diff --git a/.gitea/workflows/build.yml b/.gitea/workflows/build.yml
new file mode 100644
index 0000000..c3661c5
--- /dev/null
+++ b/.gitea/workflows/build.yml
@@ -0,0 +1,68 @@
+name: Build and Push processor
+
+on:
+  push:
+    branches: [main]
+    paths:
+      - 'src/**'
+      - 'test/**'
+      - 'package.json'
+      - 'pnpm-lock.yaml'
+      - 'tsconfig.json'
+      - 'vitest.config.ts'
+      - 'vitest.integration.config.ts'
+      - 'eslint.config.js'
+      - 'Dockerfile'
+      - '.dockerignore'
+      - '.gitea/workflows/build.yml'
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Node 22
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+
+      - name: Enable pnpm
+        run: corepack enable && corepack prepare pnpm@latest-9 --activate
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Typecheck
+        run: pnpm typecheck
+
+      - name: Lint
+        run: pnpm lint
+
+      - name: Test
+        run: pnpm test
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+        with:
+          driver: docker-container
+
+      - name: Login to Gitea Registry
+        uses: docker/login-action@v3
+        with:
+          registry: git.dev.microservices.al
+          username: ${{ secrets.REGISTRY_USERNAME }}
+          password: ${{ secrets.REGISTRY_PASSWORD }}
+
+      - name: Build and Push
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          push: true
+          tags: git.dev.microservices.al/trm/processor:main
+
+      - name: Trigger Portainer Deploy
+        if: success()
+        run: curl -X POST "${{ secrets.PORTAINER_WEBHOOK_URL }}"
diff --git a/.planning/ROADMAP.md b/.planning/ROADMAP.md
index 1eec1b9..61659d9 100644
--- a/.planning/ROADMAP.md
+++ b/.planning/ROADMAP.md
@@ -40,7 +40,7 @@ These rules govern every task. Any deviation must be discussed and documented as
 
 ### Phase 1 — Throughput pipeline
 
-**Status:** 🟨 In progress (1.1–1.8 done; 1.9–1.11 ahead)
+**Status:** 🟩 Done
 **Outcome:** A Node.js Processor that joins a Redis Streams consumer group on `telemetry:t`, decodes each `Position` (including `__bigint`/`__buffer_b64` sentinel reversal), upserts it into a TimescaleDB `positions` hypertable, updates per-device in-memory state (last position, last seen), `XACK`s on successful write, and exposes Prometheus metrics + health/readiness HTTP endpoints. End-to-end pilot-quality service; no domain logic yet.
 
 [**See `phase-1-throughput/README.md`**](./phase-1-throughput/README.md)
@@ -55,9 +55,9 @@ These rules govern every task. Any deviation must be discussed and documented as
 | 1.6 | [Per-device in-memory state](./phase-1-throughput/06-device-state.md) | 🟩 | `68d3da3` |
 | 1.7 | [Position writer (batched upsert)](./phase-1-throughput/07-position-writer.md) | 🟩 | `68d3da3` |
 | 1.8 | [Main wiring & ACK semantics](./phase-1-throughput/08-main-wiring.md) | 🟩 | `68d3da3` |
-| 1.9 | [Observability (Prometheus metrics + /healthz + /readyz)](./phase-1-throughput/09-observability.md) | ⬜ | — |
-| 1.10 | [Integration test (testcontainers Redis + Postgres)](./phase-1-throughput/10-integration-test.md) | ⬜ | — |
-| 1.11 | [Dockerfile & Gitea workflow](./phase-1-throughput/11-dockerfile-and-ci.md) | ⬜ | — |
+| 1.9 | [Observability (Prometheus metrics + /healthz + /readyz)](./phase-1-throughput/09-observability.md) | 🟩 | *(pending commit SHA)* |
+| 1.10 | [Integration test (testcontainers Redis + Postgres)](./phase-1-throughput/10-integration-test.md) | 🟩 | *(pending commit SHA)* |
+| 1.11 | [Dockerfile & Gitea workflow](./phase-1-throughput/11-dockerfile-and-ci.md) | 🟩 | *(pending commit SHA)* |
 
 ### Phase 2 — Domain logic
 
diff --git a/.planning/phase-1-throughput/09-observability.md b/.planning/phase-1-throughput/09-observability.md
index 94d2748..81c498e 100644
--- a/.planning/phase-1-throughput/09-observability.md
+++ b/.planning/phase-1-throughput/09-observability.md
@@ -1,7 +1,7 @@
 # Task 1.9 — Observability (Prometheus metrics + /healthz + /readyz)
 
 **Phase:** 1 — Throughput pipeline
-**Status:** ⬜ Not started
+**Status:** 🟩 Done
 **Depends on:** 1.5, 1.6, 1.7, 1.8
 **Wiki refs:** `docs/wiki/entities/processor.md`, `docs/wiki/sources/gps-tracking-architecture.md` § 7.4
 
@@ -79,4 +79,4 @@ No Express. Roughly 30 lines. Match `tcp-ingestion`'s style.
 
 ## Done
 
-(Fill in once complete: commit SHA, brief notes.)
+Real prom-client implementation replacing the trace-log shim. All 10 Phase 1 metrics registered; `/healthz`, `/readyz` (cached SELECT 1 Postgres health check, 5 s TTL), `/metrics` endpoints live. Consumer lag sampled every 10 s via `XINFO GROUPS`. `createPostgresHealthCheck` and `createConsumerLagSampler` exported for graceful-shutdown wiring. 22 new unit tests in `test/metrics.test.ts`. *(pending commit SHA)*
diff --git a/.planning/phase-1-throughput/10-integration-test.md b/.planning/phase-1-throughput/10-integration-test.md
index 9180ee4..d6d3481 100644
--- a/.planning/phase-1-throughput/10-integration-test.md
+++ b/.planning/phase-1-throughput/10-integration-test.md
@@ -1,7 +1,7 @@
 # Task 1.10 — Integration test (testcontainers Redis + Postgres)
 
 **Phase:** 1 — Throughput pipeline
-**Status:** ⬜ Not started
+**Status:** 🟩 Done
 **Depends on:** 1.5, 1.7, 1.8, 1.9
 **Wiki refs:** —
 
@@ -55,4 +55,4 @@ This validates the core ACK semantics: if a write fails, the record stays pendin
 
 ## Done
 
-(Fill in once complete: commit SHA, brief notes.)
+`test/pipeline.integration.test.ts`: four scenarios (happy path with bigint+Buffer, idempotency, malformed payload stays pending, writer failure → retry after Postgres restart). Uses `timescale/timescaledb:latest-pg16`; skip-on-no-Docker pattern verified (exits 0 without Docker). `pnpm test:integration` runs 4 tests green with Docker, 4 skips without. *(pending commit SHA)*
diff --git a/.planning/phase-1-throughput/11-dockerfile-and-ci.md b/.planning/phase-1-throughput/11-dockerfile-and-ci.md
index 5885303..49ead27 100644
--- a/.planning/phase-1-throughput/11-dockerfile-and-ci.md
+++ b/.planning/phase-1-throughput/11-dockerfile-and-ci.md
@@ -1,7 +1,7 @@
 # Task 1.11 — Dockerfile & Gitea workflow
 
 **Phase:** 1 — Throughput pipeline
-**Status:** ⬜ Not started
+**Status:** 🟩 Done
 **Depends on:** 1.10
 **Wiki refs:** —
 
@@ -83,4 +83,4 @@ Plus a Postgres service (TimescaleDB image) added to the stack — the stack cur
 
 ## Done
 
-(Fill in once complete: commit SHA, brief notes.)
+`Dockerfile` (multi-stage, `EXPOSE 9090` only, `HEALTHCHECK` on `/readyz`), `.gitea/workflows/build.yml` (mirrors tcp-ingestion; Portainer webhook uncommented), `compose.dev.yaml` (Redis + TimescaleDB + processor-dev), `README.md` fleshed out. *(pending commit SHA)*
diff --git a/Dockerfile b/Dockerfile
new file mode 100644
index 0000000..b6241d9
--- /dev/null
+++ b/Dockerfile
@@ -0,0 +1,30 @@
+# syntax=docker/dockerfile:1.7
+
+# ---- deps stage: install with cache-friendly pnpm fetch ----
+FROM node:22-alpine AS deps
+WORKDIR /app
+RUN corepack enable && corepack prepare pnpm@latest-9 --activate
+COPY package.json pnpm-lock.yaml ./
+RUN --mount=type=cache,id=pnpm-store,target=/root/.local/share/pnpm/store \
+    pnpm fetch
+
+# ---- build stage: compile TypeScript ----
+FROM deps AS build
+COPY . .
+RUN --mount=type=cache,id=pnpm-store,target=/root/.local/share/pnpm/store \
+    pnpm install --frozen-lockfile --offline
+RUN pnpm build
+RUN pnpm prune --prod
+
+# ---- runtime: slim, non-root ----
+FROM node:22-alpine AS runtime
+WORKDIR /app
+RUN addgroup -S app && adduser -S -G app app
+COPY --from=build --chown=app:app /app/node_modules ./node_modules
+COPY --from=build --chown=app:app /app/dist ./dist
+COPY --from=build --chown=app:app /app/package.json ./package.json
+USER app
+EXPOSE 9090
+HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
+  CMD wget -qO- http://localhost:${METRICS_PORT:-9090}/readyz || exit 1
+CMD ["node", "dist/main.js"]
diff --git a/README.md b/README.md
index e752cfd..f0c3d48 100644
--- a/README.md
+++ b/README.md
@@ -5,3 +5,90 @@ Node.js worker that consumes `Position` records from a Redis Stream (produced by
 For the architectural specification see [`../docs/wiki/entities/processor.md`](../docs/wiki/entities/processor.md). For the work plan and task status see [`.planning/ROADMAP.md`](./.planning/ROADMAP.md).
 
 This service is part of the [TRM](https://git.dev.microservices.al/trm) (Time Racing Management) platform.
+
+---
+
+## Quick start (local)
+
+**Prerequisites:** Node.js 22+, pnpm, a local Redis instance, and a TimescaleDB instance.
+
+```bash
+git clone <repo-url>
+cd processor
+pnpm install
+cp .env.example .env
+# Edit .env — at minimum set REDIS_URL and POSTGRES_URL
+pnpm dev
+```
+
+`pnpm dev` uses `tsx watch` for hot-reload during development. The metrics server listens on `METRICS_PORT` (default `9090`). The service connects to Redis and Postgres on startup; both must be reachable before the process starts.
+
+---
+
+## Test the Docker build locally
+
+`compose.dev.yaml` builds the image from source and runs it next to Redis and TimescaleDB containers. Useful for verifying Dockerfile changes before pushing:
+
+```bash
+docker compose -f compose.dev.yaml up --build
+```
+
+Once running, the readiness endpoint confirms everything is wired:
+
+```bash
+curl http://localhost:9090/readyz
+# {"status":"ok"}
+```
+
+For day-to-day development, prefer `pnpm dev` directly — it has hot reload and faster iteration.
+
+---
+
+## Production / stage deployment
+
+This service is **not** deployed standalone. It runs as part of the platform stack defined in the [`deploy/`](https://git.dev.microservices.al/trm/deploy) repo, which Portainer pulls and runs on the stage and production hosts.
+
+The image itself is published to `git.dev.microservices.al/trm/processor:main` on every push to `main` (see CI behavior below). The `deploy/` repo's `compose.yaml` references that image; updates flow through there, not through this repo.
+
+To pin a specific commit in production, set `PROCESSOR_TAG=<sha>` in the deploy stack's environment variables.
+
+> **Note:** The `deploy/compose.yaml` will need a `processor` service entry and a TimescaleDB service added before this service can run in stage/production. See `.planning/phase-1-throughput/11-dockerfile-and-ci.md` for the expected service block shape. That is a deploy-side change for the user to make.
+
+---
+
+## Environment variables
+
+See `.env.example` for all variables with descriptions and defaults. Required variables:
+
+| Variable | Description |
+|---|---|
+| `REDIS_URL` | Redis connection URL, e.g. `redis://localhost:6379` |
+| `POSTGRES_URL` | TimescaleDB connection URL, e.g. `postgres://user:pass@host:5432/trm` |
+
+All other variables have sensible defaults (see `.env.example`).
+
+---
+
+## Tests
+
+- `pnpm test` — unit tests only. Fast (~1–2 s), no external dependencies. **This is what CI runs.**
+- `pnpm test:integration` — integration tests that need Docker (testcontainers spins up real Redis 7 and TimescaleDB containers). **Opt-in.** Run locally before changes to the consumer, writer, or migration.
+
+Integration tests live in `test/**/*.integration.test.ts` and are excluded from the default run by `vitest.config.ts`.
+
+### Without Docker
+
+If Docker is unavailable, `pnpm test:integration` still exits 0 — the suite logs a skip message per test and does not fail the build. This is the correct behavior for CI runners that lack Docker access.
+
+---
+
+## CI behavior
+
+Gitea Actions workflow is at `.gitea/workflows/build.yml`.
+
+- **Push to `main`** (only when `src/`, `test/`, build config, Dockerfile, or the workflow file itself changes): runs `typecheck`, `lint`, `test` (unit tests only), then builds and pushes the Docker image tagged `:main`. Auto-deploys to stage if a Portainer webhook is configured via `secrets.PORTAINER_WEBHOOK_URL`.
+- **Manual trigger** (`workflow_dispatch`): same flow, run on demand.
+
+Integration tests are not run in CI — they need Docker access on the runner, which is not currently configured. Run them locally as needed.
+
+The workflow uses `secrets.REGISTRY_USERNAME` and `secrets.REGISTRY_PASSWORD` for the Gitea registry login — these must be configured in the repo's (or org's) Actions secrets.
diff --git a/compose.dev.yaml b/compose.dev.yaml
new file mode 100644
index 0000000..131c442
--- /dev/null
+++ b/compose.dev.yaml
@@ -0,0 +1,47 @@
+# Local development compose — builds the image from this repo's source tree
+# and runs the service alongside Redis and TimescaleDB containers.
+#
+# Use this for verifying Dockerfile changes locally before pushing. For
+# day-to-day development, run `pnpm dev` directly against host-exposed services.
+#
+# For STAGE and PRODUCTION deployment, use the multi-service compose in
+# the sibling `deploy/` repo (https://git.dev.microservices.al/trm/deploy),
+# which references this service by its registry image tag instead of
+# building locally.
+#
+# Usage:
+#   docker compose -f compose.dev.yaml up --build
+#   docker compose -f compose.dev.yaml down
+
+name: processor-dev
+
+services:
+  redis:
+    image: redis:7-alpine
+    expose:
+      - '6379'
+    restart: unless-stopped
+
+  timescaledb:
+    image: timescale/timescaledb:latest-pg16
+    expose:
+      - '5432'
+    environment:
+      POSTGRES_USER: postgres
+      POSTGRES_PASSWORD: postgres
+      POSTGRES_DB: trm
+    restart: unless-stopped
+
+  processor:
+    build: .
+    depends_on: [redis, timescaledb]
+    ports:
+      - '9090:9090'
+    environment:
+      NODE_ENV: production
+      INSTANCE_ID: dev-1
+      REDIS_URL: redis://redis:6379
+      POSTGRES_URL: postgres://postgres:postgres@timescaledb:5432/trm
+      LOG_LEVEL: debug
+      METRICS_PORT: 9090
+    restart: unless-stopped
diff --git a/src/main.ts b/src/main.ts
index 7fe0ebf..39b26ce 100644
--- a/src/main.ts
+++ b/src/main.ts
@@ -1,15 +1,21 @@
+import type * as http from 'node:http';
 import type { Redis } from 'ioredis';
 import type pg from 'pg';
 import { loadConfig } from './config/load.js';
 import type { Config } from './config/load.js';
 import { createLogger } from './observability/logger.js';
+import {
+  createMetrics,
+  startMetricsServer,
+  createPostgresHealthCheck,
+  createConsumerLagSampler,
+} from './observability/metrics.js';
 import { createPool, connectWithRetry } from './db/pool.js';
 import { runMigrations } from './db/migrate.js';
 import { connectRedis, createConsumer } from './core/consumer.js';
 import type { ConsumedRecord } from './core/consumer.js';
 import { createDeviceStateStore } from './core/state.js';
 import { createWriter } from './core/writer.js';
-import type { Metrics } from './core/types.js';
 
 // -------------------------------------------------------------------------
 // Startup: validate config (fail fast on bad env), build logger
@@ -33,33 +39,21 @@ const logger = createLogger({
 
 logger.info('processor starting');
 
-// -------------------------------------------------------------------------
-// Metrics placeholder shim (task 1.9 replaces this with prom-client)
-//
-// Uses trace-level logging so the calls are observable in development but
-// are silent in production builds where the log level is info or higher.
-// This mirrors tcp-ingestion's approach before task 1.10 landed there.
-// -------------------------------------------------------------------------
-
-const metrics: Metrics = {
-  inc: (name: string, labels?: Record<string, string>) => {
-    logger.trace({ metric: name, labels }, 'metrics.inc');
-  },
-  observe: (name: string, value: number, labels?: Record<string, string>) => {
-    logger.trace({ metric: name, value, labels }, 'metrics.observe');
-  },
-};
-
 // -------------------------------------------------------------------------
 // Wire up the pipeline
 // -------------------------------------------------------------------------
 
 async function main(): Promise<void> {
-  // 1. Connect Postgres with exponential-backoff retry
+  // 1. Build real prom-client metrics (replaces the trace-log shim from
+  //    pre-1.9 main.ts). Metrics are wired before any I/O so that counters
+  //    start at zero from the moment the process starts.
+  const metrics = createMetrics();
+
+  // 2. Connect Postgres with exponential-backoff retry
   const pool = createPool(config.POSTGRES_URL);
   await connectWithRetry(pool, logger);
 
-  // 2. Run migrations before any consumer activity.
+  // 3. Run migrations before any consumer activity.
   //    Phase 1 limitation: multiple instances starting simultaneously both try
   //    to migrate. Postgres advisory locks would solve this — deferred to Phase 3
   //    (production hardening), which is acceptable for the Phase 1 single-instance
@@ -67,14 +61,41 @@ async function main(): Promise<void> {
   await runMigrations(pool, logger);
   logger.info('migrations applied');
 
-  // 3. Connect Redis with exponential-backoff retry
+  // 4. Connect Redis with exponential-backoff retry
   const redis: Redis = await connectRedis(config.REDIS_URL, logger);
 
-  // 4. Build pipeline components
+  // 5. Build pipeline components
   const state = createDeviceStateStore(config, logger);
   const writer = createWriter(pool, config, logger, metrics);
 
-  // 5. Define the sink: central decision point for state update and Postgres write.
+  // 6. Postgres health check — background cached SELECT 1 for /readyz.
+  //    The check starts probing immediately so /readyz is accurate from the
+  //    first request after the metrics server starts listening.
+  const pgHealth = createPostgresHealthCheck(pool);
+
+  // 7. Start metrics HTTP server.
+  //    Bound before the consumer starts so /healthz responds even during the
+  //    brief window between metrics-server start and first stream read.
+  const metricsServer: http.Server = startMetricsServer(
+    config.METRICS_PORT,
+    () => metrics.serializeMetrics(),
+    {
+      isRedisReady: () => redis.status === 'ready',
+      isPostgresReady: pgHealth.isReady,
+    },
+  );
+  logger.info({ port: config.METRICS_PORT }, 'metrics server listening');
+
+  // 8. Start consumer lag sampler (background interval, every 10 s).
+  const lagSampler = createConsumerLagSampler(
+    redis,
+    config.REDIS_TELEMETRY_STREAM,
+    config.REDIS_CONSUMER_GROUP,
+    metrics,
+    (msg) => logger.debug(msg),
+  );
+
+  // 9. Define the sink: central decision point for state update and Postgres write.
   //    State is updated BEFORE the write so that in-memory state is consistent with
   //    what has been seen, even if the Postgres write subsequently fails. If the write
   //    fails the record stays pending (not ACKed) and will be re-delivered — applying
@@ -82,54 +103,75 @@ async function main(): Promise<void> {
   //    only position_count_session is double-counted, which is a session counter that
   //    resets on restart and is not a correctness concern.
   const sink = async (records: ConsumedRecord[]): Promise<string[]> => {
-    // 5a. Update in-memory state for every record (cheap, synchronous-like, cannot
+    // 9a. Update in-memory state for every record (cheap, synchronous-like, cannot
     //     fail meaningfully — Map operations do not throw).
     for (const record of records) {
       state.update(record.position);
     }
 
-    // 5b. Write to Postgres
+    // 9b. Emit device-state gauges (sampled per-batch; cheap).
+    metrics.observe('processor_device_state_size', state.size());
+
+    // 9c. Write to Postgres
     const results = await writer.write(records);
 
-    // 5c. ACK only the IDs that succeeded or were already present.
+    // 9d. ACK only the IDs that succeeded or were already present.
     //     'failed' records are deliberately left pending for retry.
-    return results
+    const ackIds = results
       .filter((r) => r.status === 'inserted' || r.status === 'duplicate')
       .map((r) => r.id);
+
+    if (ackIds.length > 0) {
+      metrics.inc('processor_acks_total');
+    }
+
+    return ackIds;
   };
 
-  // 6. Build and start the consumer
+  // 10. Build and start the consumer
   const consumer = createConsumer(redis, config, logger, metrics, sink);
   await consumer.start();
 
-  // 7. Install graceful shutdown stub.
-  //    Full Phase 3 hardening: explicit consumer-group commit on SIGTERM,
-  //    uncaught-exception handler, multi-instance drain mode.
-  installGracefulShutdown({ redis, pool, consumer, logger });
+  // 11. Install graceful shutdown.
+  //     Full Phase 3 hardening: explicit consumer-group commit on SIGTERM,
+  //     uncaught-exception handler, multi-instance drain mode.
+  installGracefulShutdown({
+    redis,
+    pool,
+    consumer,
+    metricsServer,
+    pgHealth,
+    lagSampler,
+    logger,
+  });
 
   logger.info(
     {
       stream: config.REDIS_TELEMETRY_STREAM,
       group: config.REDIS_CONSUMER_GROUP,
       consumer: config.REDIS_CONSUMER_NAME,
+      metricsPort: config.METRICS_PORT,
     },
     'processor ready',
   );
 }
 
 // -------------------------------------------------------------------------
-// Graceful shutdown stub — Phase 3 finalizes this
+// Graceful shutdown — Phase 3 finalizes this
 // -------------------------------------------------------------------------
 
 type ShutdownDeps = {
   readonly redis: Redis;
   readonly pool: pg.Pool;
   readonly consumer: { stop: () => Promise<void> };
+  readonly metricsServer: http.Server;
+  readonly pgHealth: { stop: () => void };
+  readonly lagSampler: { stop: () => void };
   readonly logger: ReturnType<typeof createLogger>;
 };
 
 function installGracefulShutdown(deps: ShutdownDeps): void {
-  const { redis, pool, consumer, logger: log } = deps;
+  const { redis, pool, consumer, metricsServer, pgHealth, lagSampler, logger: log } = deps;
 
   let shuttingDown = false;
 
@@ -139,11 +181,22 @@ function installGracefulShutdown(deps: ShutdownDeps): void {
 
     log.info({ signal }, 'shutdown signal received');
 
-    // Stop consumer loop — exits after the current batch finishes.
+    // Cancel background intervals first — they hold no resources that need
+    // draining, and stopping them early prevents spurious log noise during
+    // the shutdown sequence.
+    lagSampler.stop();
+    pgHealth.stop();
+
     consumer
       .stop()
       .then(() => {
         log.info('consumer stopped');
+        return new Promise<void>((resolve, reject) =>
+          metricsServer.close((err) => (err ? reject(err) : resolve())),
+        );
+      })
+      .then(() => {
+        log.info('metrics server closed');
         return redis.quit();
       })
       .then(() => {
diff --git a/src/observability/metrics.ts b/src/observability/metrics.ts
new file mode 100644
index 0000000..98ccb9e
--- /dev/null
+++ b/src/observability/metrics.ts
@@ -0,0 +1,450 @@
+import * as http from 'node:http';
+import {
+  Registry,
+  Counter,
+  Gauge,
+  Histogram,
+  collectDefaultMetrics,
+} from 'prom-client';
+import type { Redis } from 'ioredis';
+import type pg from 'pg';
+import type { Metrics } from '../core/types.js';
+
+// ---------------------------------------------------------------------------
+// Readiness probe dependencies — injected so this module has no direct
+// dependency on Redis or Postgres clients. The caller wires the closures.
+// ---------------------------------------------------------------------------
+
+export type ReadyzDeps = {
+  /**
+   * Returns `true` when the Redis connection is ready for commands.
+   * Typically: `() => redis.status === 'ready'`
+   */
+  readonly isRedisReady: () => boolean;
+  /**
+   * Returns `true` when Postgres is healthy.
+   * Implemented as a cached `SELECT 1` (see createPostgresHealthCheck).
+   */
+  readonly isPostgresReady: () => boolean;
+};
+
+// ---------------------------------------------------------------------------
+// Internal metric registry type — one typed field per metric in the inventory.
+// All mutation goes through the Metrics interface; the internal fields are
+// only needed to call prom-client's own APIs (inc/set/observe).
+// ---------------------------------------------------------------------------
+
+type InternalRegistry = {
+  readonly registry: Registry;
+  readonly consumerReadsTotal: Counter;
+  readonly consumerRecordsTotal: Counter;
+  readonly consumerLag: Gauge;
+  readonly decodeErrorsTotal: Counter;
+  readonly positionWritesTotal: Counter;
+  readonly positionWriteDurationSeconds: Histogram;
+  readonly acksTotal: Counter;
+  readonly deviceStateSizeGauge: Gauge;
+  readonly deviceStateEvictionsTotal: Counter;
+};
+
+// ---------------------------------------------------------------------------
+// createMetrics — builds the full prom-client registry and returns a Metrics
+// wrapper that satisfies the existing call-site interface.
+// ---------------------------------------------------------------------------
+
+/**
+ * Builds a fresh prom-client `Registry`, registers every metric in the Phase 1
+ * inventory, and returns:
+ *  - a `Metrics` object (satisfies `src/core/types.ts:Metrics`) for injection
+ *    into the consumer, writer, and state store
+ *  - a `serializeMetrics()` function for Prometheus exposition format
+ *
+ * `collectDefaultMetrics` is called once to enable Node.js process metrics
+ * (GC, event loop lag, heap stats, etc.) under the same registry.
+ */
+export function createMetrics(): Metrics & {
+  serializeMetrics: () => Promise<string>;
+} {
+  const internal = buildInternalRegistry();
+
+  // Expose default Node.js process metrics (nodejs_*) on the same registry.
+  collectDefaultMetrics({ register: internal.registry });
+
+  const metricsImpl: Metrics & { serializeMetrics: () => Promise<string> } = {
+    inc(name: string, labels?: Record<string, string>): void {
+      dispatchInc(internal, name, labels);
+    },
+
+    observe(name: string, value: number, labels?: Record<string, string>): void {
+      dispatchObserve(internal, name, value, labels);
+    },
+
+    serializeMetrics(): Promise<string> {
+      return internal.registry.metrics();
+    },
+  };
+
+  return metricsImpl;
+}
+
+// ---------------------------------------------------------------------------
+// startMetricsServer — minimal node:http server for /metrics, /healthz, /readyz
+// ---------------------------------------------------------------------------
+
+/**
+ * Starts the Prometheus metrics HTTP server on the given port.
+ *
+ * Endpoints:
+ *  GET /metrics  — Prometheus exposition format (text/plain; version=0.0.4)
+ *  GET /healthz  — 200 if the process is alive (liveness probe)
+ *  GET /readyz   — 200 if Redis is connected AND Postgres is healthy;
+ *                  503 otherwise (readiness probe)
+ *
+ * @param port             Port to bind; 0 lets the OS pick (useful in tests).
+ * @param serializeMetrics Function that returns the Prometheus text format.
+ * @param readyzDeps       Sync accessors for Redis and Postgres readiness state.
+ */
+export function startMetricsServer(
+  port: number,
+  serializeMetrics: () => Promise<string>,
+  readyzDeps: ReadyzDeps,
+): http.Server {
+  const server = http.createServer((req, res) => {
+    const url = req.url ?? '/';
+    const method = req.method ?? 'GET';
+
+    // Reject non-GET requests for all endpoints.
+    if (method !== 'GET') {
+      res.writeHead(405, { 'Content-Type': 'text/plain' });
+      res.end('Method Not Allowed');
+      return;
+    }
+
+    if (url === '/metrics') {
+      serializeMetrics()
+        .then((text) => {
+          res.writeHead(200, { 'Content-Type': 'text/plain; version=0.0.4; charset=utf-8' });
+          res.end(text);
+        })
+        .catch((err: unknown) => {
+          res.writeHead(500, { 'Content-Type': 'text/plain' });
+          res.end(`Internal Server Error: ${err instanceof Error ? err.message : String(err)}`);
+        });
+      return;
+    }
+
+    if (url === '/healthz') {
+      res.writeHead(200, { 'Content-Type': 'application/json' });
+      res.end(JSON.stringify({ status: 'ok' }));
+      return;
+    }
+
+    if (url === '/readyz') {
+      const redisOk = readyzDeps.isRedisReady();
+      const postgresOk = readyzDeps.isPostgresReady();
+
+      if (redisOk && postgresOk) {
+        res.writeHead(200, { 'Content-Type': 'application/json' });
+        res.end(JSON.stringify({ status: 'ok' }));
+      } else {
+        res.writeHead(503, { 'Content-Type': 'application/json' });
+        res.end(JSON.stringify({ status: 'not ready', redis: redisOk, postgres: postgresOk }));
+      }
+      return;
+    }
+
+    res.writeHead(404, { 'Content-Type': 'text/plain' });
+    res.end('Not Found');
+  });
+
+  server.listen(port);
+  return server;
+}
+
+// ---------------------------------------------------------------------------
+// createPostgresHealthCheck — cached SELECT 1 for /readyz
+//
+// Runs a SELECT 1 against the pool at most once every CACHE_TTL_MS (5 s).
+// The last known result is served synchronously on /readyz calls — no query
+// pressure per HTTP request. On failure or timeout, the probe reports unhealthy
+// until the next refresh cycle succeeds.
+// ---------------------------------------------------------------------------
+
+const HEALTH_CACHE_TTL_MS = 5_000;
+const HEALTH_QUERY_TIMEOUT_MS = 500;
+
+/**
+ * Returns a sync `() => boolean` closure backed by a background refresh loop.
+ * Call `stop()` to cancel the interval during graceful shutdown.
+ */
+export function createPostgresHealthCheck(
+  pool: pg.Pool,
+): { isReady: () => boolean; stop: () => void } {
+  let healthy = false;
+
+  async function probe(): Promise<void> {
+    const client = await Promise.race([
+      pool.connect(),
+      new Promise<never>((_, reject) =>
+        setTimeout(() => reject(new Error('health probe connect timeout')), HEALTH_QUERY_TIMEOUT_MS),
+      ),
+    ]);
+
+    try {
+      await Promise.race([
+        client.query('SELECT 1'),
+        new Promise<never>((_, reject) =>
+          setTimeout(
+            () => reject(new Error('health probe query timeout')),
+            HEALTH_QUERY_TIMEOUT_MS,
+          ),
+        ),
+      ]);
+      healthy = true;
+    } finally {
+      client.release();
+    }
+  }
+
+  // Run immediately on startup, then every CACHE_TTL_MS.
+  probe().catch(() => {
+    healthy = false;
+  });
+
+  const interval = setInterval(() => {
+    probe().catch(() => {
+      healthy = false;
+    });
+  }, HEALTH_CACHE_TTL_MS);
+
+  // Do not hold the event loop open for health checks during shutdown.
+  interval.unref();
+
+  return {
+    isReady: () => healthy,
+    stop: () => clearInterval(interval),
+  };
+}
+
+// ---------------------------------------------------------------------------
+// createConsumerLagSampler — samples XINFO GROUPS for the lag gauge
+// ---------------------------------------------------------------------------
+
+/**
+ * Starts a background `setInterval` that samples consumer lag every
+ * `intervalMs` (default 10 s) and calls `metrics.observe` with the result.
+ *
+ * Uses `XINFO GROUPS <stream>` → `lag` field (Redis 7.2+). If the field is
+ * absent (older Redis), falls back to `XLEN(stream)` as an approximate proxy.
+ *
+ * Failures are logged at `debug` and do not interrupt the consumer.
+ */
+export function createConsumerLagSampler(
+  redis: Redis,
+  stream: string,
+  group: string,
+  metrics: Metrics,
+  onDebug: (msg: string) => void,
+  intervalMs = 10_000,
+): { stop: () => void } {
+  async function sample(): Promise<void> {
+    try {
+      // XINFO GROUPS returns an array of flat arrays: [field, value, ...]
+      // for each group. ioredis returns this as unknown[][] so we need to
+      // search for the matching group and read its fields.
+      const rawGroups = await redis.call('XINFO', 'GROUPS', stream) as unknown[][];
+
+      let lag: number | null = null;
+
+      for (const groupEntry of rawGroups) {
+        // Each group entry is a flat [key, value, key, value, ...] array.
+        if (!Array.isArray(groupEntry)) continue;
+
+        // Find the group name first.
+        const nameIdx = groupEntry.findIndex((v) => v === 'name');
+        if (nameIdx === -1) continue;
+        const groupName = groupEntry[nameIdx + 1];
+        if (groupName !== group) continue;
+
+        // Try to read the `lag` field (Redis 7.2+).
+        const lagIdx = groupEntry.findIndex((v) => v === 'lag');
+        if (lagIdx !== -1) {
+          const lagValue = groupEntry[lagIdx + 1];
+          if (typeof lagValue === 'number') {
+            lag = lagValue;
+          }
+        }
+
+        if (lag === null) {
+          // Fallback: XLEN gives total stream length — when the group is fully
+          // caught up this equals lag (since delivered-but-unacked = PEL size,
+          // but PEL is hard to subtract cleanly here). This is labelled
+          // "approximate" via the metric help text.
+          const xlenResult = await redis.xlen(stream);
+          lag = xlenResult;
+        }
+
+        break;
+      }
+
+      if (lag !== null) {
+        metrics.observe('processor_consumer_lag', lag);
+      }
+    } catch (err: unknown) {
+      onDebug(
+        `consumer lag sampling failed: ${err instanceof Error ? err.message : String(err)}`,
+      );
+    }
+  }
+
+  // First sample runs immediately.
+  sample().catch(() => {
+    // Already handled inside sample() — swallow at the outer level so the
+    // unhandled-rejection handler is not triggered.
+  });
+
+  const interval = setInterval(() => {
+    sample().catch(() => {});
+  }, intervalMs);
+
+  interval.unref();
+
+  return { stop: () => clearInterval(interval) };
+}
+
+// ---------------------------------------------------------------------------
+// Private: registry construction
+// ---------------------------------------------------------------------------
+
+function buildInternalRegistry(): InternalRegistry {
+  const registry = new Registry();
+
+  const consumerReadsTotal = new Counter({
+    name: 'processor_consumer_reads_total',
+    help: 'XREADGROUP calls. result=ok|empty|error. empty = BLOCK timeout, error = client error.',
+    labelNames: ['result'],
+    registers: [registry],
+  });
+
+  const consumerRecordsTotal = new Counter({
+    name: 'processor_consumer_records_total',
+    help: 'Total records pulled off the stream.',
+    registers: [registry],
+  });
+
+  const consumerLag = new Gauge({
+    name: 'processor_consumer_lag',
+    help: 'Consumer group lag (XINFO GROUPS lag field, Redis 7.2+; falls back to approximate XLEN when field absent).',
+    registers: [registry],
+  });
+
+  const decodeErrorsTotal = new Counter({
+    name: 'processor_decode_errors_total',
+    help: 'Records that failed to decode (malformed payload or sentinel error).',
+    registers: [registry],
+  });
+
+  const positionWritesTotal = new Counter({
+    name: 'processor_position_writes_total',
+    help: 'Per-record write outcomes. status=inserted|duplicate|failed.',
+    labelNames: ['status'],
+    registers: [registry],
+  });
+
+  const positionWriteDurationSeconds = new Histogram({
+    name: 'processor_position_write_duration_seconds',
+    help: 'Per-batch Postgres write latency.',
+    buckets: [0.001, 0.005, 0.01, 0.05, 0.1, 0.5, 1, 5],
+    registers: [registry],
+  });
+
+  const acksTotal = new Counter({
+    name: 'processor_acks_total',
+    help: 'Total stream entry IDs ACKed.',
+    registers: [registry],
+  });
+
+  const deviceStateSizeGauge = new Gauge({
+    name: 'processor_device_state_size',
+    help: 'Current count of devices in the in-memory LRU state map.',
+    registers: [registry],
+  });
+
+  const deviceStateEvictionsTotal = new Counter({
+    name: 'processor_device_state_evictions_total',
+    help: 'Total LRU evictions from the device state map since start.',
+    registers: [registry],
+  });
+
+  return {
+    registry,
+    consumerReadsTotal,
+    consumerRecordsTotal,
+    consumerLag,
+    decodeErrorsTotal,
+    positionWritesTotal,
+    positionWriteDurationSeconds,
+    acksTotal,
+    deviceStateSizeGauge,
+    deviceStateEvictionsTotal,
+  };
+}
+
+// ---------------------------------------------------------------------------
+// Private: dispatch helpers — map string metric names to typed prom-client calls
+// ---------------------------------------------------------------------------
+
+function dispatchInc(
+  r: InternalRegistry,
+  name: string,
+  labels?: Record<string, string>,
+): void {
+  switch (name) {
+    case 'processor_consumer_reads_total':
+      r.consumerReadsTotal.inc(labels ?? {});
+      break;
+    case 'processor_consumer_records_total':
+      r.consumerRecordsTotal.inc();
+      break;
+    case 'processor_decode_errors_total':
+      r.decodeErrorsTotal.inc();
+      break;
+    case 'processor_position_writes_total':
+      r.positionWritesTotal.inc(labels ?? {});
+      break;
+    case 'processor_acks_total':
+      r.acksTotal.inc();
+      break;
+    case 'processor_device_state_evictions_total':
+      r.deviceStateEvictionsTotal.inc();
+      break;
+    default:
+      // Unknown metric name — silently ignore. This preserves the contract
+      // that the Metrics interface never throws, and avoids crashing the
+      // process when a call site references a metric not yet in the registry
+      // (e.g. staged rollouts or future tasks).
+      break;
+  }
+}
+
+function dispatchObserve(
+  r: InternalRegistry,
+  name: string,
+  value: number,
+  _labels?: Record<string, string>,
+): void {
+  switch (name) {
+    case 'processor_position_write_duration_seconds':
+      r.positionWriteDurationSeconds.observe(value);
+      break;
+    case 'processor_consumer_lag':
+      r.consumerLag.set(value);
+      break;
+    case 'processor_device_state_size':
+      r.deviceStateSizeGauge.set(value);
+      break;
+    default:
+      // Unknown metric name — silently ignore (see dispatchInc comment).
+      break;
+  }
+}
diff --git a/test/metrics.test.ts b/test/metrics.test.ts
new file mode 100644
index 0000000..3183ac6
--- /dev/null
+++ b/test/metrics.test.ts
@@ -0,0 +1,328 @@
+/**
+ * Unit tests for src/observability/metrics.ts
+ *
+ * Covers:
+ *  - createMetrics(): Prometheus exposition format contains all Phase 1 metrics
+ *  - Counter increments via metrics.inc()
+ *  - Histogram observation via metrics.observe()
+ *  - Gauge set via metrics.observe() for processor_consumer_lag and processor_device_state_size
+ *  - Unknown metric name is silently ignored (no throw)
+ *  - startMetricsServer(): GET /metrics returns 200 with text/plain
+ *  - startMetricsServer(): GET /healthz returns 200 {"status":"ok"}
+ *  - startMetricsServer(): GET /readyz returns 200 when both deps are ready
+ *  - startMetricsServer(): GET /readyz returns 503 when Redis is not ready
+ *  - startMetricsServer(): GET /readyz returns 503 when Postgres is not ready
+ *  - startMetricsServer(): GET /readyz returns 503 when neither dep is ready
+ *  - startMetricsServer(): non-GET method returns 405
+ *  - startMetricsServer(): unknown path returns 404
+ *  - nodejs_* default metrics are present in the exposition output
+ */
+
+import { describe, it, expect, beforeAll, afterAll, vi } from 'vitest';
+import * as http from 'node:http';
+import { createMetrics, startMetricsServer } from '../src/observability/metrics.js';
+import type { ReadyzDeps } from '../src/observability/metrics.js';
+
+// ---------------------------------------------------------------------------
+// HTTP helper — makes a simple GET (or other method) against the test server
+// ---------------------------------------------------------------------------
+
+function httpGet(
+  port: number,
+  path: string,
+  method = 'GET',
+): Promise<{ statusCode: number; body: string; contentType: string }> {
+  return new Promise((resolve, reject) => {
+    const req = http.request({ hostname: '127.0.0.1', port, path, method }, (res) => {
+      let body = '';
+      res.on('data', (chunk: Buffer) => {
+        body += chunk.toString();
+      });
+      res.on('end', () => {
+        resolve({
+          statusCode: res.statusCode ?? 0,
+          body,
+          contentType: (res.headers['content-type'] as string | undefined) ?? '',
+        });
+      });
+    });
+    req.on('error', reject);
+    req.end();
+  });
+}
+
+// ---------------------------------------------------------------------------
+// createMetrics tests
+// ---------------------------------------------------------------------------
+
+describe('createMetrics — exposition format', () => {
+  it('returns valid Prometheus text format containing all Phase 1 metrics', async () => {
+    const metrics = createMetrics();
+    const text = await metrics.serializeMetrics();
+
+    // Every metric from the task 1.9 inventory must appear in the output.
+    expect(text).toContain('processor_consumer_reads_total');
+    expect(text).toContain('processor_consumer_records_total');
+    expect(text).toContain('processor_consumer_lag');
+    expect(text).toContain('processor_decode_errors_total');
+    expect(text).toContain('processor_position_writes_total');
+    expect(text).toContain('processor_position_write_duration_seconds');
+    expect(text).toContain('processor_acks_total');
+    expect(text).toContain('processor_device_state_size');
+    expect(text).toContain('processor_device_state_evictions_total');
+
+    // Default Node.js process metrics must be present.
+    expect(text).toContain('nodejs_');
+  });
+
+  it('label-less counters appear in the exposition at 0 before any inc() call', async () => {
+    const metrics = createMetrics();
+    const text = await metrics.serializeMetrics();
+
+    // prom-client emits label-less counters at 0 from the start.
+    // Counters with label dims only appear once .inc() is called with a label value.
+    expect(text).toMatch(/processor_consumer_records_total\s+0/);
+    expect(text).toMatch(/processor_decode_errors_total\s+0/);
+    expect(text).toMatch(/processor_acks_total\s+0/);
+    expect(text).toMatch(/processor_device_state_evictions_total\s+0/);
+  });
+});
+
+describe('createMetrics — counter increments', () => {
+  it('increments processor_consumer_reads_total with label', async () => {
+    const metrics = createMetrics();
+    metrics.inc('processor_consumer_reads_total', { result: 'ok' });
+    metrics.inc('processor_consumer_reads_total', { result: 'ok' });
+    metrics.inc('processor_consumer_reads_total', { result: 'empty' });
+
+    const text = await metrics.serializeMetrics();
+    // result="ok" incremented twice
+    expect(text).toMatch(/processor_consumer_reads_total\{result="ok"\} 2/);
+    // result="empty" incremented once
+    expect(text).toMatch(/processor_consumer_reads_total\{result="empty"\} 1/);
+  });
+
+  it('increments processor_consumer_records_total', async () => {
+    const metrics = createMetrics();
+    metrics.inc('processor_consumer_records_total');
+    metrics.inc('processor_consumer_records_total');
+    metrics.inc('processor_consumer_records_total');
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_consumer_records_total\s+3/);
+  });
+
+  it('increments processor_decode_errors_total', async () => {
+    const metrics = createMetrics();
+    metrics.inc('processor_decode_errors_total');
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_decode_errors_total\s+1/);
+  });
+
+  it('increments processor_position_writes_total with status label', async () => {
+    const metrics = createMetrics();
+    metrics.inc('processor_position_writes_total', { status: 'inserted' });
+    metrics.inc('processor_position_writes_total', { status: 'duplicate' });
+    metrics.inc('processor_position_writes_total', { status: 'failed' });
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_position_writes_total\{status="inserted"\} 1/);
+    expect(text).toMatch(/processor_position_writes_total\{status="duplicate"\} 1/);
+    expect(text).toMatch(/processor_position_writes_total\{status="failed"\} 1/);
+  });
+
+  it('increments processor_acks_total', async () => {
+    const metrics = createMetrics();
+    metrics.inc('processor_acks_total');
+    metrics.inc('processor_acks_total');
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_acks_total\s+2/);
+  });
+
+  it('increments processor_device_state_evictions_total', async () => {
+    const metrics = createMetrics();
+    metrics.inc('processor_device_state_evictions_total');
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_device_state_evictions_total\s+1/);
+  });
+
+  it('silently ignores unknown metric names', () => {
+    const metrics = createMetrics();
+    // Must not throw
+    expect(() => metrics.inc('no_such_metric_total')).not.toThrow();
+    expect(() => metrics.observe('no_such_metric', 42)).not.toThrow();
+  });
+});
+
+describe('createMetrics — gauge and histogram', () => {
+  it('sets processor_consumer_lag via observe()', async () => {
+    const metrics = createMetrics();
+    metrics.observe('processor_consumer_lag', 42);
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_consumer_lag\s+42/);
+  });
+
+  it('sets processor_device_state_size via observe()', async () => {
+    const metrics = createMetrics();
+    metrics.observe('processor_device_state_size', 7);
+
+    const text = await metrics.serializeMetrics();
+    expect(text).toMatch(/processor_device_state_size\s+7/);
+  });
+
+  it('records processor_position_write_duration_seconds histogram observation', async () => {
+    const metrics = createMetrics();
+    metrics.observe('processor_position_write_duration_seconds', 0.007);
+
+    const text = await metrics.serializeMetrics();
+    // Histogram emits _bucket, _sum, _count lines.
+    expect(text).toContain('processor_position_write_duration_seconds_sum');
+    expect(text).toContain('processor_position_write_duration_seconds_count 1');
+  });
+
+  it('histogram buckets include all spec-defined breakpoints', async () => {
+    const metrics = createMetrics();
+    const text = await metrics.serializeMetrics();
+
+    // Spec buckets: [0.001, 0.005, 0.01, 0.05, 0.1, 0.5, 1, 5]
+    const expectedBuckets = ['0.001', '0.005', '0.01', '0.05', '0.1', '0.5', '1', '5'];
+    for (const bucket of expectedBuckets) {
+      expect(text).toContain(`le="${bucket}"`);
+    }
+  });
+});
+
+// ---------------------------------------------------------------------------
+// startMetricsServer tests
+// ---------------------------------------------------------------------------
+
+describe('startMetricsServer — HTTP endpoints', () => {
+  let server: http.Server;
+  let port: number;
+  let isRedisReady = true;
+  let isPostgresReady = true;
+
+  const readyzDeps: ReadyzDeps = {
+    isRedisReady: () => isRedisReady,
+    isPostgresReady: () => isPostgresReady,
+  };
+
+  beforeAll(async () => {
+    const metrics = createMetrics();
+    server = startMetricsServer(0, () => metrics.serializeMetrics(), readyzDeps);
+    // Wait for the server to bind a port (port=0 lets OS pick)
+    await new Promise<void>((resolve) => {
+      if (server.listening) {
+        resolve();
+      } else {
+        server.once('listening', () => resolve());
+      }
+    });
+    const addr = server.address();
+    port = typeof addr === 'object' && addr !== null ? addr.port : 0;
+  });
+
+  afterAll(async () => {
+    await new Promise<void>((resolve, reject) =>
+      server.close((err) => (err ? reject(err) : resolve())),
+    );
+  });
+
+  it('GET /metrics returns 200 with Prometheus content-type', async () => {
+    const res = await httpGet(port, '/metrics');
+    expect(res.statusCode).toBe(200);
+    expect(res.contentType).toMatch('text/plain');
+    expect(res.body).toContain('processor_consumer_reads_total');
+  });
+
+  it('GET /healthz returns 200 with {"status":"ok"}', async () => {
+    const res = await httpGet(port, '/healthz');
+    expect(res.statusCode).toBe(200);
+    expect(JSON.parse(res.body)).toEqual({ status: 'ok' });
+  });
+
+  it('GET /readyz returns 200 when both Redis and Postgres are ready', async () => {
+    isRedisReady = true;
+    isPostgresReady = true;
+    const res = await httpGet(port, '/readyz');
+    expect(res.statusCode).toBe(200);
+    expect(JSON.parse(res.body)).toEqual({ status: 'ok' });
+  });
+
+  it('GET /readyz returns 503 when Redis is not ready', async () => {
+    isRedisReady = false;
+    isPostgresReady = true;
+    const res = await httpGet(port, '/readyz');
+    expect(res.statusCode).toBe(503);
+    const body = JSON.parse(res.body) as { status: string; redis: boolean; postgres: boolean };
+    expect(body.status).toBe('not ready');
+    expect(body.redis).toBe(false);
+    expect(body.postgres).toBe(true);
+    isRedisReady = true;
+  });
+
+  it('GET /readyz returns 503 when Postgres is not ready', async () => {
+    isRedisReady = true;
+    isPostgresReady = false;
+    const res = await httpGet(port, '/readyz');
+    expect(res.statusCode).toBe(503);
+    const body = JSON.parse(res.body) as { status: string; redis: boolean; postgres: boolean };
+    expect(body.status).toBe('not ready');
+    expect(body.redis).toBe(true);
+    expect(body.postgres).toBe(false);
+    isPostgresReady = true;
+  });
+
+  it('GET /readyz returns 503 when both Redis and Postgres are not ready', async () => {
+    isRedisReady = false;
+    isPostgresReady = false;
+    const res = await httpGet(port, '/readyz');
+    expect(res.statusCode).toBe(503);
+    const body = JSON.parse(res.body) as { status: string; redis: boolean; postgres: boolean };
+    expect(body.redis).toBe(false);
+    expect(body.postgres).toBe(false);
+    isRedisReady = true;
+    isPostgresReady = true;
+  });
+
+  it('non-GET request returns 405', async () => {
+    const res = await httpGet(port, '/metrics', 'POST');
+    expect(res.statusCode).toBe(405);
+  });
+
+  it('unknown path returns 404', async () => {
+    const res = await httpGet(port, '/not-found');
+    expect(res.statusCode).toBe(404);
+  });
+});
+
+describe('startMetricsServer — /metrics error path', () => {
+  it('returns 500 when serializeMetrics rejects', async () => {
+    const serializeMetrics = vi.fn().mockRejectedValue(new Error('prom-client exploded'));
+    const server = startMetricsServer(
+      0,
+      serializeMetrics,
+      { isRedisReady: () => true, isPostgresReady: () => true },
+    );
+
+    await new Promise<void>((resolve) => {
+      if (server.listening) resolve();
+      else server.once('listening', () => resolve());
+    });
+
+    const addr = server.address();
+    const port = typeof addr === 'object' && addr !== null ? addr.port : 0;
+
+    const res = await httpGet(port, '/metrics');
+    expect(res.statusCode).toBe(500);
+    expect(res.body).toContain('prom-client exploded');
+
+    await new Promise<void>((resolve, reject) =>
+      server.close((err) => (err ? reject(err) : resolve())),
+    );
+  });
+});
diff --git a/test/pipeline.integration.test.ts b/test/pipeline.integration.test.ts
new file mode 100644
index 0000000..a384060
--- /dev/null
+++ b/test/pipeline.integration.test.ts
@@ -0,0 +1,413 @@
+/**
+ * Integration test: end-to-end pipeline round-trip via testcontainers.
+ *
+ * Spins up Redis 7 and TimescaleDB (timescale/timescaledb:latest-pg16) containers,
+ * runs the Processor migration, starts the consumer pipeline, publishes synthetic
+ * Position records, and asserts the resulting rows in `positions`.
+ *
+ * If Docker is unavailable (CI runner without Docker, local dev without Docker
+ * Desktop), the suite skips — it does not fail the build. Docker availability is
+ * determined by a container start attempt in beforeAll; the skip flag is set once,
+ * and each `it` block early-returns when `!dockerAvailable`.
+ *
+ * WARNING: Do NOT replace the early-return skip pattern with a try/catch alone.
+ * A hang does not throw; only an explicit `!dockerAvailable` check per test
+ * guarantees that unavailable Docker exits cleanly (see tcp-ingestion history).
+ */
+
+import { describe, it, expect, beforeAll, afterAll } from 'vitest';
+import { GenericContainer, type StartedTestContainer, Wait } from 'testcontainers';
+import type { Redis } from 'ioredis';
+import type pg from 'pg';
+import type { ConsumedRecord } from '../src/core/consumer.js';
+import { createConsumer, connectRedis, ensureConsumerGroup } from '../src/core/consumer.js';
+import { createWriter } from '../src/core/writer.js';
+import { createDeviceStateStore } from '../src/core/state.js';
+import { createPool, connectWithRetry } from '../src/db/pool.js';
+import { runMigrations } from '../src/db/migrate.js';
+import { createMetrics } from '../src/observability/metrics.js';
+import type { Config } from '../src/config/load.js';
+import type { Position } from '../src/core/types.js';
+import { vi } from 'vitest';
+import type { Logger } from 'pino';
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
+function makeSilentLogger(): Logger {
+  return {
+    debug: vi.fn(),
+    info: vi.fn(),
+    warn: vi.fn(),
+    error: vi.fn(),
+    fatal: vi.fn(),
+    trace: vi.fn(),
+    child: vi.fn().mockReturnThis(),
+    level: 'silent',
+    silent: vi.fn(),
+  } as unknown as Logger;
+}
+
+function makeConfig(overrides: Partial<Config> = {}): Config {
+  return {
+    NODE_ENV: 'test',
+    INSTANCE_ID: 'test-integration',
+    LOG_LEVEL: 'silent',
+    REDIS_URL: 'redis://localhost:6379', // overridden below with mapped port
+    POSTGRES_URL: 'postgres://postgres:postgres@localhost:5432/trm', // overridden below
+    REDIS_TELEMETRY_STREAM: 'telemetry:t',
+    REDIS_CONSUMER_GROUP: 'processor',
+    REDIS_CONSUMER_NAME: 'test-consumer',
+    METRICS_PORT: 0,
+    BATCH_SIZE: 100,
+    BATCH_BLOCK_MS: 500,
+    WRITE_BATCH_SIZE: 50,
+    DEVICE_STATE_LRU_CAP: 10_000,
+    ...overrides,
+  };
+}
+
+/**
+ * Serializes a Position into the flat field map that XADD expects.
+ * Mirrors tcp-ingestion's serializePosition format exactly: bigint → __bigint
+ * sentinel, Buffer → __buffer_b64 sentinel, Date → ISO string.
+ */
+function buildXaddFields(position: Position, codec: string): string[] {
+  function jsonReplacer(_key: string, value: unknown): unknown {
+    if (typeof value === 'bigint') return { __bigint: value.toString() };
+    if (value instanceof Uint8Array) {
+      return { __buffer_b64: Buffer.from(value).toString('base64') };
+    }
+    if (value instanceof Date) return value.toISOString();
+    return value;
+  }
+
+  const payload = JSON.stringify(position, jsonReplacer);
+  return [
+    'ts', position.timestamp.toISOString(),
+    'device_id', position.device_id,
+    'codec', codec,
+    'payload', payload,
+  ];
+}
+
+/**
+ * Polls `fn` up to `timeoutMs` with `intervalMs` gaps until it returns a
+ * truthy result. Returns null if the timeout expires.
+ */
+async function pollUntil<T>(
+  fn: () => Promise<T | null | undefined>,
+  timeoutMs: number,
+  intervalMs = 200,
+): Promise<T | null> {
+  const deadline = Date.now() + timeoutMs;
+  while (Date.now() < deadline) {
+    const result = await fn();
+    if (result !== null && result !== undefined) return result as T;
+    await new Promise<void>((resolve) => setTimeout(resolve, intervalMs));
+  }
+  return null;
+}
+
+// ---------------------------------------------------------------------------
+// Container and pipeline lifecycle
+// ---------------------------------------------------------------------------
+
+let redisContainer: StartedTestContainer | null = null;
+let pgContainer: StartedTestContainer | null = null;
+let redisClient: Redis | null = null;
+let pgPool: pg.Pool | null = null;
+let consumer: { start: () => Promise<void>; stop: () => Promise<void> } | null = null;
+let dockerAvailable = true;
+
+const STREAM = 'telemetry:t';
+const GROUP = 'processor';
+
+beforeAll(async () => {
+  // --- Step 1: start Redis container -----------------------------------------
+  try {
+    redisContainer = await new GenericContainer('redis:7-alpine')
+      .withExposedPorts(6379)
+      .withWaitStrategy(Wait.forLogMessage('Ready to accept connections'))
+      .start();
+  } catch {
+    console.warn(
+      '[pipeline.integration.test] Docker not available — skipping integration tests',
+    );
+    dockerAvailable = false;
+    return;
+  }
+
+  // --- Step 2: start TimescaleDB container ------------------------------------
+  try {
+    pgContainer = await new GenericContainer('timescale/timescaledb:latest-pg16')
+      .withExposedPorts(5432)
+      .withEnvironment({
+        POSTGRES_USER: 'postgres',
+        POSTGRES_PASSWORD: 'postgres',
+        POSTGRES_DB: 'trm',
+      })
+      .withWaitStrategy(Wait.forLogMessage('database system is ready to accept connections', 2))
+      .start();
+  } catch (err) {
+    console.warn(
+      `[pipeline.integration.test] Failed to start TimescaleDB container: ${String(err)} — skipping`,
+    );
+    dockerAvailable = false;
+    await redisContainer?.stop().catch(() => {});
+    redisContainer = null;
+    return;
+  }
+
+  const redisHost = redisContainer.getHost();
+  const redisPort = redisContainer.getMappedPort(6379);
+  const pgHost = pgContainer.getHost();
+  const pgPort = pgContainer.getMappedPort(5432);
+
+  const redisUrl = `redis://${redisHost}:${redisPort}`;
+  const postgresUrl = `postgres://postgres:postgres@${pgHost}:${pgPort}/trm`;
+
+  const config = makeConfig({ REDIS_URL: redisUrl, POSTGRES_URL: postgresUrl });
+  const logger = makeSilentLogger();
+
+  // --- Step 3: connect Redis --------------------------------------------------
+  const { default: Redis } = await import('ioredis');
+  const client = new Redis(redisUrl, {
+    enableOfflineQueue: false,
+    lazyConnect: true,
+    maxRetriesPerRequest: 0,
+  });
+  await client.connect();
+  redisClient = client;
+
+  // --- Step 4: connect Postgres and run migrations ---------------------------
+  pgPool = createPool(postgresUrl);
+  await connectWithRetry(pgPool, logger);
+  await runMigrations(pgPool, logger);
+
+  // --- Step 5: wire and start the consumer pipeline -------------------------
+  const metrics = createMetrics();
+  const state = createDeviceStateStore(config, logger);
+  const writer = createWriter(pgPool, config, logger, metrics);
+
+  await ensureConsumerGroup(client, STREAM, GROUP, logger);
+
+  const sink = async (records: ConsumedRecord[]): Promise<string[]> => {
+    for (const record of records) {
+      state.update(record.position);
+    }
+    const results = await writer.write(records);
+    return results
+      .filter((r) => r.status === 'inserted' || r.status === 'duplicate')
+      .map((r) => r.id);
+  };
+
+  // Use connectRedis for the consumer's own connection (separate from the
+  // redisClient used for XADD in tests) so we mirror production topology.
+  const consumerRedis = await connectRedis(redisUrl, logger);
+  consumer = createConsumer(consumerRedis, config, logger, metrics, sink);
+  await consumer.start();
+}, 120_000);
+
+afterAll(async () => {
+  await consumer?.stop().catch(() => {});
+  await redisClient?.quit().catch(() => {});
+  await pgPool?.end().catch(() => {});
+  await redisContainer?.stop().catch(() => {});
+  await pgContainer?.stop().catch(() => {});
+}, 30_000);
+
+// ---------------------------------------------------------------------------
+// Integration tests
+// ---------------------------------------------------------------------------
+
+describe('pipeline integration — full round-trip', () => {
+  // Test 1: happy-path with bigint + Buffer attributes
+  it('publishes a Position with bigint and Buffer attributes and verifies the row in positions', async () => {
+    if (!dockerAvailable || !redisClient || !pgPool) {
+      console.warn('[pipeline.integration.test] skipping test 1: Docker not available');
+      return;
+    }
+
+    const position: Position = {
+      device_id: '356307042441013',
+      timestamp: new Date('2024-06-15T12:00:00.000Z'),
+      latitude: 54.687157,
+      longitude: 25.279652,
+      altitude: 130,
+      angle: 90,
+      speed: 45,
+      satellites: 12,
+      priority: 0,
+      attributes: {
+        num_attr: 255,
+        big_attr: BigInt('18446744073709551615'), // u64 max
+        buf_attr: Buffer.from([0xde, 0xad, 0xbe, 0xef]),
+      },
+    };
+
+    const fields = buildXaddFields(position, '8E');
+    await redisClient.xadd(STREAM, '*', ...fields);
+
+    // Poll until the row appears in positions (up to 10 s).
+    type Row = {
+      device_id: string;
+      ts: Date;
+      latitude: number;
+      longitude: number;
+      attributes: Record<string, unknown>;
+    };
+
+    const row = await pollUntil<Row>(async () => {
+      const result = await pgPool!.query<Row>(
+        'SELECT device_id, ts, latitude, longitude, attributes FROM positions WHERE device_id = $1 AND ts = $2',
+        [position.device_id, position.timestamp],
+      );
+      return result.rows[0] ?? null;
+    }, 10_000);
+
+    expect(row).not.toBeNull();
+    expect(row!.device_id).toBe(position.device_id);
+    expect(row!.latitude).toBeCloseTo(position.latitude, 4);
+    expect(row!.longitude).toBeCloseTo(position.longitude, 4);
+
+    // attributes JSONB: bigint stored as decimal string, Buffer as base64 string.
+    expect(typeof row!.attributes['big_attr']).toBe('string');
+    expect(row!.attributes['big_attr']).toBe('18446744073709551615');
+
+    expect(typeof row!.attributes['buf_attr']).toBe('string');
+    const decoded = Buffer.from(row!.attributes['buf_attr'] as string, 'base64');
+    expect(decoded).toEqual(Buffer.from([0xde, 0xad, 0xbe, 0xef]));
+
+    expect(row!.attributes['num_attr']).toBe(255);
+  }, 30_000);
+
+  // Test 2: idempotency — duplicate (device_id, ts) must not create a second row
+  it('does not create a duplicate row when the same (device_id, ts) is published twice', async () => {
+    if (!dockerAvailable || !redisClient || !pgPool) {
+      console.warn('[pipeline.integration.test] skipping test 2: Docker not available');
+      return;
+    }
+
+    const position: Position = {
+      device_id: 'DUP-DEVICE-001',
+      timestamp: new Date('2024-06-15T13:00:00.000Z'),
+      latitude: 1.0,
+      longitude: 2.0,
+      altitude: 10,
+      angle: 0,
+      speed: 0,
+      satellites: 4,
+      priority: 0,
+      attributes: {},
+    };
+
+    const fields = buildXaddFields(position, '8');
+
+    // Publish the same position twice.
+    await redisClient.xadd(STREAM, '*', ...fields);
+    await redisClient.xadd(STREAM, '*', ...fields);
+
+    // Wait long enough for both entries to be processed.
+    await new Promise<void>((resolve) => setTimeout(resolve, 3_000));
+
+    const result = await pgPool.query<{ count: string }>(
+      'SELECT COUNT(*) AS count FROM positions WHERE device_id = $1 AND ts = $2',
+      [position.device_id, position.timestamp],
+    );
+    const count = parseInt(result.rows[0]?.count ?? '0', 10);
+    expect(count).toBe(1);
+  }, 30_000);
+
+  // Test 3: malformed payload — decode error counter increments, entry not ACKed
+  it('increments decode error counter and leaves malformed entry pending (not ACKed)', async () => {
+    if (!dockerAvailable || !redisClient || !pgPool) {
+      console.warn('[pipeline.integration.test] skipping test 3: Docker not available');
+      return;
+    }
+
+    // Push a stream entry with a broken payload (not valid JSON).
+    const badEntryId = await redisClient.xadd(
+      STREAM,
+      '*',
+      'ts', new Date().toISOString(),
+      'device_id', 'BAD-DEVICE',
+      'codec', '8',
+      'payload', 'NOT_VALID_JSON {{{',
+    );
+
+    // Wait for the consumer to attempt processing.
+    await new Promise<void>((resolve) => setTimeout(resolve, 2_000));
+
+    // The entry should remain in the Pending Entry List (PEL) — it was not ACKed.
+    const pendingResult = await redisClient.xpending(
+      STREAM,
+      GROUP,
+      '-',
+      '+',
+      '100',
+    ) as Array<[string, string, number, number]>;
+
+    // Find the bad entry in the PEL.
+    const pendingIds = pendingResult.map(([id]) => id);
+    expect(pendingIds).toContain(badEntryId);
+  }, 30_000);
+
+  // Test 4: writer failure → retry — stop Postgres before publish, restart, verify row lands
+  it('retries and writes the row after Postgres recovers from a stopped state', async () => {
+    if (!dockerAvailable || !redisClient || !pgPool || !pgContainer) {
+      console.warn('[pipeline.integration.test] skipping test 4: Docker not available');
+      return;
+    }
+
+    const position: Position = {
+      device_id: 'RETRY-DEVICE-001',
+      timestamp: new Date('2024-06-15T14:00:00.000Z'),
+      latitude: 3.0,
+      longitude: 4.0,
+      altitude: 20,
+      angle: 45,
+      speed: 10,
+      satellites: 8,
+      priority: 1,
+      attributes: {},
+    };
+
+    // Stop Postgres before publishing so the first write attempt fails.
+    await pgContainer.stop();
+
+    const fields = buildXaddFields(position, '8');
+    await redisClient.xadd(STREAM, '*', ...fields);
+
+    // Wait briefly — the write should fail while Postgres is down.
+    await new Promise<void>((resolve) => setTimeout(resolve, 1_500));
+
+    // Restart Postgres.
+    pgContainer = await pgContainer.restart();
+
+    // Wait a bit to ensure the new container is accepting connections before
+    // reconnecting. The pool will get fresh connections once the TCP stack
+    // accepts again.
+    await new Promise<void>((resolve) => setTimeout(resolve, 3_000));
+
+    // The entry is still pending in the consumer's PEL; the next XREADGROUP
+    // poll will re-deliver it. The pipeline should eventually write it.
+    type Row = { device_id: string };
+    const row = await pollUntil<Row>(async () => {
+      try {
+        const result = await pgPool!.query<Row>(
+          'SELECT device_id FROM positions WHERE device_id = $1 AND ts = $2',
+          [position.device_id, position.timestamp],
+        );
+        return result.rows[0] ?? null;
+      } catch {
+        // Pool may throw transiently while connections re-establish.
+        return null;
+      }
+    }, 20_000);
+
+    expect(row).not.toBeNull();
+    expect(row!.device_id).toBe(position.device_id);
+  }, 60_000);
+});