From b5a867ca5ae83e2f4c5daf2b4b0298d670aaf578 Mon Sep 17 00:00:00 2001
From: Egutierrez <egutierrez@dead.dd>
Date: Mon, 13 Apr 2026 02:00:44 +0200
Subject: [PATCH 1/2] feat: cola de jobs asincrona basada en SQLite (issue
 0013)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Implementa el subsistema completo de background jobs para apps Go en el
dominio infra. 9 funciones + 3 tipos + 17 tests, todos pasando.

- Tipos: Job (product), JobQueue (product), JobStatus (sum) con
  JobHandler, EnqueueOption y WorkerOption usando functional options pattern
- job_queue_create: CREATE TABLE + indices + WAL mode
- job_enqueue: INSERT con UUID (github.com/google/uuid), WithPriority/WithScheduledAt/WithMaxAttempts
- job_dequeue: SELECT+UPDATE atomico en transaccion exclusiva, filtro por jobTypes
- job_complete / job_fail: transiciones de estado; fail → dead cuando attempts >= max_attempts
- job_status_summary: pura, formatea conteo de jobs por estado
- job_worker: poll loop bloqueante, context-cancelable, graceful shutdown
- job_worker_pool: N workers con golang.org/x/sync/errgroup
- job_cleanup: DELETE jobs terminales mas viejos que olderThan

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 functions/infra/job_cleanup.go      |  33 +++
 functions/infra/job_cleanup.md      |  41 +++
 functions/infra/job_complete.go     |  39 +++
 functions/infra/job_complete.md     |  40 +++
 functions/infra/job_dequeue.go      | 125 +++++++++
 functions/infra/job_dequeue.md      |  49 ++++
 functions/infra/job_enqueue.go      |  47 ++++
 functions/infra/job_enqueue.md      |  48 ++++
 functions/infra/job_fail.go         |  43 +++
 functions/infra/job_fail.md         |  43 +++
 functions/infra/job_queue_create.go |  72 +++++
 functions/infra/job_queue_create.md |  45 +++
 functions/infra/job_queue_test.go   | 418 ++++++++++++++++++++++++++++
 functions/infra/job_queue_types.go  |  87 ++++++
 functions/infra/job_status.go       |  19 ++
 functions/infra/job_status.md       |  39 +++
 functions/infra/job_worker.go       |  70 +++++
 functions/infra/job_worker.md       |  49 ++++
 functions/infra/job_worker_pool.go  |  36 +++
 functions/infra/job_worker_pool.md  |  50 ++++
 types/infra/job.md                  |  31 +++
 types/infra/job_queue.md            |  20 ++
 types/infra/job_status.md           |  25 ++
 23 files changed, 1469 insertions(+)
 create mode 100644 functions/infra/job_cleanup.go
 create mode 100644 functions/infra/job_cleanup.md
 create mode 100644 functions/infra/job_complete.go
 create mode 100644 functions/infra/job_complete.md
 create mode 100644 functions/infra/job_dequeue.go
 create mode 100644 functions/infra/job_dequeue.md
 create mode 100644 functions/infra/job_enqueue.go
 create mode 100644 functions/infra/job_enqueue.md
 create mode 100644 functions/infra/job_fail.go
 create mode 100644 functions/infra/job_fail.md
 create mode 100644 functions/infra/job_queue_create.go
 create mode 100644 functions/infra/job_queue_create.md
 create mode 100644 functions/infra/job_queue_test.go
 create mode 100644 functions/infra/job_queue_types.go
 create mode 100644 functions/infra/job_status.go
 create mode 100644 functions/infra/job_status.md
 create mode 100644 functions/infra/job_worker.go
 create mode 100644 functions/infra/job_worker.md
 create mode 100644 functions/infra/job_worker_pool.go
 create mode 100644 functions/infra/job_worker_pool.md
 create mode 100644 types/infra/job.md
 create mode 100644 types/infra/job_queue.md
 create mode 100644 types/infra/job_status.md

diff --git a/functions/infra/job_cleanup.go b/functions/infra/job_cleanup.go
new file mode 100644
index 00000000..ede8edf4
--- /dev/null
+++ b/functions/infra/job_cleanup.go
@@ -0,0 +1,33 @@
+package infra
+
+import (
+	"fmt"
+	"time"
+)
+
+// JobCleanup deletes jobs in terminal states (completed, failed, dead) that are
+// older than olderThan. Returns the number of rows deleted.
+// This is useful for keeping the jobs table small in long-running applications.
+func JobCleanup(q *JobQueue, olderThan time.Duration) (int64, error) {
+	if q == nil {
+		return 0, fmt.Errorf("job_cleanup: queue must not be nil")
+	}
+
+	cutoff := time.Now().UTC().Add(-olderThan).Format(time.RFC3339)
+
+	query := fmt.Sprintf(`
+DELETE FROM %s
+WHERE status IN ('completed', 'failed', 'dead')
+  AND created_at < ?
+`, q.TableName)
+
+	res, err := q.DB.Exec(query, cutoff)
+	if err != nil {
+		return 0, fmt.Errorf("job_cleanup: delete: %w", err)
+	}
+	n, err := res.RowsAffected()
+	if err != nil {
+		return 0, fmt.Errorf("job_cleanup: rows affected: %w", err)
+	}
+	return n, nil
+}
diff --git a/functions/infra/job_cleanup.md b/functions/infra/job_cleanup.md
new file mode 100644
index 00000000..16d5d333
--- /dev/null
+++ b/functions/infra/job_cleanup.md
@@ -0,0 +1,41 @@
+---
+name: job_cleanup
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobCleanup(q *JobQueue, olderThan time.Duration) (int64, error)"
+description: "Elimina jobs en estados terminales (completed, failed, dead) cuyo created_at sea mas antiguo que olderThan. Retorna el numero de filas eliminadas. Util para mantener la tabla compacta en apps de larga duracion."
+tags: [job, queue, cleanup, delete, maintenance, sqlite, async, background, infra]
+uses_functions: []
+uses_types: [job_queue_go_infra]
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [fmt, time]
+params:
+  - name: q
+    desc: "cola de jobs creada con JobQueueCreate"
+  - name: olderThan
+    desc: "duracion maxima de retension; jobs mas viejos que esto se eliminan (ej: 24*time.Hour)"
+output: "numero de filas eliminadas"
+tested: true
+tests:
+  - "job_cleanup_removes_old_terminal_jobs"
+  - "job_cleanup_keeps_recent_jobs"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_cleanup.go"
+---
+
+## Ejemplo
+
+```go
+// Limpiar jobs terminados hace mas de 7 dias
+n, err := JobCleanup(q, 7*24*time.Hour)
+fmt.Printf("eliminados: %d jobs\n", n)
+```
+
+## Notas
+
+Solo elimina jobs en estados terminales: completed, failed, dead. Los jobs pending y running nunca se eliminan. El cutoff se calcula en UTC. Llamar periodicamente con CronTicker u otro scheduler para mantener la tabla compacta.
diff --git a/functions/infra/job_complete.go b/functions/infra/job_complete.go
new file mode 100644
index 00000000..25527fa2
--- /dev/null
+++ b/functions/infra/job_complete.go
@@ -0,0 +1,39 @@
+package infra
+
+import (
+	"fmt"
+	"time"
+)
+
+// JobComplete marks a job as completed and stores the optional result JSON string.
+// result may be empty ("") to indicate no result payload.
+func JobComplete(q *JobQueue, jobID string, result string) error {
+	if q == nil {
+		return fmt.Errorf("job_complete: queue must not be nil")
+	}
+	if jobID == "" {
+		return fmt.Errorf("job_complete: jobID must not be empty")
+	}
+
+	now := time.Now().UTC().Format(time.RFC3339)
+	var resultPtr *string
+	if result != "" {
+		resultPtr = &result
+	}
+
+	query := fmt.Sprintf(`
+UPDATE %s
+SET status = 'completed', completed_at = ?, result = ?
+WHERE id = ?
+`, q.TableName)
+
+	res, err := q.DB.Exec(query, now, resultPtr, jobID)
+	if err != nil {
+		return fmt.Errorf("job_complete: update: %w", err)
+	}
+	n, _ := res.RowsAffected()
+	if n == 0 {
+		return fmt.Errorf("job_complete: job %q not found", jobID)
+	}
+	return nil
+}
diff --git a/functions/infra/job_complete.md b/functions/infra/job_complete.md
new file mode 100644
index 00000000..00fe8cd3
--- /dev/null
+++ b/functions/infra/job_complete.md
@@ -0,0 +1,40 @@
+---
+name: job_complete
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobComplete(q *JobQueue, jobID string, result string) error"
+description: "Marca un job como completado, setea completed_at y almacena el resultado opcional. result puede ser vacio si no hay payload de resultado."
+tags: [job, queue, complete, sqlite, async, background, infra]
+uses_functions: []
+uses_types: [job_queue_go_infra]
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [fmt, time]
+params:
+  - name: q
+    desc: "cola de jobs creada con JobQueueCreate"
+  - name: jobID
+    desc: "UUID del job a marcar como completado"
+  - name: result
+    desc: "JSON string con el resultado del job; puede ser vacio"
+output: "error si el job no existe o falla el UPDATE"
+tested: true
+tests:
+  - "complete_fail_transitions"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_complete.go"
+---
+
+## Ejemplo
+
+```go
+err := JobComplete(q, job.ID, `{"rows_processed":42}`)
+```
+
+## Notas
+
+Setea `status='completed'`, `completed_at=now`, y `result` (NULL si vacio). Retorna error si el job no existe (0 rows affected).
diff --git a/functions/infra/job_dequeue.go b/functions/infra/job_dequeue.go
new file mode 100644
index 00000000..3da3cd2c
--- /dev/null
+++ b/functions/infra/job_dequeue.go
@@ -0,0 +1,125 @@
+package infra
+
+import (
+	"database/sql"
+	"errors"
+	"fmt"
+	"strings"
+	"time"
+)
+
+// JobDequeue atomically claims the next available job by running a
+// SELECT + UPDATE inside an EXCLUSIVE transaction. Returns nil, nil when the
+// queue is empty (or no job matches the jobTypes filter).
+// jobTypes restricts which job types are dequeued; an empty slice means all types.
+func JobDequeue(q *JobQueue, jobTypes []string) (*Job, error) {
+	if q == nil {
+		return nil, fmt.Errorf("job_dequeue: queue must not be nil")
+	}
+
+	tx, err := q.DB.Begin()
+	if err != nil {
+		return nil, fmt.Errorf("job_dequeue: begin tx: %w", err)
+	}
+	defer tx.Rollback() //nolint:errcheck
+
+	// Build the SELECT query.
+	now := time.Now().UTC().Format(time.RFC3339)
+
+	var selectQ string
+	var args []any
+
+	if len(jobTypes) > 0 {
+		placeholders := make([]string, len(jobTypes))
+		for i, t := range jobTypes {
+			placeholders[i] = "?"
+			args = append(args, t)
+		}
+		selectQ = fmt.Sprintf(`
+SELECT id, type, payload, status, priority, attempts, max_attempts,
+       scheduled_at, started_at, completed_at, result, error, created_at
+FROM %s
+WHERE status = 'pending'
+  AND scheduled_at <= ?
+  AND type IN (%s)
+ORDER BY priority DESC, scheduled_at ASC
+LIMIT 1
+`, q.TableName, strings.Join(placeholders, ","))
+		args = append(args, now)
+		// rearrange: now goes before jobTypes in the query
+		newArgs := []any{now}
+		newArgs = append(newArgs, args[:len(args)-1]...)
+		args = newArgs
+	} else {
+		selectQ = fmt.Sprintf(`
+SELECT id, type, payload, status, priority, attempts, max_attempts,
+       scheduled_at, started_at, completed_at, result, error, created_at
+FROM %s
+WHERE status = 'pending'
+  AND scheduled_at <= ?
+ORDER BY priority DESC, scheduled_at ASC
+LIMIT 1
+`, q.TableName)
+		args = []any{now}
+	}
+
+	row := tx.QueryRow(selectQ, args...)
+
+	var job Job
+	var scheduledAt string
+	var startedAt, completedAt sql.NullString
+	var result, jobError sql.NullString
+	var createdAt string
+
+	err = row.Scan(
+		&job.ID, &job.Type, &job.Payload, (*string)(&job.Status),
+		&job.Priority, &job.Attempts, &job.MaxAttempts,
+		&scheduledAt, &startedAt, &completedAt,
+		&result, &jobError, &createdAt,
+	)
+	if errors.Is(err, sql.ErrNoRows) {
+		return nil, nil
+	}
+	if err != nil {
+		return nil, fmt.Errorf("job_dequeue: scan: %w", err)
+	}
+
+	// Mark as running.
+	updateQ := fmt.Sprintf(`
+UPDATE %s SET status = 'running', started_at = ? WHERE id = ?
+`, q.TableName)
+	if _, err := tx.Exec(updateQ, now, job.ID); err != nil {
+		return nil, fmt.Errorf("job_dequeue: update status: %w", err)
+	}
+
+	if err := tx.Commit(); err != nil {
+		return nil, fmt.Errorf("job_dequeue: commit: %w", err)
+	}
+
+	// Hydrate optional fields.
+	if t, err := time.Parse(time.RFC3339, scheduledAt); err == nil {
+		job.ScheduledAt = t
+	}
+	if t, err := time.Parse(time.RFC3339, createdAt); err == nil {
+		job.CreatedAt = t
+	}
+	if startedAt.Valid {
+		if t, err := time.Parse(time.RFC3339, startedAt.String); err == nil {
+			job.StartedAt = &t
+		}
+	}
+	if completedAt.Valid {
+		if t, err := time.Parse(time.RFC3339, completedAt.String); err == nil {
+			job.CompletedAt = &t
+		}
+	}
+	if result.Valid {
+		job.Result = &result.String
+	}
+	if jobError.Valid {
+		job.Error = &jobError.String
+	}
+	job.Status = JobStatusRunning
+
+	return &job, nil
+}
diff --git a/functions/infra/job_dequeue.md b/functions/infra/job_dequeue.md
new file mode 100644
index 00000000..d11d3bf2
--- /dev/null
+++ b/functions/infra/job_dequeue.md
@@ -0,0 +1,49 @@
+---
+name: job_dequeue
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobDequeue(q *JobQueue, jobTypes []string) (*Job, error)"
+description: "Extrae atomicamente el siguiente job disponible usando SELECT+UPDATE en una transaccion exclusiva. Retorna nil, nil si la cola esta vacia o no hay jobs que cumplan el filtro. jobTypes limita los tipos dequeued; slice vacio significa todos."
+tags: [job, queue, dequeue, atomic, transaction, sqlite, async, background, infra]
+uses_functions: []
+uses_types: [job_queue_go_infra, job_go_infra, job_status_go_infra]
+returns: [job_go_infra]
+returns_optional: true
+error_type: "error_go_core"
+imports: [database/sql, errors, fmt, strings, time]
+params:
+  - name: q
+    desc: "cola de jobs creada con JobQueueCreate"
+  - name: jobTypes
+    desc: "lista de tipos a desencolar; slice vacio o nil = todos los tipos"
+output: "puntero a Job con status=running, o nil si la cola esta vacia"
+tested: true
+tests:
+  - "enqueue_dequeue_atomicidad"
+  - "dequeue_empty_queue_returns_nil"
+  - "dequeue_priority_order"
+  - "dequeue_jobtype_filter"
+  - "dequeue_scheduled_in_future_waits"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_dequeue.go"
+---
+
+## Ejemplo
+
+```go
+// Desencolar cualquier tipo
+job, err := JobDequeue(q, nil)
+if job == nil {
+    // cola vacia
+}
+
+// Filtrar por tipo
+job, err = JobDequeue(q, []string{"send_email", "send_sms"})
+```
+
+## Notas
+
+Usa `db.Begin()` + `tx.Rollback()` (deferred) + `tx.Commit()`. El SELECT filtra por `status='pending' AND scheduled_at <= now` y ordena por `priority DESC, scheduled_at ASC`. El UPDATE atomico cambia el status a 'running' y setea `started_at`. Safe para multiples workers concurrentes — SQLite serializa la transaccion.
diff --git a/functions/infra/job_enqueue.go b/functions/infra/job_enqueue.go
new file mode 100644
index 00000000..aa11199a
--- /dev/null
+++ b/functions/infra/job_enqueue.go
@@ -0,0 +1,47 @@
+package infra
+
+import (
+	"fmt"
+	"time"
+
+	"github.com/google/uuid"
+)
+
+// JobEnqueue inserts a new job into the queue and returns its UUID.
+// jobType identifies the kind of work (e.g. "send_email", "resize_image").
+// payload is a JSON string with the job data (defaults to "{}" if empty).
+// Options: WithPriority, WithScheduledAt, WithMaxAttempts.
+func JobEnqueue(q *JobQueue, jobType string, payload string, opts ...EnqueueOption) (string, error) {
+	if q == nil {
+		return "", fmt.Errorf("job_enqueue: queue must not be nil")
+	}
+	if jobType == "" {
+		return "", fmt.Errorf("job_enqueue: jobType must not be empty")
+	}
+	if payload == "" {
+		payload = "{}"
+	}
+
+	cfg := &enqueueConfig{
+		priority:    0,
+		scheduledAt: time.Now().UTC(),
+		maxAttempts: 3,
+	}
+	for _, o := range opts {
+		o(cfg)
+	}
+
+	id := uuid.New().String()
+	scheduledAt := cfg.scheduledAt.Format(time.RFC3339)
+
+	q2 := fmt.Sprintf(`
+INSERT INTO %s (id, type, payload, status, priority, max_attempts, scheduled_at)
+VALUES (?, ?, ?, 'pending', ?, ?, ?)
+`, q.TableName)
+
+	_, err := q.DB.Exec(q2, id, jobType, payload, cfg.priority, cfg.maxAttempts, scheduledAt)
+	if err != nil {
+		return "", fmt.Errorf("job_enqueue: insert: %w", err)
+	}
+	return id, nil
+}
diff --git a/functions/infra/job_enqueue.md b/functions/infra/job_enqueue.md
new file mode 100644
index 00000000..78ce1f86
--- /dev/null
+++ b/functions/infra/job_enqueue.md
@@ -0,0 +1,48 @@
+---
+name: job_enqueue
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobEnqueue(q *JobQueue, jobType string, payload string, opts ...EnqueueOption) (string, error)"
+description: "Inserta un nuevo job en la cola con UUID generado, tipo, payload JSON y opciones. Retorna el UUID del job. Opciones: WithPriority, WithScheduledAt, WithMaxAttempts."
+tags: [job, queue, enqueue, insert, async, background, infra, sqlite]
+uses_functions: []
+uses_types: [job_queue_go_infra]
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [fmt, time, github.com/google/uuid]
+params:
+  - name: q
+    desc: "cola de jobs creada con JobQueueCreate"
+  - name: jobType
+    desc: "identificador del tipo de trabajo (ej: 'send_email', 'resize_image')"
+  - name: payload
+    desc: "JSON string con los datos del job; si vacio se usa '{}'"
+  - name: opts
+    desc: "opciones: WithPriority(n), WithScheduledAt(t), WithMaxAttempts(n)"
+output: "UUID del job insertado como string"
+tested: true
+tests:
+  - "enqueue_dequeue_atomicidad"
+  - "dequeue_priority_order"
+  - "dequeue_jobtype_filter"
+  - "dequeue_scheduled_in_future_waits"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_enqueue.go"
+---
+
+## Ejemplo
+
+```go
+id, err := JobEnqueue(q, "send_email", `{"to":"user@example.com"}`,
+    WithPriority(5),
+    WithMaxAttempts(5),
+)
+```
+
+## Notas
+
+Usa `github.com/google/uuid` para generar el ID. El payload por defecto es "{}". El campo `scheduled_at` por defecto es `time.Now().UTC()`. La prioridad mas alta se procesa primero (ORDER BY priority DESC).
diff --git a/functions/infra/job_fail.go b/functions/infra/job_fail.go
new file mode 100644
index 00000000..99efebb8
--- /dev/null
+++ b/functions/infra/job_fail.go
@@ -0,0 +1,43 @@
+package infra
+
+import (
+	"fmt"
+	"time"
+)
+
+// JobFail increments the attempt counter and transitions the job to:
+//   - "failed"  when attempts < max_attempts (eligible for retry)
+//   - "dead"    when attempts >= max_attempts (no more retries)
+//
+// errMsg is stored in the error column for debugging.
+func JobFail(q *JobQueue, jobID string, errMsg string) error {
+	if q == nil {
+		return fmt.Errorf("job_fail: queue must not be nil")
+	}
+	if jobID == "" {
+		return fmt.Errorf("job_fail: jobID must not be empty")
+	}
+
+	now := time.Now().UTC().Format(time.RFC3339)
+
+	// Atomically increment attempts and decide status.
+	query := fmt.Sprintf(`
+UPDATE %s
+SET
+    attempts    = attempts + 1,
+    status      = CASE WHEN (attempts + 1) >= max_attempts THEN 'dead' ELSE 'failed' END,
+    error       = ?,
+    completed_at = ?
+WHERE id = ?
+`, q.TableName)
+
+	res, err := q.DB.Exec(query, errMsg, now, jobID)
+	if err != nil {
+		return fmt.Errorf("job_fail: update: %w", err)
+	}
+	n, _ := res.RowsAffected()
+	if n == 0 {
+		return fmt.Errorf("job_fail: job %q not found", jobID)
+	}
+	return nil
+}
diff --git a/functions/infra/job_fail.md b/functions/infra/job_fail.md
new file mode 100644
index 00000000..094b995a
--- /dev/null
+++ b/functions/infra/job_fail.md
@@ -0,0 +1,43 @@
+---
+name: job_fail
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobFail(q *JobQueue, jobID string, errMsg string) error"
+description: "Incrementa el contador de intentos y transiciona el job a 'failed' (reintentable) o 'dead' (sin mas intentos) segun si attempts >= max_attempts. Almacena el mensaje de error."
+tags: [job, queue, fail, retry, dead, sqlite, async, background, infra]
+uses_functions: []
+uses_types: [job_queue_go_infra]
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [fmt, time]
+params:
+  - name: q
+    desc: "cola de jobs creada con JobQueueCreate"
+  - name: jobID
+    desc: "UUID del job que fallo"
+  - name: errMsg
+    desc: "mensaje de error para almacenar en la columna error"
+output: "error si el job no existe o falla el UPDATE"
+tested: true
+tests:
+  - "fail_increments_attempts"
+  - "fail_transitions_to_dead"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_fail.go"
+---
+
+## Ejemplo
+
+```go
+if err := handler(job); err != nil {
+    _ = JobFail(q, job.ID, err.Error())
+}
+```
+
+## Notas
+
+El UPDATE es atomico: incrementa attempts y decide el nuevo status en un solo SQL con CASE WHEN. Si `(attempts + 1) >= max_attempts` → status='dead', sino → status='failed'. Los jobs failed pueden ser reintentados manualmente reseteando su status a 'pending'. Los jobs dead no se procesan automaticamente.
diff --git a/functions/infra/job_queue_create.go b/functions/infra/job_queue_create.go
new file mode 100644
index 00000000..9a99dd8c
--- /dev/null
+++ b/functions/infra/job_queue_create.go
@@ -0,0 +1,72 @@
+package infra
+
+import (
+	"database/sql"
+	"fmt"
+)
+
+// JobQueueCreate creates (or verifies) the jobs table and required indices in
+// the given SQLite database, activates WAL mode, and returns a ready *JobQueue.
+// tableName is typically "jobs" but can be any valid SQLite identifier.
+//
+// Schema created:
+//
+//	CREATE TABLE IF NOT EXISTS jobs (
+//	    id TEXT PRIMARY KEY,
+//	    type TEXT NOT NULL,
+//	    payload TEXT NOT NULL DEFAULT '{}',
+//	    status TEXT NOT NULL DEFAULT 'pending',
+//	    priority INTEGER NOT NULL DEFAULT 0,
+//	    attempts INTEGER NOT NULL DEFAULT 0,
+//	    max_attempts INTEGER NOT NULL DEFAULT 3,
+//	    scheduled_at TEXT NOT NULL,
+//	    started_at TEXT,
+//	    completed_at TEXT,
+//	    result TEXT,
+//	    error TEXT,
+//	    created_at TEXT NOT NULL DEFAULT (strftime('%Y-%m-%dT%H:%M:%SZ', 'now'))
+//	);
+func JobQueueCreate(db *sql.DB, tableName string) (*JobQueue, error) {
+	if db == nil {
+		return nil, fmt.Errorf("job_queue_create: db must not be nil")
+	}
+	if tableName == "" {
+		tableName = "jobs"
+	}
+
+	// Enable WAL mode for concurrent read/write access.
+	if _, err := db.Exec(`PRAGMA journal_mode=WAL`); err != nil {
+		return nil, fmt.Errorf("job_queue_create: enable WAL: %w", err)
+	}
+
+	schema := fmt.Sprintf(`
+CREATE TABLE IF NOT EXISTS %s (
+    id TEXT PRIMARY KEY,
+    type TEXT NOT NULL,
+    payload TEXT NOT NULL DEFAULT '{}',
+    status TEXT NOT NULL DEFAULT 'pending',
+    priority INTEGER NOT NULL DEFAULT 0,
+    attempts INTEGER NOT NULL DEFAULT 0,
+    max_attempts INTEGER NOT NULL DEFAULT 3,
+    scheduled_at TEXT NOT NULL,
+    started_at TEXT,
+    completed_at TEXT,
+    result TEXT,
+    error TEXT,
+    created_at TEXT NOT NULL DEFAULT (strftime('%%Y-%%m-%%dT%%H:%%M:%%SZ', 'now'))
+);
+CREATE INDEX IF NOT EXISTS idx_%s_dequeue ON %s (status, priority DESC, scheduled_at ASC) WHERE status = 'pending';
+CREATE INDEX IF NOT EXISTS idx_%s_status ON %s (status);
+CREATE INDEX IF NOT EXISTS idx_%s_type ON %s (type);
+`, tableName,
+		tableName, tableName,
+		tableName, tableName,
+		tableName, tableName,
+	)
+
+	if _, err := db.Exec(schema); err != nil {
+		return nil, fmt.Errorf("job_queue_create: create schema: %w", err)
+	}
+
+	return &JobQueue{DB: db, TableName: tableName}, nil
+}
diff --git a/functions/infra/job_queue_create.md b/functions/infra/job_queue_create.md
new file mode 100644
index 00000000..965a6f50
--- /dev/null
+++ b/functions/infra/job_queue_create.md
@@ -0,0 +1,45 @@
+---
+name: job_queue_create
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobQueueCreate(db *sql.DB, tableName string) (*JobQueue, error)"
+description: "Crea (o verifica) la tabla de jobs con indices en SQLite, activa WAL mode, y retorna un *JobQueue listo para usar. tableName puede ser cualquier identificador SQLite valido (default: 'jobs')."
+tags: [job, queue, sqlite, async, background, infra, create, schema]
+uses_functions: []
+uses_types: [job_queue_go_infra]
+returns: [job_queue_go_infra]
+returns_optional: false
+error_type: "error_go_core"
+imports: [database/sql, fmt]
+params:
+  - name: db
+    desc: "conexion SQLite abierta (puede ser :memory: o file-based)"
+  - name: tableName
+    desc: "nombre de la tabla de jobs; si esta vacio se usa 'jobs'"
+output: "puntero a JobQueue con DB y TableName configurados, listo para enqueue/dequeue"
+tested: true
+tests:
+  - "test_queue_create"
+  - "test_queue_create_default_table_name"
+  - "test_queue_create_nil_db"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_queue_create.go"
+---
+
+## Ejemplo
+
+```go
+db, _ := sql.Open("sqlite3", "jobs.db?_journal_mode=WAL")
+q, err := JobQueueCreate(db, "jobs")
+if err != nil {
+    log.Fatal(err)
+}
+// q esta listo para JobEnqueue, JobDequeue, etc.
+```
+
+## Notas
+
+Activa WAL mode en la DB. Usa `CREATE TABLE IF NOT EXISTS` y `CREATE INDEX IF NOT EXISTS` — es seguro llamarlo multiples veces. Indices creados: `idx_{table}_dequeue` (filtrado partial por status='pending'), `idx_{table}_status`, `idx_{table}_type`.
diff --git a/functions/infra/job_queue_test.go b/functions/infra/job_queue_test.go
new file mode 100644
index 00000000..f478bbc0
--- /dev/null
+++ b/functions/infra/job_queue_test.go
@@ -0,0 +1,418 @@
+package infra
+
+import (
+	"context"
+	"database/sql"
+	"fmt"
+	"sync"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	_ "github.com/mattn/go-sqlite3"
+)
+
+// openTestDB opens an in-memory SQLite database for testing.
+// MaxOpenConns is set to 1 to ensure all goroutines share the same connection
+// (required for in-memory SQLite which is per-connection otherwise).
+func openTestDB(t *testing.T) *sql.DB {
+	t.Helper()
+	db, err := sql.Open("sqlite3", ":memory:?_journal_mode=WAL")
+	if err != nil {
+		t.Fatalf("open test db: %v", err)
+	}
+	db.SetMaxOpenConns(1)
+	t.Cleanup(func() { db.Close() })
+	return db
+}
+
+func TestJobQueueCreate(t *testing.T) {
+	t.Run("test_queue_create", func(t *testing.T) {
+		db := openTestDB(t)
+		q, err := JobQueueCreate(db, "jobs")
+		if err != nil {
+			t.Fatalf("JobQueueCreate: %v", err)
+		}
+		if q == nil {
+			t.Fatal("expected non-nil JobQueue")
+		}
+		if q.TableName != "jobs" {
+			t.Errorf("TableName = %q, want %q", q.TableName, "jobs")
+		}
+
+		// Verify table exists by inserting a row directly.
+		_, err = db.Exec(`INSERT INTO jobs (id, type, scheduled_at) VALUES ('x', 'test', ?)`,
+			time.Now().UTC().Format(time.RFC3339))
+		if err != nil {
+			t.Fatalf("table not created: %v", err)
+		}
+	})
+
+	t.Run("test_queue_create_default_table_name", func(t *testing.T) {
+		db := openTestDB(t)
+		q, err := JobQueueCreate(db, "")
+		if err != nil {
+			t.Fatalf("JobQueueCreate with empty name: %v", err)
+		}
+		if q.TableName != "jobs" {
+			t.Errorf("TableName = %q, want %q", q.TableName, "jobs")
+		}
+	})
+
+	t.Run("test_queue_create_nil_db", func(t *testing.T) {
+		_, err := JobQueueCreate(nil, "jobs")
+		if err == nil {
+			t.Fatal("expected error for nil db")
+		}
+	})
+}
+
+func TestJobEnqueueDequeue(t *testing.T) {
+	t.Run("enqueue_dequeue_atomicidad", func(t *testing.T) {
+		db := openTestDB(t)
+		q, err := JobQueueCreate(db, "jobs")
+		if err != nil {
+			t.Fatalf("create: %v", err)
+		}
+
+		id, err := JobEnqueue(q, "send_email", `{"to":"a@b.com"}`)
+		if err != nil {
+			t.Fatalf("enqueue: %v", err)
+		}
+		if id == "" {
+			t.Fatal("expected non-empty job ID")
+		}
+
+		job, err := JobDequeue(q, nil)
+		if err != nil {
+			t.Fatalf("dequeue: %v", err)
+		}
+		if job == nil {
+			t.Fatal("expected a job, got nil")
+		}
+		if job.ID != id {
+			t.Errorf("job ID = %q, want %q", job.ID, id)
+		}
+		if job.Type != "send_email" {
+			t.Errorf("job Type = %q, want %q", job.Type, "send_email")
+		}
+		if job.Status != JobStatusRunning {
+			t.Errorf("job Status = %q, want %q", job.Status, JobStatusRunning)
+		}
+
+		// Second dequeue should return nil (job is now running, not pending).
+		job2, err := JobDequeue(q, nil)
+		if err != nil {
+			t.Fatalf("second dequeue: %v", err)
+		}
+		if job2 != nil {
+			t.Errorf("expected nil second dequeue, got job %q", job2.ID)
+		}
+	})
+
+	t.Run("dequeue_empty_queue_returns_nil", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		job, err := JobDequeue(q, nil)
+		if err != nil {
+			t.Fatalf("dequeue empty: %v", err)
+		}
+		if job != nil {
+			t.Errorf("expected nil, got job %q", job.ID)
+		}
+	})
+
+	t.Run("dequeue_priority_order", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		_, _ = JobEnqueue(q, "low", "{}", WithPriority(0))
+		_, _ = JobEnqueue(q, "high", "{}", WithPriority(10))
+		_, _ = JobEnqueue(q, "mid", "{}", WithPriority(5))
+
+		job, _ := JobDequeue(q, nil)
+		if job == nil || job.Type != "high" {
+			t.Errorf("expected high-priority job first, got %v", job)
+		}
+	})
+
+	t.Run("dequeue_jobtype_filter", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		_, _ = JobEnqueue(q, "email", "{}")
+		_, _ = JobEnqueue(q, "sms", "{}")
+
+		// Only dequeue "sms" jobs.
+		job, err := JobDequeue(q, []string{"sms"})
+		if err != nil {
+			t.Fatalf("dequeue with filter: %v", err)
+		}
+		if job == nil || job.Type != "sms" {
+			t.Errorf("expected sms job, got %v", job)
+		}
+	})
+
+	t.Run("dequeue_scheduled_in_future_waits", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		future := time.Now().Add(1 * time.Hour)
+		_, _ = JobEnqueue(q, "future", "{}", WithScheduledAt(future))
+
+		job, err := JobDequeue(q, nil)
+		if err != nil {
+			t.Fatalf("dequeue future: %v", err)
+		}
+		if job != nil {
+			t.Errorf("expected nil (future job not yet due), got job %q", job.ID)
+		}
+	})
+}
+
+func TestJobCompleteAndFail(t *testing.T) {
+	t.Run("complete_fail_transitions", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		id, _ := JobEnqueue(q, "work", "{}", WithMaxAttempts(3))
+		job, _ := JobDequeue(q, nil)
+		if job == nil {
+			t.Fatal("expected job")
+		}
+
+		// Complete it.
+		if err := JobComplete(q, id, `{"ok":true}`); err != nil {
+			t.Fatalf("complete: %v", err)
+		}
+
+		// Verify status in DB.
+		var status string
+		_ = db.QueryRow(`SELECT status FROM jobs WHERE id = ?`, id).Scan(&status)
+		if status != "completed" {
+			t.Errorf("status = %q, want completed", status)
+		}
+	})
+
+	t.Run("fail_increments_attempts", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		id, _ := JobEnqueue(q, "work", "{}", WithMaxAttempts(3))
+		_, _ = JobDequeue(q, nil)
+
+		// First failure → status = failed (attempts=1 < max=3).
+		if err := JobFail(q, id, "oops"); err != nil {
+			t.Fatalf("fail: %v", err)
+		}
+		var status string
+		var attempts int
+		_ = db.QueryRow(`SELECT status, attempts FROM jobs WHERE id = ?`, id).Scan(&status, &attempts)
+		if status != "failed" {
+			t.Errorf("status = %q, want failed", status)
+		}
+		if attempts != 1 {
+			t.Errorf("attempts = %d, want 1", attempts)
+		}
+	})
+
+	t.Run("fail_transitions_to_dead", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		id, _ := JobEnqueue(q, "work", "{}", WithMaxAttempts(2))
+
+		// Fail twice → dead.
+		for i := 0; i < 2; i++ {
+			// Re-set to pending so we can dequeue again (simulate retry logic).
+			if i > 0 {
+				_, _ = db.Exec(`UPDATE jobs SET status = 'pending' WHERE id = ?`, id)
+			}
+			_, _ = JobDequeue(q, nil)
+			_ = JobFail(q, id, fmt.Sprintf("error %d", i+1))
+		}
+
+		var status string
+		_ = db.QueryRow(`SELECT status FROM jobs WHERE id = ?`, id).Scan(&status)
+		if status != "dead" {
+			t.Errorf("status = %q, want dead after max_attempts reached", status)
+		}
+	})
+}
+
+func TestJobStatusSummaryFn(t *testing.T) {
+	t.Run("job_status_summary_format", func(t *testing.T) {
+		counts := map[string]int{
+			"pending":   5,
+			"running":   2,
+			"completed": 10,
+			"failed":    1,
+			"dead":      0,
+		}
+		got := JobStatusSummary(counts)
+		want := "pending: 5, running: 2, completed: 10, failed: 1, dead: 0"
+		if got != want {
+			t.Errorf("got %q, want %q", got, want)
+		}
+	})
+
+	t.Run("job_status_summary_empty_map", func(t *testing.T) {
+		got := JobStatusSummary(map[string]int{})
+		want := "pending: 0, running: 0, completed: 0, failed: 0, dead: 0"
+		if got != want {
+			t.Errorf("got %q, want %q", got, want)
+		}
+	})
+}
+
+func TestJobWorkerGracefulShutdown(t *testing.T) {
+	t.Run("worker_graceful_shutdown", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		ctx, cancel := context.WithCancel(context.Background())
+
+		done := make(chan struct{})
+		go func() {
+			defer close(done)
+			_ = JobWorker(ctx, q, func(j Job) error {
+				return nil
+			}, WithPollInterval(10*time.Millisecond))
+		}()
+
+		// Let the worker poll a couple of times.
+		time.Sleep(50 * time.Millisecond)
+		cancel()
+
+		select {
+		case <-done:
+			// OK — worker exited cleanly.
+		case <-time.After(2 * time.Second):
+			t.Fatal("worker did not stop within timeout after context cancellation")
+		}
+	})
+
+	t.Run("worker_processes_jobs", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		var processed atomic.Int32
+		for i := 0; i < 3; i++ {
+			_, _ = JobEnqueue(q, "task", `{}`)
+		}
+
+		ctx, cancel := context.WithCancel(context.Background())
+		go func() {
+			_ = JobWorker(ctx, q, func(j Job) error {
+				processed.Add(1)
+				return nil
+			}, WithPollInterval(10*time.Millisecond))
+		}()
+
+		// Wait for all 3 to be processed (with timeout).
+		deadline := time.Now().Add(2 * time.Second)
+		for time.Now().Before(deadline) {
+			if processed.Load() >= 3 {
+				break
+			}
+			time.Sleep(10 * time.Millisecond)
+		}
+		cancel()
+
+		if n := processed.Load(); n < 3 {
+			t.Errorf("processed %d jobs, want 3", n)
+		}
+	})
+}
+
+func TestJobConcurrency(t *testing.T) {
+	t.Run("concurrency_multiples_goroutines_dequeue", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		const numJobs = 50
+		for i := 0; i < numJobs; i++ {
+			_, _ = JobEnqueue(q, "task", fmt.Sprintf(`{"i":%d}`, i))
+		}
+
+		var processed atomic.Int32
+		var wg sync.WaitGroup
+
+		ctx, cancel := context.WithCancel(context.Background())
+		const numWorkers = 5
+		for i := 0; i < numWorkers; i++ {
+			wg.Add(1)
+			go func() {
+				defer wg.Done()
+				_ = JobWorker(ctx, q, func(j Job) error {
+					processed.Add(1)
+					return nil
+				}, WithPollInterval(5*time.Millisecond))
+			}()
+		}
+
+		// Wait until all jobs are processed.
+		deadline := time.Now().Add(5 * time.Second)
+		for time.Now().Before(deadline) {
+			if processed.Load() >= numJobs {
+				break
+			}
+			time.Sleep(10 * time.Millisecond)
+		}
+		cancel()
+		wg.Wait()
+
+		if n := processed.Load(); n != numJobs {
+			t.Errorf("processed %d, want %d (no double-processing with concurrent workers)", n, numJobs)
+		}
+	})
+}
+
+func TestJobCleanupFn(t *testing.T) {
+	t.Run("job_cleanup_removes_old_terminal_jobs", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		id, _ := JobEnqueue(q, "work", "{}")
+		_, _ = JobDequeue(q, nil)
+		_ = JobComplete(q, id, "")
+
+		// Make the job look old (set created_at in the past, always UTC to match JobCleanup).
+		past := time.Now().UTC().Add(-25 * time.Hour).Format(time.RFC3339)
+		_, _ = db.Exec(`UPDATE jobs SET created_at = ? WHERE id = ?`, past, id)
+
+		n, err := JobCleanup(q, 24*time.Hour)
+		if err != nil {
+			t.Fatalf("cleanup: %v", err)
+		}
+		if n != 1 {
+			t.Errorf("deleted %d rows, want 1", n)
+		}
+
+		// Job should be gone.
+		var count int
+		_ = db.QueryRow(`SELECT COUNT(*) FROM jobs WHERE id = ?`, id).Scan(&count)
+		if count != 0 {
+			t.Errorf("job still in DB after cleanup")
+		}
+	})
+
+	t.Run("job_cleanup_keeps_recent_jobs", func(t *testing.T) {
+		db := openTestDB(t)
+		q, _ := JobQueueCreate(db, "jobs")
+
+		id, _ := JobEnqueue(q, "work", "{}")
+		_, _ = JobDequeue(q, nil)
+		_ = JobComplete(q, id, "")
+
+		// Job was just created — should not be cleaned up.
+		n, err := JobCleanup(q, 24*time.Hour)
+		if err != nil {
+			t.Fatalf("cleanup: %v", err)
+		}
+		if n != 0 {
+			t.Errorf("deleted %d rows, want 0 (job is recent)", n)
+		}
+	})
+}
diff --git a/functions/infra/job_queue_types.go b/functions/infra/job_queue_types.go
new file mode 100644
index 00000000..458193af
--- /dev/null
+++ b/functions/infra/job_queue_types.go
@@ -0,0 +1,87 @@
+package infra
+
+import (
+	"database/sql"
+	"time"
+)
+
+// JobStatus represents the lifecycle state of a job.
+type JobStatus string
+
+const (
+	JobStatusPending   JobStatus = "pending"
+	JobStatusRunning   JobStatus = "running"
+	JobStatusCompleted JobStatus = "completed"
+	JobStatusFailed    JobStatus = "failed"
+	JobStatusDead      JobStatus = "dead"
+)
+
+// Job represents a unit of asynchronous work stored in SQLite.
+type Job struct {
+	ID          string
+	Type        string
+	Payload     string    // JSON string
+	Status      JobStatus
+	Priority    int
+	Attempts    int
+	MaxAttempts int
+	ScheduledAt time.Time
+	StartedAt   *time.Time
+	CompletedAt *time.Time
+	Result      *string
+	Error       *string
+	CreatedAt   time.Time
+}
+
+// JobQueue wraps a *sql.DB and the jobs table name.
+type JobQueue struct {
+	DB        *sql.DB
+	TableName string
+}
+
+// JobHandler is the function signature for processing a job.
+// Return a non-nil error to trigger retry / fail logic.
+type JobHandler func(job Job) error
+
+// EnqueueOption configures optional behaviour of JobEnqueue.
+type EnqueueOption func(*enqueueConfig)
+
+type enqueueConfig struct {
+	priority    int
+	scheduledAt time.Time
+	maxAttempts int
+}
+
+// WithPriority sets the job priority (higher = processed first).
+func WithPriority(p int) EnqueueOption {
+	return func(c *enqueueConfig) { c.priority = p }
+}
+
+// WithScheduledAt delays execution until the given time.
+func WithScheduledAt(t time.Time) EnqueueOption {
+	return func(c *enqueueConfig) { c.scheduledAt = t }
+}
+
+// WithMaxAttempts sets how many times the job may be attempted before dying.
+func WithMaxAttempts(n int) EnqueueOption {
+	return func(c *enqueueConfig) { c.maxAttempts = n }
+}
+
+// WorkerOption configures optional behaviour of JobWorker / JobWorkerPool.
+type WorkerOption func(*workerConfig)
+
+type workerConfig struct {
+	pollInterval time.Duration
+	jobTypes     []string
+}
+
+// WithPollInterval sets how long the worker sleeps between dequeue attempts.
+func WithPollInterval(d time.Duration) WorkerOption {
+	return func(c *workerConfig) { c.pollInterval = d }
+}
+
+// WithJobTypes restricts the worker to the given job types.
+// An empty list means all types.
+func WithJobTypes(types ...string) WorkerOption {
+	return func(c *workerConfig) { c.jobTypes = types }
+}
diff --git a/functions/infra/job_status.go b/functions/infra/job_status.go
new file mode 100644
index 00000000..a0cfb653
--- /dev/null
+++ b/functions/infra/job_status.go
@@ -0,0 +1,19 @@
+package infra
+
+import (
+	"fmt"
+	"strings"
+)
+
+// JobStatus formats a map of status → count into a human-readable summary string.
+// Example output: "pending: 5, running: 2, completed: 10, failed: 1, dead: 0"
+// The output always includes all five canonical statuses in a fixed order.
+// This is a pure function — no I/O, no state.
+func JobStatusSummary(counts map[string]int) string {
+	statuses := []string{"pending", "running", "completed", "failed", "dead"}
+	parts := make([]string, 0, len(statuses))
+	for _, s := range statuses {
+		parts = append(parts, fmt.Sprintf("%s: %d", s, counts[s]))
+	}
+	return strings.Join(parts, ", ")
+}
diff --git a/functions/infra/job_status.md b/functions/infra/job_status.md
new file mode 100644
index 00000000..3eec0965
--- /dev/null
+++ b/functions/infra/job_status.md
@@ -0,0 +1,39 @@
+---
+name: job_status_summary
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: pure
+signature: "func JobStatusSummary(counts map[string]int) string"
+description: "Formatea un mapa de status→conteo en un resumen legible. Siempre incluye los cinco estados canonicos en orden fijo: pending, running, completed, failed, dead."
+tags: [job, queue, status, summary, format, pure, infra]
+uses_functions: []
+uses_types: []
+returns: []
+returns_optional: false
+error_type: ""
+imports: [fmt, strings]
+params:
+  - name: counts
+    desc: "mapa de nombre de status (pending/running/completed/failed/dead) a numero de jobs en ese estado"
+output: "string formateado 'pending: N, running: N, completed: N, failed: N, dead: N'"
+tested: true
+tests:
+  - "job_status_summary_format"
+  - "job_status_summary_empty_map"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_status.go"
+---
+
+## Ejemplo
+
+```go
+counts := map[string]int{"pending": 5, "running": 2, "completed": 10}
+summary := JobStatusSummary(counts)
+// "pending: 5, running: 2, completed: 10, failed: 0, dead: 0"
+```
+
+## Notas
+
+Funcion pura. Los estados no presentes en el mapa se tratan como 0. El orden de salida es siempre fijo: pending, running, completed, failed, dead.
diff --git a/functions/infra/job_worker.go b/functions/infra/job_worker.go
new file mode 100644
index 00000000..405fe12a
--- /dev/null
+++ b/functions/infra/job_worker.go
@@ -0,0 +1,70 @@
+package infra
+
+import (
+	"context"
+	"fmt"
+	"time"
+)
+
+// JobWorker runs a poll loop in the calling goroutine, dequeuing and processing
+// jobs until ctx is cancelled. handler is called for each job; its return value
+// drives complete vs fail logic. The worker sleeps pollInterval between empty
+// dequeue attempts.
+//
+// Options: WithPollInterval (default 1s), WithJobTypes (default all types).
+//
+// This function blocks until ctx is done. Run it in a goroutine:
+//
+//	go JobWorker(ctx, q, handler, WithPollInterval(2*time.Second))
+func JobWorker(ctx context.Context, q *JobQueue, handler JobHandler, opts ...WorkerOption) error {
+	if q == nil {
+		return fmt.Errorf("job_worker: queue must not be nil")
+	}
+	if handler == nil {
+		return fmt.Errorf("job_worker: handler must not be nil")
+	}
+
+	cfg := &workerConfig{
+		pollInterval: time.Second,
+	}
+	for _, o := range opts {
+		o(cfg)
+	}
+
+	for {
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		default:
+		}
+
+		job, err := JobDequeue(q, cfg.jobTypes)
+		if err != nil {
+			// Transient DB error — wait and retry.
+			select {
+			case <-ctx.Done():
+				return ctx.Err()
+			case <-time.After(cfg.pollInterval):
+			}
+			continue
+		}
+
+		if job == nil {
+			// Queue empty — wait before polling again.
+			select {
+			case <-ctx.Done():
+				return ctx.Err()
+			case <-time.After(cfg.pollInterval):
+			}
+			continue
+		}
+
+		// Process the job.
+		handlerErr := handler(*job)
+		if handlerErr != nil {
+			_ = JobFail(q, job.ID, handlerErr.Error())
+		} else {
+			_ = JobComplete(q, job.ID, "")
+		}
+	}
+}
diff --git a/functions/infra/job_worker.md b/functions/infra/job_worker.md
new file mode 100644
index 00000000..947107f1
--- /dev/null
+++ b/functions/infra/job_worker.md
@@ -0,0 +1,49 @@
+---
+name: job_worker
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobWorker(ctx context.Context, q *JobQueue, handler JobHandler, opts ...WorkerOption) error"
+description: "Ejecuta un poll loop bloqueante que desencola y procesa jobs hasta que el context sea cancelado. El handler determina complete vs fail. Opciones: WithPollInterval (default 1s), WithJobTypes."
+tags: [job, queue, worker, goroutine, poll, async, background, infra, concurrency]
+uses_functions: [job_dequeue_go_infra, job_complete_go_infra, job_fail_go_infra]
+uses_types: [job_queue_go_infra, job_go_infra]
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [context, fmt, time]
+params:
+  - name: ctx
+    desc: "context que al cancelarse detiene el worker limpiamente (graceful shutdown)"
+  - name: q
+    desc: "cola de jobs creada con JobQueueCreate"
+  - name: handler
+    desc: "funcion que procesa un job; retorno nil = completado, error = fallo"
+  - name: opts
+    desc: "opciones: WithPollInterval(d), WithJobTypes(types...)"
+output: "ctx.Err() cuando el context se cancela, o error de configuracion"
+tested: true
+tests:
+  - "worker_graceful_shutdown"
+  - "worker_processes_jobs"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_worker.go"
+---
+
+## Ejemplo
+
+```go
+ctx, cancel := context.WithCancel(context.Background())
+defer cancel()
+
+go JobWorker(ctx, q, func(job Job) error {
+    fmt.Println("processing", job.ID, job.Type)
+    return nil // complete
+}, WithPollInterval(500*time.Millisecond), WithJobTypes("send_email"))
+```
+
+## Notas
+
+Funcion bloqueante — ejecutar en una goroutine. Retorna `ctx.Err()` cuando el context se cancela (graceful shutdown). Los errores transitorios de dequeue (DB) generan un sleep y retry. El handler tiene la responsabilidad completa del procesamiento; JobWorker se encarga de complete/fail.
diff --git a/functions/infra/job_worker_pool.go b/functions/infra/job_worker_pool.go
new file mode 100644
index 00000000..cf02cdea
--- /dev/null
+++ b/functions/infra/job_worker_pool.go
@@ -0,0 +1,36 @@
+package infra
+
+import (
+	"context"
+	"fmt"
+
+	"golang.org/x/sync/errgroup"
+)
+
+// JobWorkerPool starts n workers concurrently, all sharing the same queue and
+// handler. It uses errgroup for structured concurrency and supports graceful
+// shutdown via ctx cancellation.
+//
+// The function blocks until all workers exit. It returns the first non-context
+// error encountered, or ctx.Err() if all workers stopped due to cancellation.
+//
+// Options: WithPollInterval, WithJobTypes (applied to every worker).
+func JobWorkerPool(ctx context.Context, q *JobQueue, n int, handler JobHandler, opts ...WorkerOption) error {
+	if q == nil {
+		return fmt.Errorf("job_worker_pool: queue must not be nil")
+	}
+	if handler == nil {
+		return fmt.Errorf("job_worker_pool: handler must not be nil")
+	}
+	if n <= 0 {
+		return fmt.Errorf("job_worker_pool: n must be > 0, got %d", n)
+	}
+
+	g, gctx := errgroup.WithContext(ctx)
+	for i := 0; i < n; i++ {
+		g.Go(func() error {
+			return JobWorker(gctx, q, handler, opts...)
+		})
+	}
+	return g.Wait()
+}
diff --git a/functions/infra/job_worker_pool.md b/functions/infra/job_worker_pool.md
new file mode 100644
index 00000000..7ca60ef0
--- /dev/null
+++ b/functions/infra/job_worker_pool.md
@@ -0,0 +1,50 @@
+---
+name: job_worker_pool
+kind: function
+lang: go
+domain: infra
+version: "1.0.0"
+purity: impure
+signature: "func JobWorkerPool(ctx context.Context, q *JobQueue, n int, handler JobHandler, opts ...WorkerOption) error"
+description: "Lanza n workers concurrentes que comparten la misma cola y handler. Usa errgroup para concurrencia estructurada y soporta graceful shutdown via context. Bloquea hasta que todos los workers terminen."
+tags: [job, queue, worker, pool, concurrency, errgroup, async, background, infra, graceful-shutdown]
+uses_functions: [job_worker_go_infra]
+uses_types: [job_queue_go_infra]
+returns: []
+returns_optional: false
+error_type: "error_go_core"
+imports: [context, fmt, golang.org/x/sync/errgroup]
+params:
+  - name: ctx
+    desc: "context que al cancelarse detiene todos los workers; si uno falla, los demas tambien se detienen"
+  - name: q
+    desc: "cola de jobs compartida por todos los workers"
+  - name: n
+    desc: "numero de workers concurrentes (debe ser > 0)"
+  - name: handler
+    desc: "funcion de procesamiento compartida por todos los workers"
+  - name: opts
+    desc: "opciones aplicadas a cada worker: WithPollInterval, WithJobTypes"
+output: "primer error no-context de cualquier worker, o ctx.Err() si todos terminaron por cancelacion"
+tested: true
+tests:
+  - "concurrency_multiples_goroutines_dequeue"
+test_file_path: "functions/infra/job_queue_test.go"
+file_path: "functions/infra/job_worker_pool.go"
+---
+
+## Ejemplo
+
+```go
+ctx, cancel := context.WithCancel(context.Background())
+defer cancel()
+
+// 5 workers procesando emails en paralelo
+err := JobWorkerPool(ctx, q, 5, func(job Job) error {
+    return sendEmail(job)
+}, WithPollInterval(100*time.Millisecond))
+```
+
+## Notas
+
+Usa `golang.org/x/sync/errgroup`. Si un worker retorna un error no-context, `errgroup` cancela el context derivado y todos los demas workers se detienen. El graceful shutdown ocurre al cancelar el context original.
diff --git a/types/infra/job.md b/types/infra/job.md
new file mode 100644
index 00000000..c4f69597
--- /dev/null
+++ b/types/infra/job.md
@@ -0,0 +1,31 @@
+---
+name: job
+lang: go
+domain: infra
+version: "1.0.0"
+algebraic: product
+definition: |
+  type Job struct {
+      ID          string
+      Type        string
+      Payload     string
+      Status      JobStatus
+      Priority    int
+      Attempts    int
+      MaxAttempts int
+      ScheduledAt time.Time
+      StartedAt   *time.Time
+      CompletedAt *time.Time
+      Result      *string
+      Error       *string
+      CreatedAt   time.Time
+  }
+description: "Unidad de trabajo asincrono almacenada en SQLite. Payload es un JSON string. Status evoluciona de pending a running y luego a completed, failed o dead."
+tags: [job, queue, async, background, sqlite, infra]
+uses_types: [job_status_go_infra]
+file_path: "functions/infra/job_queue_types.go"
+---
+
+## Notas
+
+Tipo producto. Los campos `StartedAt`, `CompletedAt`, `Result` y `Error` son punteros — son nil hasta que el job alcanza el estado correspondiente. `Payload` es siempre un JSON string (minimo "{}").
diff --git a/types/infra/job_queue.md b/types/infra/job_queue.md
new file mode 100644
index 00000000..39f2c77a
--- /dev/null
+++ b/types/infra/job_queue.md
@@ -0,0 +1,20 @@
+---
+name: job_queue
+lang: go
+domain: infra
+version: "1.0.0"
+algebraic: product
+definition: |
+  type JobQueue struct {
+      DB        *sql.DB
+      TableName string
+  }
+description: "Handle para una cola de jobs SQLite. Wrappea un *sql.DB y el nombre de la tabla de jobs. Se crea con JobQueueCreate."
+tags: [job, queue, sqlite, infra, async]
+uses_types: []
+file_path: "functions/infra/job_queue_types.go"
+---
+
+## Notas
+
+Tipo producto minimo — todas las funciones del subsistema de jobs lo reciben como primer argumento. TableName permite tener multiples colas en la misma DB. Se crea con `JobQueueCreate`.
diff --git a/types/infra/job_status.md b/types/infra/job_status.md
new file mode 100644
index 00000000..27c5648b
--- /dev/null
+++ b/types/infra/job_status.md
@@ -0,0 +1,25 @@
+---
+name: job_status
+lang: go
+domain: infra
+version: "1.0.0"
+algebraic: sum
+definition: |
+  type JobStatus string
+
+  const (
+      JobStatusPending   JobStatus = "pending"
+      JobStatusRunning   JobStatus = "running"
+      JobStatusCompleted JobStatus = "completed"
+      JobStatusFailed    JobStatus = "failed"
+      JobStatusDead      JobStatus = "dead"
+  )
+description: "Estado del ciclo de vida de un job. Tipo suma con cinco variantes: pending, running, completed, failed, dead."
+tags: [job, status, queue, async, infra]
+uses_types: []
+file_path: "functions/infra/job_queue_types.go"
+---
+
+## Notas
+
+Tipo suma. Ciclo de vida normal: pending → running → completed. En error: pending → running → failed (reintentos posibles) → dead (sin mas intentos). Un job con `attempts >= max_attempts` pasa directamente a dead.

From 13e45aa22ef1444a101235c60fd9e95e7530f279 Mon Sep 17 00:00:00 2001
From: Egutierrez <egutierrez@dead.dd>
Date: Mon, 13 Apr 2026 02:01:03 +0200
Subject: [PATCH 2/2] docs: cerrar issue 0013

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 dev/issues/completed/0013-background-jobs.md | 409 +++++++++++++++++++
 1 file changed, 409 insertions(+)
 create mode 100644 dev/issues/completed/0013-background-jobs.md

diff --git a/dev/issues/completed/0013-background-jobs.md b/dev/issues/completed/0013-background-jobs.md
new file mode 100644
index 00000000..7c9f9d9f
--- /dev/null
+++ b/dev/issues/completed/0013-background-jobs.md
@@ -0,0 +1,409 @@
+# 0013 — Background Job Queue
+
+## Metadata
+
+| Campo | Valor |
+|-------|-------|
+| **ID** | 0013 |
+| **Estado** | pendiente |
+| **Prioridad** | alta |
+| **Tipo** | feature |
+
+## Dependencias
+
+Ninguna.
+
+---
+
+## Objetivo
+
+Cola de trabajos asincrona basada en SQLite para apps Go del registry. Permite a cualquier app encolar tareas (enviar emails, procesar archivos, ejecutar pipelines) y procesarlas en background con workers concurrentes, reintentos automaticos y dead-letter queue — todo respaldado por SQLite sin depender de Redis, RabbitMQ ni ningun servicio externo.
+
+## Contexto
+
+- Actualmente no existen funciones de job queue en el registry. Cada app que necesita procesamiento asincrono tiene que construir su propia solucion ad-hoc.
+- SQLite encaja perfectamente con el stack existente: `registry.db` y `operations.db` ya demuestran que SQLite escala para los volúmenes del proyecto.
+- Las apps Go del registry ya usan `mattn/go-sqlite3` con FTS5 y WAL mode — no hay dependencias nuevas.
+- `map_concurrent_go_core` ofrece paralelismo sincrono pero no persistencia: si el proceso muere, se pierde el trabajo. Una cola persistente en SQLite sobrevive a reinicios.
+- Patron existente en el ecosistema: `gocraft/work`, `faktory`, `river` — pero todos requieren Redis o Postgres. Una solucion SQLite-native es mas coherente con el proyecto.
+
+## Arquitectura
+
+```
+functions/infra/
+├── job_queue_create.go         — NEW: crea tabla de jobs en SQLite, retorna handle
+├── job_queue_create.md         — NEW
+├── job_enqueue.go              — NEW: añade job a la cola
+├── job_enqueue.md              — NEW
+├── job_dequeue.go              — NEW: reclama proximo job pendiente (atomico)
+├── job_dequeue.md              — NEW
+├── job_complete.go             — NEW: marca job completado con resultado
+├── job_complete.md             — NEW
+├── job_fail.go                 — NEW: marca job fallido, incrementa intentos
+├── job_fail.md                 — NEW
+├── job_worker.go               — NEW: goroutine que pollea y procesa jobs
+├── job_worker.md               — NEW
+├── job_worker_pool.go          — NEW: pool de N workers con graceful shutdown
+├── job_worker_pool.md          — NEW
+├── job_cleanup.go              — NEW: elimina jobs antiguos completados/fallidos
+├── job_cleanup.md              — NEW
+├── job_status.go               — NEW: resumen de estado de la cola (pura)
+├── job_status.md               — NEW
+
+types/infra/
+├── job.md                      — NEW: metadata del tipo Job
+├── job_queue.md                — NEW: metadata del tipo JobQueue
+├── job_status.md               — NEW: metadata del tipo JobStatus (sum)
+├── job_handler.md              — NEW: metadata del tipo JobHandler
+```
+
+### Patron pure core / impure shell
+
+- **Pure:** `job_status` — formatea conteos en un resumen sin I/O.
+- **Impure:** Todo lo demas — interactúa con SQLite (I/O de disco) y goroutines (concurrencia).
+
+## Diseno
+
+### Tipos
+
+```go
+// JobStatus es un sum type: estados posibles de un job
+type JobStatus string
+
+const (
+    JobStatusPending   JobStatus = "pending"
+    JobStatusRunning   JobStatus = "running"
+    JobStatusCompleted JobStatus = "completed"
+    JobStatusFailed    JobStatus = "failed"
+    JobStatusDead      JobStatus = "dead" // max_attempts superado
+)
+
+// Job representa una unidad de trabajo en la cola
+type Job struct {
+    ID           string    `json:"id"`            // UUID generado al encolar
+    Type         string    `json:"type"`          // tipo del job (ej: "send_email", "process_file")
+    Payload      string    `json:"payload"`       // JSON arbitrario con los datos del job
+    Status       JobStatus `json:"status"`        // pending | running | completed | failed | dead
+    Priority     int       `json:"priority"`      // 0 = normal, mayor = mas prioritario
+    Attempts     int       `json:"attempts"`      // intentos ejecutados hasta ahora
+    MaxAttempts  int       `json:"max_attempts"`  // maximo de intentos antes de dead-letter
+    ScheduledAt  time.Time `json:"scheduled_at"`  // cuando debe ejecutarse (permite scheduling futuro)
+    StartedAt    *time.Time `json:"started_at"`   // cuando empezo a ejecutarse (nil si pending)
+    CompletedAt  *time.Time `json:"completed_at"` // cuando termino (nil si no completado)
+    Result       string    `json:"result"`        // JSON con resultado de ejecucion exitosa
+    Error        string    `json:"error"`         // mensaje de error del ultimo intento
+    CreatedAt    time.Time `json:"created_at"`    // timestamp de creacion
+}
+
+// JobQueue es el handle a una cola respaldada por SQLite
+type JobQueue struct {
+    DB        *sql.DB // conexion SQLite con WAL mode
+    TableName string  // nombre de la tabla (permite multiples colas en una BD)
+}
+
+// JobHandler es la funcion que procesa un job
+// Recibe el job completo y retorna error si falla.
+type JobHandler func(job Job) error
+```
+
+### Schema SQLite
+
+```sql
+CREATE TABLE IF NOT EXISTS jobs (
+    id           TEXT PRIMARY KEY,
+    type         TEXT NOT NULL,
+    payload      TEXT NOT NULL DEFAULT '{}',
+    status       TEXT NOT NULL DEFAULT 'pending',
+    priority     INTEGER NOT NULL DEFAULT 0,
+    attempts     INTEGER NOT NULL DEFAULT 0,
+    max_attempts INTEGER NOT NULL DEFAULT 3,
+    scheduled_at TEXT NOT NULL,
+    started_at   TEXT,
+    completed_at TEXT,
+    result       TEXT,
+    error        TEXT,
+    created_at   TEXT NOT NULL DEFAULT (strftime('%Y-%m-%dT%H:%M:%SZ', 'now'))
+);
+
+CREATE INDEX IF NOT EXISTS idx_jobs_dequeue
+    ON jobs (status, priority DESC, scheduled_at ASC)
+    WHERE status = 'pending';
+
+CREATE INDEX IF NOT EXISTS idx_jobs_status ON jobs (status);
+CREATE INDEX IF NOT EXISTS idx_jobs_type ON jobs (type);
+CREATE INDEX IF NOT EXISTS idx_jobs_cleanup ON jobs (status, completed_at)
+    WHERE status IN ('completed', 'failed', 'dead');
+```
+
+El indice parcial `idx_jobs_dequeue` es clave: SQLite lo usa exclusivamente para el `SELECT` de dequeue, manteniendo el scan rapido incluso con millones de jobs historicos en la tabla.
+
+### Funciones
+
+| Funcion | Purity | Firma (simplificada) | Descripcion |
+|---------|--------|---------------------|-------------|
+| `job_queue_create` | impure | `(db *sql.DB, tableName string) (*JobQueue, error)` | Crea tabla de jobs con indices, activa WAL mode, retorna handle |
+| `job_enqueue` | impure | `(q *JobQueue, jobType string, payload string, opts ...EnqueueOption) (string, error)` | Inserta job con priority, scheduled_at, max_attempts. Retorna job ID |
+| `job_dequeue` | impure | `(q *JobQueue, jobTypes ...string) (*Job, error)` | Reclama atomicamente el proximo job pending. Nil si cola vacia |
+| `job_complete` | impure | `(q *JobQueue, jobID string, result string) error` | Marca job como completed con resultado JSON |
+| `job_fail` | impure | `(q *JobQueue, jobID string, errMsg string) error` | Marca como failed, incrementa attempts. Si attempts >= max_attempts, marca como dead |
+| `job_worker` | impure | `(ctx context.Context, q *JobQueue, handler JobHandler, opts ...WorkerOption) error` | Goroutine que pollea la cola y ejecuta handler. Para con context |
+| `job_worker_pool` | impure | `(ctx context.Context, q *JobQueue, handler JobHandler, n int, opts ...WorkerOption) error` | Lanza N workers, espera a que todos terminen con graceful shutdown |
+| `job_cleanup` | impure | `(q *JobQueue, olderThan time.Duration) (int64, error)` | Elimina jobs completed/failed/dead mas antiguos que la retencion |
+| `job_status` | pure | `(counts map[JobStatus]int) string` | Formatea resumen legible: "pending: 5, running: 2, completed: 100, failed: 3, dead: 1" |
+
+### Dequeue atomico
+
+La operacion critica es `job_dequeue` — debe reclamar un job sin condiciones de carrera cuando hay multiples workers. SQLite no tiene `SELECT FOR UPDATE`, pero se puede lograr con una transaccion exclusiva:
+
+```go
+func JobDequeue(q *JobQueue, jobTypes ...string) (*Job, error) {
+    tx, err := q.DB.Begin()
+    if err != nil {
+        return nil, err
+    }
+    defer tx.Rollback()
+
+    // SELECT el proximo job elegible
+    query := fmt.Sprintf(`
+        SELECT id, type, payload, status, priority, attempts, max_attempts,
+               scheduled_at, result, error, created_at
+        FROM %s
+        WHERE status = 'pending'
+          AND scheduled_at <= strftime('%%Y-%%m-%%dT%%H:%%M:%%SZ', 'now')
+        ORDER BY priority DESC, scheduled_at ASC
+        LIMIT 1
+    `, q.TableName)
+
+    // Si se especifican tipos, filtrar
+    if len(jobTypes) > 0 {
+        // ... añadir AND type IN (...)
+    }
+
+    var job Job
+    err = tx.QueryRow(query).Scan(/* ... */)
+    if err == sql.ErrNoRows {
+        return nil, nil // cola vacia, no es error
+    }
+    if err != nil {
+        return nil, err
+    }
+
+    // UPDATE atomico dentro de la misma transaccion
+    _, err = tx.Exec(fmt.Sprintf(`
+        UPDATE %s SET status = 'running', started_at = strftime('%%Y-%%m-%%dT%%H:%%M:%%SZ', 'now')
+        WHERE id = ?
+    `, q.TableName), job.ID)
+    if err != nil {
+        return nil, err
+    }
+
+    return &job, tx.Commit()
+}
+```
+
+Con WAL mode, los writers serializan transacciones de escritura pero los readers no bloquean. Dado que `dequeue` es la unica operacion que necesita exclusividad de escritura y es extremadamente rapida (un SELECT indexado + un UPDATE por PK), la contención es minima incluso con varios workers.
+
+### Backoff exponencial en reintentos
+
+Cuando un job falla, `job_fail` incrementa `attempts` y lo marca como `failed`. El worker, al reclamar un job previamente fallido, respeta un backoff basado en el numero de intentos:
+
+```go
+// Delay antes de reintentar: base * 2^(attempts-1)
+// Intento 1: 5s, intento 2: 10s, intento 3: 20s
+backoff := baseDelay * time.Duration(1<<uint(job.Attempts-1))
+```
+
+Cuando `attempts >= max_attempts`, el job pasa a status `dead` (dead-letter). Los jobs dead no se reintentan automaticamente — quedan para inspeccion manual o re-enqueue explicito.
+
+### Worker loop
+
+```go
+func JobWorker(ctx context.Context, q *JobQueue, handler JobHandler, opts ...WorkerOption) error {
+    cfg := defaultWorkerConfig() // pollInterval: 1s, jobTypes: nil (todos)
+    for _, opt := range opts {
+        opt(&cfg)
+    }
+
+    for {
+        select {
+        case <-ctx.Done():
+            return ctx.Err()
+        default:
+        }
+
+        job, err := JobDequeue(q, cfg.jobTypes...)
+        if err != nil {
+            // log error, backoff, continue
+            time.Sleep(cfg.pollInterval)
+            continue
+        }
+
+        if job == nil {
+            // cola vacia, esperar
+            time.Sleep(cfg.pollInterval)
+            continue
+        }
+
+        // Ejecutar handler
+        if err := handler(*job); err != nil {
+            JobFail(q, job.ID, err.Error())
+        } else {
+            JobComplete(q, job.ID, "") // resultado vacio por defecto
+        }
+    }
+}
+```
+
+## Tareas
+
+### Fase 1: Tipos
+
+- [ ] **1.1** Crear tipo `Job` en `functions/infra/job.go` con `.md` en `types/infra/job.md`
+- [ ] **1.2** Crear tipo `JobQueue` en `functions/infra/job_queue.go` con `.md` en `types/infra/job_queue.md`
+- [ ] **1.3** Crear tipo `JobStatus` (sum) en `functions/infra/job_status_type.go` con `.md` en `types/infra/job_status.md`
+- [ ] **1.4** Crear tipo `JobHandler` en `functions/infra/job_handler.go` con `.md` en `types/infra/job_handler.md`
+
+### Fase 2: Funciones core (CRUD de jobs)
+
+- [ ] **2.1** `job_queue_create` — crea tabla con schema e indices, configura WAL mode, retorna `*JobQueue`
+- [ ] **2.2** `job_enqueue` — INSERT con UUID generado, soporte para options (priority, scheduled_at, max_attempts)
+- [ ] **2.3** `job_dequeue` — SELECT + UPDATE atomico en transaccion, nil si cola vacia
+- [ ] **2.4** `job_complete` — UPDATE status=completed, set completed_at y result
+- [ ] **2.5** `job_fail` — UPDATE status=failed|dead, incrementar attempts, set error
+- [ ] **2.6** `job_status` (pure) — recibe map de conteos, retorna string formateado
+
+### Fase 3: Workers y mantenimiento
+
+- [ ] **3.1** `job_worker` — goroutine con poll loop, context cancelable, backoff configurable
+- [ ] **3.2** `job_worker_pool` — lanza N workers con `errgroup`, graceful shutdown via context
+- [ ] **3.3** `job_cleanup` — DELETE de jobs completed/failed/dead mas antiguos que la retencion
+
+### Fase 4: Tests y validacion
+
+- [ ] **4.1** Tests para `job_queue_create` — verifica tabla e indices creados
+- [ ] **4.2** Tests para enqueue/dequeue — verifica atomicidad, prioridad, scheduling futuro
+- [ ] **4.3** Tests para complete/fail — verifica transiciones de estado, dead-letter tras max_attempts
+- [ ] **4.4** Tests para worker — verifica procesamiento, reintentos, graceful shutdown con context
+- [ ] **4.5** Tests de concurrencia — multiples goroutines dequeue simultaneo, ningun job procesado dos veces
+- [ ] **4.6** `fn index` y verificar que todas las funciones y tipos aparecen en registry.db
+- [ ] **4.7** Verificar `go vet -tags fts5` y `go test -tags fts5 -race ./functions/infra/`
+
+---
+
+## Ejemplo de uso
+
+### Cola de envio de emails
+
+```go
+// En una app que necesita enviar emails en background
+
+// 1. Crear la cola
+db, _ := sql.Open("sqlite3", "operations.db?_journal_mode=wal")
+queue, _ := infra.JobQueueCreate(db, "jobs")
+
+// 2. Encolar emails desde un handler HTTP
+func handleSignup(w http.ResponseWriter, r *http.Request) {
+    user := createUser(r)
+
+    payload, _ := json.Marshal(map[string]string{
+        "to":      user.Email,
+        "subject": "Bienvenido",
+        "body":    "Gracias por registrarte...",
+    })
+
+    jobID, _ := infra.JobEnqueue(queue, "send_email", string(payload),
+        infra.WithPriority(1),
+        infra.WithMaxAttempts(5),
+    )
+
+    log.Printf("email encolado: job=%s user=%s", jobID, user.Email)
+    infra.HttpJsonResponse(w, 201, user)
+}
+
+// 3. Worker que procesa los emails
+func emailHandler(job infra.Job) error {
+    var email struct {
+        To      string `json:"to"`
+        Subject string `json:"subject"`
+        Body    string `json:"body"`
+    }
+    json.Unmarshal([]byte(job.Payload), &email)
+    return sendEmail(email.To, email.Subject, email.Body)
+}
+
+// 4. Arrancar pool de workers
+ctx, cancel := signal.NotifyContext(context.Background(), os.Interrupt)
+defer cancel()
+
+infra.JobWorkerPool(ctx, queue, emailHandler, 3,
+    infra.WithPollInterval(2*time.Second),
+    infra.WithJobTypes("send_email"),
+)
+```
+
+### Cola de procesamiento de archivos
+
+```go
+// Encolar procesamiento de archivos con prioridad y scheduling
+
+// Job de alta prioridad: procesar ahora
+infra.JobEnqueue(queue, "process_csv", `{"path": "/data/ventas.csv", "format": "monthly"}`,
+    infra.WithPriority(10),
+)
+
+// Job schedulado para las 3am
+infra.JobEnqueue(queue, "generate_report", `{"type": "weekly", "format": "pdf"}`,
+    infra.WithScheduledAt(nextAt3AM()),
+)
+
+// Worker que despacha por tipo
+func fileHandler(job infra.Job) error {
+    switch job.Type {
+    case "process_csv":
+        return processCSV(job.Payload)
+    case "generate_report":
+        return generateReport(job.Payload)
+    default:
+        return fmt.Errorf("tipo de job desconocido: %s", job.Type)
+    }
+}
+```
+
+### Monitoreo y limpieza
+
+```go
+// Consultar estado de la cola
+rows, _ := db.Query(`
+    SELECT status, COUNT(*) FROM jobs GROUP BY status
+`)
+counts := map[infra.JobStatus]int{}
+for rows.Next() {
+    var s string; var c int
+    rows.Scan(&s, &c)
+    counts[infra.JobStatus(s)] = c
+}
+fmt.Println(infra.JobStatusSummary(counts))
+// → "pending: 12, running: 3, completed: 1458, failed: 7, dead: 2"
+
+// Limpiar jobs viejos (retener 7 dias)
+deleted, _ := infra.JobCleanup(queue, 7*24*time.Hour)
+log.Printf("limpiados %d jobs antiguos", deleted)
+```
+
+## Decisiones de diseno
+
+- **SQLite como backend unico:** Coherente con el stack del proyecto. No introduce dependencias externas (Redis, Postgres). WAL mode permite lecturas concurrentes sin bloquear al writer.
+- **Una tabla por cola, no una BD por cola:** Permite que una app tenga su cola en su propio `operations.db` usando un table name custom, o que varias colas convivan en la misma BD con distintos table names.
+- **Dequeue con transaccion, no con SKIP LOCKED:** SQLite no soporta `SKIP LOCKED`. La transaccion exclusiva es correcta y suficiente para el nivel de concurrencia esperado (decenas de workers, no miles).
+- **Poll-based, no event-driven:** SQLite no tiene `LISTEN/NOTIFY`. El polling con intervalo configurable (default 1s) es simple y predecible. Para la escala de uso esperada (apps personales, no sistemas de alta frecuencia), el overhead es despreciable.
+- **Functional options pattern (WithPriority, WithMaxAttempts):** Permite extender sin romper firmas existentes. Idomático en Go.
+- **Dead-letter como status, no como tabla separada:** Los jobs dead quedan en la misma tabla para inspeccion. `job_cleanup` los borra despues del periodo de retencion.
+- **Tipos nativos en firmas Go:** Siguiendo la regla del registry, las funciones usan `*sql.DB`, `string`, `time.Duration` — no tipos del registry en la firma. Los tipos (`Job`, `JobQueue`, etc.) se documentan en `uses_types`/`returns` del frontmatter.
+
+## Riesgos
+
+- **Contención de escritura con muchos workers:** SQLite serializa escrituras. Con WAL mode y transacciones cortas (microsegundos), esto no es problema para decenas de workers. Si alguna app necesitara miles de dequeues/segundo, habria que migrar a Postgres — pero eso esta fuera del scope de este registry.
+- **Polling waste con cola vacia:** Workers consumen CPU haciendo SELECT cada segundo aunque no haya trabajo. Mitigado con backoff adaptativo: si N polls consecutivos retornan nil, el intervalo crece gradualmente (hasta un techo configurable). Se resetea al primer job encontrado.
+- **Jobs zombies (running forever):** Si un worker muere sin completar ni fallar el job, queda en status `running` indefinidamente. Solucion: `job_worker` registra un timeout por job; si el handler no retorna en ese tiempo, lo marca como failed. Complementar con un sweep periodico que busque jobs `running` con `started_at` mas antiguo que un umbral.
+- **Payload sin schema:** El payload es JSON libre (`string`). No hay validacion de estructura al encolar. Esto es intencional — cada `JobHandler` valida su propio payload. Pero un typo en el payload causa errores en runtime, no en compile time.
+- **Limpieza manual:** `job_cleanup` debe invocarse explicitamente (desde un cron, un job periodico, o al arrancar la app). No hay garbage collection automatico. Documentar en el ejemplo de uso.