 2, schema 1)
node --test dist/__tests__/db-client.test.js  → 4/4 passed (no regression in the parent factory)

Stacking

Base: agent/db-client-factory (PR #209). When #209 merges to main, GitHub will auto-rebase this PR's base to main.

Test plan

• CI green on the stacked branch
• After feat(native): DbClient factory — MYCELIUM_USE_PGLITE switch (#176 follow-up) #209 merges, PR auto-retargets to main and tests still pass
• Visual review: middleware + service layer style consistency

Related

• PR feat(native): DbClient factory — MYCELIUM_USE_PGLITE switch (#176 follow-up) #209 — DbClient factory (parent)
• Issue epic: native standalone app (no Docker) — macOS / Windows / Linux #176 — Native standalone app epic (sub-task 3)

🤖 Generated with Claude Code

[Dewinator]

Local empirical validation 2026-05-05 (227th-tick) — first SwarmPin DbClient migration green

PR #210 had no prior comments; the stack-base PR #209 was revalidated against 8b6a203 earlier today, so this comment closes the gap on the stacked piece itself.

Setup

• Fresh checkout of agent/swarm-pin-dbclient-stacked at HEAD 1bf6d3f
• Mac M4, 16 GB RAM
• rm -rf dist && npm run build && npm test against the existing .env (Docker/Supabase backend; the migrated service uses the PostgrestClient branch of DbClient)

Results

> tsc                         # build clean, no diagnostics
ℹ tests 1047
ℹ pass 1046
ℹ fail 0
ℹ skipped 1                   # same baseline as PR #209 — no new skips, no regression
ℹ duration_ms 67642

Focused replay of the SwarmPin suite:

node --test dist/__tests__/swarm-pin.test.js
ℹ tests 14
ℹ pass 14
ℹ fail 0
ℹ duration_ms 357

Migration shape — empirical cross-check

Diff vs. 8b6a203 (PR #209 head): services/swarm-pin.ts constructor goes from (supabaseUrl, supabaseKey) to (db: DbClient) (L103). The service no longer constructs its own SupabaseClient — it consumes the one its caller hands in. defaultDeps() widens db to { from(t: string): any } locally (L152-L153) because TS can't unify the .from() generic across PostgrestClient and PGliteAdapter — the chain shape itself (.select()/.insert()/.update().eq().maybeSingle()) is identical at runtime, and that runtime equivalence is what the PR #209 pglite-adapter contract tests assert.

Stack readiness — no blockers from this PR

• 79-migration walk on PG 17.5 + pgvector 0.8.1 stays green (docs(native-pg): re-validate spike — all 79 migrations green on PGlite (#177) #194/test(native): pin PR #194 spike — 79-migration walk as CI gate #204 inherited via feat(native): DbClient factory — MYCELIUM_USE_PGLITE switch (#176 follow-up) #209).
• No new SQL migration in feat(native): migrate SwarmPinService onto DbClient (#176, stacked on #209) #210 — pure service-side migration.
• index.ts wiring is the only ripple at the orchestrator boundary (12 lines).
• 22 → 21 services after feat(native): migrate SwarmPinService onto DbClient (#176, stacked on #209) #210 lands: SwarmPin moves out of the direct-Supabase set; SkillsService follows in stacked PR feat(native): migrate SkillsService onto DbClient + shared dbClient (#176, stacked on #210) #211.

Recommendation

Merge order remains #209 → #210 → #211 (linear stack). #210 is locally clean against today's main and the stack-base; no merge-conflict signals from git log.

🤖 Posted by autonomous tick (mycelium 227th — local PR-validation, no new PR opened, queue-bound)