feat(queue): async delivery via Cloudflare Queues with retry backoff and DLQ

AGENTS.md

@@ -24,6 +24,7 @@ Core pipeline: GitHub Webhook → Worker (verify + filter + format) → Discord
 - Audit log: every admin operation (logins, group/route/member/invite changes) recorded in D1 `audit_logs`; pruned by the scheduled trigger after `AUDIT_RETENTION_DAYS` (default 90)
 - Group webhook log channel: optional `Group.logTarget` (Discord channel/thread or Telegram chat/topic) receives one summary message per webhook the group's routes dispatched (event, repo, delivery id, per route×target ✅/❌ outcome, green/red color); best-effort, not recorded in `send_logs`
 - Per-group forge branding: optional `Group.forgeSources` (a list of `{ host, type: "github" | "gitea", name? }` the group defines itself) labels each message's footer with the first entry whose type matches the event's provider and whose host matches the repository URL's hostname (GitHub matches `github.com`, so two Gitea instances can be `git1.example.com`/`git2.example.com`); the label is the entry's optional `name` (fallback: host); links are derived from the repo URL and footer icons use raster PNGs Discord renders (GitHub's `fluidicon.png`, Gitea's `/assets/img/favicon.png` — `.ico` favicons are silently ignored); Discord shows the footer icon + name, Telegram a linked name
+- Delivery queue: when the `QUEUE` binding is present, webhook ingress enqueues one message per event (not per target) to a Cloudflare Queue (`webhooker-delivery`); a Nitro `cloudflare:queue` plugin consumes batches, dispatches, and retries retryable failures (5xx/network/429-exhaustion) with exponential backoff (5s/30s/2m/10m) up to the queue `max_retries`, after which the DLQ (`webhooker-delivery-dlq`) marks the delivery dead. Oversized payloads (>~100 KB) are parked in KV (`queue:payload:*`) and resolved by the consumer; delivery state is tracked in KV (`delivery-state:*`). Without the `QUEUE` binding, dispatch stays inline (existing behavior)
 - Local dev: wrangler + Miniflare
 
 ## Architecture
@@ -42,13 +43,17 @@ server/                  # Nitro server
 ├── routes/              # H3 handlers: /health, /webhook[/:groupId], /discord/interactions, /telegram/webhook,
 │                        # /auth/github*, /admin/{login,logout,invite,api/**}, /api/{comment,merge,close,react,richheader}
 ├── tasks/               # scheduled (cron */5): discord-sync, telegram-sync, audit-prune
+├── plugins/             # Nitro plugins: queue-consumer (hooks cloudflare:queue → handleQueueBatch)
 ├── error-handler.ts     # JSON error handler
 └── lib/
     ├── types.ts         # Env, Config, Route, Filter, Group, WebhookEvent, NeutralMessage
     ├── config.ts        # loadRoutes/saveRoutes (KV config:routes, cache w/ 60s TTL), loadConfig from env
     ├── cf.ts            # cfEnv(event) — env bindings from event.context.cloudflare
     ├── http.ts          # shared HTTP helpers
-    ├── webhook.ts       # processWebhook/handleWebhook: tenant lookup, provider detect/verify/parse, dedup, scoped dispatch
+    ├── webhook.ts       # processWebhook/handleWebhook: tenant lookup, provider detect/verify/parse, dedup, enqueue (or inline dispatch)
+    ├── queue/           # Cloudflare Queue delivery pipeline
+    │   ├── delivery.ts  # DeliveryMessage, enqueueWebhook, retry backoff, delivery-state KV, payload-overflow parking
+    │   └── consumer.ts  # handleQueueBatch: resolve payload → dispatch → classify → ack/retry/DLQ
     ├── core/
     │   └── dispatch.ts  # Platform-neutral dispatch: match routes → formatEvent → driver.send/edit (recordSend + group filter + per-group webhook log)
     ├── events/
@@ -133,6 +138,8 @@ tests/                   # bun test unit tests (webhook, formatter, discord, tel
 - Record every dispatch attempt to D1 `send_logs` (route id, event, target, ok/error, duration, error code)
 - Serve a per-group webhook ingress (`POST /webhook/{groupId}`, per-group secret in KV `tenant:{groupId}`) for Gitea/classic-GitHub/custom senders; only that group's routes fire; dedup keys are provider- and tenant-scoped (`delivery:{provider}:{groupId}:{id}` via `kvIdempotencyStore`)
 - Issue a per-request correlation id (`requestId`) in webhook responses and dispatch logs
+- When the `QUEUE` binding is present, enqueue each verified webhook as a single Queue message (`webhooker-delivery`) instead of dispatching inline; the consumer resolves the payload, re-scopes routes to the tenant group, and dispatches; retryable failures (5xx/network/429-exhaustion) are retried with exponential backoff (5s/30s/2m/10m) up to the queue `max_retries`, then the DLQ marks the delivery dead
+- Track delivery state in KV (`delivery-state:*`: pending/processing/delivered/retrying/failed/dead) so redelivered messages are skipped idempotently; oversized payloads are parked in KV (`queue:payload:*`) and deleted after dispatch
 - Send a per-event summary (event, repo, delivery id, per route×target ✅/❌ outcome) to the group's `logTarget` when configured
 - Serve `/gh` slash commands + message context-menu commands + PR merge/close buttons + comment modals
 - Serve Telegram `/gh` commands (login/logout/comment/merge/close) via reply-message parsing
@@ -200,6 +207,7 @@ Rule: no functional change ships without its documentation; docs and code must n
 - **Routes**: KV key `config:routes` (JSON array, empty until configured)
 - **KV namespace**: Required binding for token/state/config/session storage
 - **D1 database**: Binding `DB` (database `webhooker`, id `214a0104-3235-47c0-b7bf-ddda95f3c8ac`) for `send_logs` + `audit_logs` + `discord_links` + `telegram_links` tables
+- **Queue**: optional `QUEUE` producer binding plus consumers `webhooker-delivery` and its DLQ `webhooker-delivery-dlq` (declared in `wrangler.jsonc`); when absent, webhook dispatch stays inline
 - **Access control**: `ADMIN_USER_IDS` (super admins), `ALLOW_SELF_SIGNUP` (optional personal group on first login), `AUDIT_RETENTION_DAYS` (default 90) — all plain env vars, not secrets
 - **Discord**: `DISCORD_PUBLIC_KEY` (Interactions Endpoint signature verification, from Discord Developer Portal) and `DISCORD_APPLICATION_ID` (optional, auto-resolved via `GET /oauth2/applications/@me` when omitted) are required for interactions
 - **Telegram**: `TELEGRAM_TOKEN` (Bot API token from BotFather) required for Telegram routes; `TELEGRAM_WEBHOOK_SECRET` (optional secret token for `POST /telegram/webhook` verification); avatars are sent as a link-preview card via the built-in `GET /api/richheader` (overridable with `TELEGRAM_RICH_HEADER_HOST`)
@@ -215,6 +223,9 @@ bunx wrangler kv namespace create KV
 # Update wrangler.jsonc with KV ID
 bunx wrangler d1 create webhooker
 # Update wrangler.jsonc d1_databases with the database ID
+bunx wrangler queues create webhooker-delivery
+bunx wrangler queues create webhooker-delivery-dlq
+# Queues are declared in wrangler.jsonc (QUEUE binding); no env var needed
 bun run db:migrate:prod   # wrangler d1 migrations apply webhooker --remote (migrations/0001..0005)
 bunx wrangler deploy
 ```

README.md

@@ -19,13 +19,14 @@ GitHub / Gitea webhook → Discord / Telegram dispatcher. Receives webhook event
 - **Discord Interactions Endpoint** (Ed25519-verified) for `/gh` slash commands, message context-menu commands, PR merge/close buttons, and comment modals
 - **Telegram `/gh` commands** (login/logout/comment/merge/close) via the Telegram webhook, with avatar link-preview cards
 - Cloudflare KV for token/state/config/session storage + D1 for send logs and platform account links
+- **Async delivery via Cloudflare Queues** — when the `QUEUE` binding is present, verified webhooks are enqueued to `webhooker-delivery` and dispatched by a consumer with exponential retry backoff (5s/30s/2m/10m) and a dead-letter queue (`webhooker-delivery-dlq`); oversized payloads are parked in KV. Without the binding, dispatch stays inline
 - Graceful degradation (webhook-only mode if Discord unavailable)
 
 ## Architecture
 
 ```text
 GitHub Webhook → Cloudflare Worker (Nuxt 4 / Nitro)
-                 ├── POST /webhook → verify → dedup → filter → format → Discord (REST) / Telegram (Bot API)
+                 ├── POST /webhook → verify → dedup → enqueue (Queue) → dispatch → Discord (REST) / Telegram (Bot API)
                  ├── POST /discord/interactions → verify (Ed25519) → handle command/button/modal
                  ├── POST /telegram/webhook → verify (secret token) → handle /gh commands
                  ├── GET  /auth/github → OAuth flow
@@ -37,8 +38,9 @@ GitHub Webhook → Cloudflare Worker (Nuxt 4 / Nitro)
 
 - **Cloudflare Worker** — HTTP ingress, signature verification, routing, platform dispatch
 - **Interactions Endpoint** — HTTPS callback (no Discord Gateway connection, no Durable Object); the bot stays offline and commands are registered via the API
-- **KV** — token storage (`token:{userId}`), OAuth state (`state:{hex}`), route config (`config:routes`), group config (`config:groups`), admin sessions (`session:{id}`), delivery dedup (`delivery:{id}`), message-update tracking (`msg:*`)
+- **KV** — token storage (`token:{userId}`), OAuth state (`state:{hex}`), route config (`config:routes`), group config (`config:groups`), admin sessions (`session:{id}`), delivery dedup (`delivery:{provider}:{groupId}:{id}`), delivery state (`delivery-state:*`), message-update tracking (`msg:*`)
 - **D1** — send logs (`send_logs`), Discord↔GitHub links (`discord_links`), Telegram↔GitHub links (`telegram_links`)
+- **Queue** — async delivery when `QUEUE` is bound: `webhooker-delivery` (exponential retry) + DLQ `webhooker-delivery-dlq`; oversized payloads parked in KV (`queue:payload:*`)
 
 ## Quick Start
 
@@ -156,7 +158,7 @@ Every filter supports plain text, `*`/`?` globs, and `/regex/` patterns (case-in
 - **GitHub App** — create the app, subscribe to events, configure OAuth, and set the _Setup URL_ for tenant isolation: see [GitHub App Setup](https://webhooker.docs.worldexecute.me/guide/deployment#github-app-setup)
 - **Discord bot** — create the bot, invite it with `applications.commands` (combined permission integer `274877910016`), and configure the Interactions Endpoint: see [Discord Bot Setup](https://webhooker.docs.worldexecute.me/guide/deployment#discord-bot-setup). The bot never connects to the Discord Gateway, so it shows as **offline** — messaging is unaffected (always REST).
 - **Telegram bot** — create the bot with [@BotFather](https://t.me/BotFather), set `TELEGRAM_TOKEN` (optional `TELEGRAM_WEBHOOK_SECRET`); the webhook is synced automatically by the scheduled trigger: see [Telegram Bot Setup](https://webhooker.docs.worldexecute.me/guide/deployment#telegram-bot-setup)
-- **Deployment** — KV namespace, D1 database + migrations, secrets, and deploy: see the [Deployment guide](https://webhooker.docs.worldexecute.me/guide/deployment)
+- **Deployment** — KV namespace, D1 database + migrations, optional Queues, secrets, and deploy: see the [Deployment guide](https://webhooker.docs.worldexecute.me/guide/deployment)
 
 ### Bot Commands (comment on GitHub as yourself)
 

README.zh.md

@@ -19,13 +19,14 @@ GitHub / Gitea webhook → Discord / Telegram 分发服务。通过 Cloudflare W
 - **Discord Interactions Endpoint**（Ed25519 验签）支持 `/gh` 斜杠命令、消息右键菜单命令、PR 合并/关闭按钮与评论 modal
 - **Telegram `/gh` 命令**（login/logout/comment/merge/close），通过 Telegram webhook 接收，头像以链接预览卡片呈现
 - Cloudflare KV 存储 token/状态/配置/会话 + D1 存储发送日志与平台账号绑定
+- **Cloudflare Queues 异步投递** —— 绑定 `QUEUE` 时，已验签的 webhook 会入队到 `webhooker-delivery`，由消费者分发，带指数退避重试（5s/30s/2m/10m）与死信队列（`webhooker-delivery-dlq`）；超大负载暂存于 KV。未绑定则保持同步分发
 - 优雅降级（Discord 不可用时仅 webhook 模式）
 
 ## 架构
 
 ```text
 GitHub Webhook → Cloudflare Worker (Nuxt 4 / Nitro)
-                 ├── POST /webhook → 验证 → 去重 → 过滤 → 格式化 → Discord (REST) / Telegram (Bot API)
+                 ├── POST /webhook → 验证 → 去重 → 入队 (Queue) → 分发 → Discord (REST) / Telegram (Bot API)
                  ├── POST /discord/interactions → 验证 (Ed25519) → 处理命令/按钮/modal
                  ├── POST /telegram/webhook → 验证 (secret token) → 处理 /gh 命令
                  ├── GET  /auth/github → OAuth 流程
@@ -37,8 +38,9 @@ GitHub Webhook → Cloudflare Worker (Nuxt 4 / Nitro)
 
 - **Cloudflare Worker** — HTTP 入口、签名验证、路由分发
 - **Interactions Endpoint** — HTTPS 回调（无 Discord Gateway 连接、无 Durable Object）；bot 保持离线，命令通过 API 注册
-- **KV** — Token 存储（`token:{userId}`）、OAuth state（`state:{hex}`）、路由配置（`config:routes`）、分组配置（`config:groups`）、管理员会话（`session:{id}`）、投递去重（`delivery:{id}`）、消息更新追踪（`msg:*`）
+- **KV** — Token 存储（`token:{userId}`）、OAuth state（`state:{hex}`）、路由配置（`config:routes`）、分组配置（`config:groups`）、管理员会话（`session:{id}`）、投递去重（`delivery:{provider}:{groupId}:{id}`）、投递状态（`delivery-state:*`）、消息更新追踪（`msg:*`）
 - **D1** — 发送日志（`send_logs`）、Discord↔GitHub 绑定（`discord_links`）、Telegram↔GitHub 绑定（`telegram_links`）
+- **Queue** — 绑定 `QUEUE` 时异步投递：`webhooker-delivery`（指数退避重试）+ 死信队列 `webhooker-delivery-dlq`；超大负载暂存于 KV（`queue:payload:*`）
 
 ## 快速开始
 
@@ -156,7 +158,7 @@ bunx wrangler dev    # 启动本地开发服务器
 - **GitHub App** — 创建应用、订阅事件、配置 OAuth 与 _Setup URL_（租户隔离）：见 [GitHub App 配置](https://webhooker.docs.worldexecute.me/zh/guide/deployment#github-app-设置)
 - **Discord 机器人** — 创建机器人、以 `applications.commands` scope 邀请（组合权限整数 `274877910016`）、配置 Interactions Endpoint：见 [Discord Bot 配置](https://webhooker.docs.worldexecute.me/zh/guide/deployment#discord-bot-设置)。bot 从不连接 Discord Gateway，因此显示为**离线**——消息推送不受影响（始终走 REST）。
 - **Telegram 机器人** — 用 [@BotFather](https://t.me/BotFather) 创建机器人，设置 `TELEGRAM_TOKEN`（可选 `TELEGRAM_WEBHOOK_SECRET`）；webhook 由定时任务自动同步：见 [Telegram 机器人配置](https://webhooker.docs.worldexecute.me/zh/guide/deployment#telegram-机器人配置)
-- **部署** — KV 命名空间、D1 数据库与迁移、密钥、部署：见[部署指南](https://webhooker.docs.worldexecute.me/zh/guide/deployment)
+- **部署** — KV 命名空间、D1 数据库与迁移、可选 Queues、密钥、部署：见[部署指南](https://webhooker.docs.worldexecute.me/zh/guide/deployment)
 
 ### Bot 指令（以本人身份评论 GitHub）
 

docs/guide/deployment.md

@@ -85,7 +85,18 @@ bunx wrangler d1 execute webhooker --remote --file ./migrations/0005_audit_logs.
 
 :::
 
-### 4. Deploy
+### 4. Create Queues (Optional)
+
+The `QUEUE` binding routes webhook delivery through Cloudflare Queues (async dispatch with retry backoff and a dead-letter queue). Skip this step to keep dispatch inline (synchronous).
+
+```bash
+bunx wrangler queues create webhooker-delivery
+bunx wrangler queues create webhooker-delivery-dlq
+```
+
+The queues are already declared in `wrangler.jsonc` (`queues.producers` / `queues.consumers`), so no binding change is needed. The `webhooker-delivery` consumer retries retryable failures with exponential backoff (5s/30s/2m/10m) up to `max_retries`, after which the message is moved to `webhooker-delivery-dlq` and marked dead.
+
+### 5. Deploy
 
 ```bash
 bunx wrangler deploy
@@ -93,13 +104,13 @@ bunx wrangler deploy
 
 Your worker is now live at `https://webhooker.<your-subdomain>.workers.dev`.
 
-### 5. Configure GitHub Webhook
+### 6. Configure GitHub Webhook
 
 1. Go to your GitHub App settings
 2. Set **Webhook URL** to `https://webhooker.<your-subdomain>.workers.dev/webhook`
 3. Set **Webhook secret** to match `GITHUB_WEBHOOK_SECRET`
 
-### 6. (Optional) Configure Gitea Webhook
+### 7. (Optional) Configure Gitea Webhook
 
 1. In your Gitea repo, go to **Settings → Webhooks → Add Webhook → Gitea**
 2. Set **Target URL** to `https://webhooker.<your-subdomain>.workers.dev/webhook`
@@ -181,4 +192,4 @@ To use a custom domain instead of `*.workers.dev`:
 3. Update `BASE_URL` to match
 
 > [!NOTE]
-> This project is a Cloudflare Worker. It requires the KV and D1 bindings declared in `wrangler.jsonc`, so it cannot run as a standalone Node/container process.
+> This project is a Cloudflare Worker. It requires the KV and D1 bindings declared in `wrangler.jsonc`, so it cannot run as a standalone Node/container process. The Queues binding (`QUEUE`) is optional — without it, webhook dispatch stays inline.

docs/guide/storage.md

@@ -12,8 +12,10 @@
 | `state:{hex}`                  | `{ redirectTo, expiresAt, discordUserId?, telegramUserId?, telegramChatId? }` | 600 seconds        |
 | `invite:{token}`               | `{ groupId, role, expiresAt, createdBy, note? }`                              | 7 days             |
 | `invite:group:{id}`            | Token index per group (keeps invite listing consistent)                       | Permanent          |
-| `delivery:{id}`                | Webhook delivery id (dedup marker)                                            | 300 seconds        |
+| `delivery:{provider}:{groupId}:{id}` | Webhook delivery dedup (provider- and tenant-scoped)                              | 300 seconds        |
-| `delivery:{groupId}:{id}`      | Tenant-scoped delivery dedup for the per-group webhook ingress                | 300 seconds        |
+| `delivery-state:{provider}:{groupId}:{id}` | Queue delivery state (`pending`/`processing`/`delivered`/`retrying`/`failed`/`dead`) | 1 day              |
+| `queue:payload:{provider}:{groupId}:{id}`  | Oversized webhook payload parked for the queue consumer                        | 1 day              |
+| `nonce:{nonce}`                | Custom-webhook replay protection nonce (single use)                            | 600 seconds        |
 | `tenant:{groupId}`             | Per-group webhook secret (64-char hex, generated from the console)             | Permanent          |
 | `msg:{routeId}:{key}:{target}` | Message id tracking for in-place updates (e.g. `workflow_run` / `check_run`)  | 7 days             |
 | `cmd:guild:{id}`               | Guild id whose commands were registered (dedup)                               | Permanent          |

docs/zh/guide/deployment.md

@@ -85,7 +85,18 @@ bunx wrangler d1 execute webhooker --remote --file ./migrations/0005_audit_logs.
 
 :::
 
-### 4. 部署
+### 4. 创建队列（可选）
+
+`QUEUE` 绑定会通过 Cloudflare Queues 投递 webhook（异步分发，带重试退避与死信队列）。跳过此步则保持同步内联分发。
+
+```bash
+bunx wrangler queues create webhooker-delivery
+bunx wrangler queues create webhooker-delivery-dlq
+```
+
+队列已在 `wrangler.jsonc` 中声明（`queues.producers` / `queues.consumers`），无需修改绑定。`webhooker-delivery` 消费者对可重试失败做指数退避重试（5s/30s/2m/10m），达到 `max_retries` 后消息进入 `webhooker-delivery-dlq` 并标记为 dead。
+
+### 5. 部署
 
 ```bash
 bunx wrangler deploy
@@ -93,13 +104,13 @@ bunx wrangler deploy
 
 Worker 现在可通过 `https://webhooker.<your-subdomain>.workers.dev` 访问。
 
-### 5. 配置 GitHub Webhook
+### 6. 配置 GitHub Webhook
 
 1. 进入 GitHub App 设置页面
 2. 设置 **Webhook URL** 为 `https://webhooker.<your-subdomain>.workers.dev/webhook`
 3. 设置 **Webhook secret** 与 `GITHUB_WEBHOOK_SECRET` 一致
 
-### 6.（可选）配置 Gitea Webhook
+### 7.（可选）配置 Gitea Webhook
 
 1. 在 Gitea 仓库中进入 **设置 → Web 钩子 → 添加 Web 钩子 → Gitea**
 2. 设置 **目标 URL** 为 `https://webhooker.<your-subdomain>.workers.dev/webhook`
@@ -181,4 +192,4 @@ Worker 现在可通过 `https://webhooker.<your-subdomain>.workers.dev` 访问
 3. 更新 `BASE_URL` 以匹配
 
 > [!NOTE]
-> 本项目是一个 Cloudflare Worker，依赖 `wrangler.jsonc` 中声明的 KV 与 D1 绑定，无法作为独立的 Node/容器进程运行。
+> 本项目是一个 Cloudflare Worker，依赖 `wrangler.jsonc` 中声明的 KV 与 D1 绑定，无法作为独立的 Node/容器进程运行。Queues 绑定（`QUEUE`）可选——未绑定时 webhook 分发保持内联同步。

docs/zh/guide/storage.md

@@ -12,8 +12,10 @@
 | `state:{hex}`                  | `{ redirectTo, expiresAt, discordUserId?, telegramUserId?, telegramChatId? }` | 600 秒             |
 | `invite:{token}`               | `{ groupId, role, expiresAt, createdBy, note? }`                              | 7 天               |
 | `invite:group:{id}`            | 每组的 Token 索引（保证邀请列表一致性）                                       | 永久               |
-| `delivery:{id}`                | Webhook 投递 id（去重标记）                                                   | 300 秒             |
+| `delivery:{provider}:{groupId}:{id}` | Webhook 投递去重（按 provider 与租户隔离）                                   | 300 秒             |
-| `delivery:{groupId}:{id}`      | 分组级 webhook 入口的租户级投递去重                                           | 300 秒             |
+| `delivery-state:{provider}:{groupId}:{id}` | 队列投递状态（`pending`/`processing`/`delivered`/`retrying`/`failed`/`dead`） | 1 天               |
+| `queue:payload:{provider}:{groupId}:{id}`  | 暂存供队列消费者读取的超大 webhook 负载                                     | 1 天               |
+| `nonce:{nonce}`                | 自定义 webhook 重放防护 nonce（一次性）                                      | 600 秒             |
 | `tenant:{groupId}`             | 分组 webhook secret（64 位 hex，控制台生成）                                 | 永久               |
 | `msg:{routeId}:{key}:{target}` | 原地更新用消息 id 追踪（如 `workflow_run` / `check_run`）                     | 7 天               |
 | `cmd:guild:{id}`               | 已注册命令的服务器 id（去重）                                                 | 永久               |

server/lib/core/dispatch.ts

@@ -13,6 +13,7 @@ import {
 } from "../web/groups";
 import { getDriver } from "../drivers";
 import type { SendResult } from "../drivers/types";
+import type { DispatchFailure, DispatchSummary } from "../queue/delivery";
 
 /** One dispatch attempt (route × target), collected for the group webhook log. */
 interface DispatchAttempt {
@@ -22,6 +23,8 @@ interface DispatchAttempt {
   target: string;
   ok: boolean;
   error?: string;
+  errorCode?: string;
+  status?: number;
 }
 
 export async function dispatchEvent(
@@ -29,7 +32,7 @@ export async function dispatchEvent(
   event: WebhookEvent,
   env: Env,
   groups?: Group[],
-): Promise<void> {
+): Promise<DispatchSummary> {
   const loadedGroups = groups ?? (await loadGroups(env.KV));
   const groupById = new Map(loadedGroups.map((g) => [g.id, g]));
 
@@ -76,6 +79,16 @@ export async function dispatchEvent(
 
   await sendGroupLogs(attempts);
 
+  const failures: DispatchFailure[] = attempts
+    .filter((a) => !a.ok)
+    .map((a) => ({
+      target: a.target,
+      error: a.error,
+      errorCode: a.errorCode,
+      status: a.status,
+    }));
+  return { attempts: attempts.length, failures };
+
   async function sendGroupLogs(list: DispatchAttempt[]): Promise<void> {
     const byGroup = new Map<string, DispatchAttempt[]>();
     for (const a of list) {
@@ -309,6 +322,8 @@ export async function dispatchEvent(
             target: targetStr,
             ok: false,
             error,
+            errorCode: result.errorCode,
+            status: result.status,
           });
           await recordSend(env.DB, {
             ...base,

server/lib/queue/consumer.ts

@@ -0,0 +1,90 @@
+import type { Env, WebhookEvent, WebhookProvider } from "../types";
+import { dispatchEvent } from "../core/dispatch";
+import { loadConfig } from "../config";
+import { loadGroups } from "../web/groups";
+import { log } from "../lib/log";
+import {
+  DELIVERY_DLQ,
+  type DeliveryMessage,
+  classifyDelivery,
+  deliveryStateKey,
+  discardPayload,
+  getDeliveryState,
+  resolvePayload,
+  retryDelay,
+  setDeliveryState,
+} from "./delivery";
+
+export async function handleQueueBatch(
+  batch: MessageBatch<DeliveryMessage>,
+  env: Env,
+): Promise<void> {
+  for (const message of batch.messages) {
+    const body = message.body;
+    if (batch.queue === DELIVERY_DLQ) {
+      await markDead(env, body);
+      message.ack();
+      continue;
+    }
+    await processMessage(env, body, message);
+  }
+}
+
+async function processMessage(
+  env: Env,
+  body: DeliveryMessage,
+  message: Message<DeliveryMessage>,
+): Promise<void> {
+  const key = deliveryStateKey(body.provider, body.groupId, body.deliveryId);
+  const prior = await getDeliveryState(env, key);
+  if (prior === "delivered" || prior === "dead") {
+    message.ack();
+    return;
+  }
+
+  await setDeliveryState(env, key, "processing");
+
+  const payload = await resolvePayload(env, body);
+  const event: WebhookEvent = {
+    event: body.event,
+    payload,
+    deliveryId: body.deliveryId,
+    provider: body.provider as WebhookProvider,
+    installationId: body.installationId,
+  };
+
+  try {
+    const config = await loadConfig(env);
+    if (body.groupId) {
+      config.routes = config.routes.filter((r) => r.groupId === body.groupId);
+    }
+    const groups = await loadGroups(env.KV);
+    const summary = await dispatchEvent(config, event, env, groups);
+    const { failed, retryable } = classifyDelivery(summary);
+
+    if (!failed) {
+      await setDeliveryState(env, key, "delivered");
+      await discardPayload(env, body);
+      message.ack();
+      return;
+    }
+    if (retryable) {
+      await setDeliveryState(env, key, "retrying");
+      message.retry({ delaySeconds: retryDelay(message.attempts) });
+      return;
+    }
+    await setDeliveryState(env, key, "failed");
+    await discardPayload(env, body);
+    message.ack();
+  } catch (err) {
+    log.error({ deliveryId: body.deliveryId, err }, "Queue delivery failed");
+    await setDeliveryState(env, key, "retrying");
+    message.retry({ delaySeconds: retryDelay(message.attempts) });
+  }
+}
+
+async function markDead(env: Env, body: DeliveryMessage): Promise<void> {
+  const key = deliveryStateKey(body.provider, body.groupId, body.deliveryId);
+  await setDeliveryState(env, key, "dead");
+  await discardPayload(env, body);
+}

server/lib/queue/delivery.ts

@@ -0,0 +1,133 @@
+import type { Env } from "../types";
+
+export type DeliveryStatus =
+  "pending" | "processing" | "delivered" | "retrying" | "failed" | "dead";
+
+export interface DeliveryMessage {
+  deliveryId: string;
+  groupId?: string;
+  provider: string;
+  event: string;
+  payload?: Record<string, unknown>;
+  payloadRef?: string;
+  installationId?: number;
+  receivedAt: number;
+  requestId?: string;
+}
+
+export interface DispatchFailure {
+  target: string;
+  error?: string;
+  errorCode?: string;
+  status?: number;
+}
+
+export interface DispatchSummary {
+  attempts: number;
+  failures: DispatchFailure[];
+}
+
+export const DELIVERY_QUEUE = "webhooker-delivery";
+export const DELIVERY_DLQ = "webhooker-delivery-dlq";
+
+const MAX_QUEUE_MESSAGE_BYTES = 100_000;
+const PAYLOAD_KV_TTL_SECONDS = 60 * 60 * 24;
+const STATE_KV_TTL_SECONDS = 60 * 60 * 24;
+
+const RETRYABLE_ERROR_CODES = new Set(["DISCORD_5XX", "TELEGRAM_5XX", "NETWORK", "RETRIES"]);
+
+const RETRY_DELAYS_SECONDS = [5, 30, 120, 600];
+
+export function isRetryableError(code?: string): boolean {
+  return code == null || RETRYABLE_ERROR_CODES.has(code);
+}
+
+export function classifyDelivery(summary: DispatchSummary): {
+  failed: boolean;
+  retryable: boolean;
+} {
+  if (summary.failures.length === 0) return { failed: false, retryable: false };
+  const retryable = summary.failures.every((f) => isRetryableError(f.errorCode));
+  const permanent = summary.failures.some((f) => !isRetryableError(f.errorCode));
+  return { failed: true, retryable: retryable && !permanent };
+}
+
+export function retryDelay(attempt: number): number {
+  if (attempt < 1) return RETRY_DELAYS_SECONDS[0];
+  const idx = Math.min(attempt - 1, RETRY_DELAYS_SECONDS.length - 1);
+  return RETRY_DELAYS_SECONDS[idx];
+}
+
+function scopeKey(provider: string, groupId: string | undefined, deliveryId: string): string {
+  return `${provider}:${groupId ?? "global"}:${deliveryId}`;
+}
+
+export function deliveryStateKey(
+  provider: string,
+  groupId: string | undefined,
+  deliveryId: string,
+): string {
+  return `delivery-state:${scopeKey(provider, groupId, deliveryId)}`;
+}
+
+function payloadKey(provider: string, groupId: string | undefined, deliveryId: string): string {
+  return `queue:payload:${scopeKey(provider, groupId, deliveryId)}`;
+}
+
+export async function setDeliveryState(
+  env: Env,
+  key: string,
+  status: DeliveryStatus,
+): Promise<void> {
+  await env.KV.put(key, JSON.stringify({ status, at: Date.now() }), {
+    expirationTtl: STATE_KV_TTL_SECONDS,
+  });
+}
+
+export async function getDeliveryState(env: Env, key: string): Promise<DeliveryStatus | null> {
+  const raw = await env.KV.get(key);
+  if (!raw) return null;
+  try {
+    return (JSON.parse(raw) as { status?: DeliveryStatus }).status ?? null;
+  } catch {
+    return null;
+  }
+}
+
+export async function enqueueWebhook(env: Env, message: DeliveryMessage): Promise<void> {
+  const queue = env.QUEUE;
+  if (!queue) throw new Error("QUEUE binding is not configured");
+  const { payload, ...rest } = message;
+  const direct: DeliveryMessage = { ...rest, payload };
+  if (JSON.stringify(direct).length <= MAX_QUEUE_MESSAGE_BYTES) {
+    await queue.send(direct);
+    return;
+  }
+  const payloadRef = payloadKey(message.provider, message.groupId, message.deliveryId);
+  await env.KV.put(payloadRef, JSON.stringify(payload ?? {}), {
+    expirationTtl: PAYLOAD_KV_TTL_SECONDS,
+  });
+  await queue.send({ ...rest, payloadRef });
+}
+
+export async function resolvePayload(
+  env: Env,
+  message: DeliveryMessage,
+): Promise<Record<string, unknown>> {
+  if (message.payload) return message.payload;
+  if (message.payloadRef) {
+    const raw = await env.KV.get(message.payloadRef);
+    if (raw) {
+      try {
+        return JSON.parse(raw) as Record<string, unknown>;
+      } catch {
+        return {};
+      }
+    }
+  }
+  return {};
+}
+
+export async function discardPayload(env: Env, message: DeliveryMessage): Promise<void> {
+  if (message.payloadRef) await env.KV.delete(message.payloadRef);
+}

server/lib/types.ts

@@ -27,6 +27,7 @@ export interface Env {
   ASSETS?: Fetcher;
   KV: KVNamespace;
   DB: D1Database;
+  QUEUE?: Queue;
 }
 
 export interface Config {

server/lib/webhook.ts

@@ -11,6 +11,7 @@ import { cfEnv, cfWaitUntil, headersFrom } from "./cf";
 import { log } from "./lib/log";
 import { deliveryKey, kvIdempotencyStore } from "./lib/idempotency";
 import { newCorrelationId } from "./lib/correlation";
+import { enqueueWebhook, type DeliveryMessage } from "./queue/delivery";
 
 const MAX_BODY_SIZE = 1024 * 1024;
 
@@ -135,6 +136,24 @@ export async function processWebhook(
     config.routes = config.routes.filter((r) => r.groupId === tenantId);
   }
 
+  if (env.QUEUE) {
+    const message: DeliveryMessage = {
+      deliveryId: event.deliveryId ?? requestId,
+      groupId: tenantId,
+      provider: provider.id,
+      event: event.event,
+      payload: event.payload,
+      installationId: event.installationId,
+      receivedAt: Date.now(),
+      requestId,
+    };
+    const enqueue = enqueueWebhook(env, message).catch((err) =>
+      log.error({ requestId, err }, "Failed to enqueue webhook"),
+    );
+    waitUntil(enqueue);
+    return { status: 200, body: { ok: true, requestId } };
+  }
+
   const dispatch = dispatchEvent(config, event, env, groups).catch((err) =>
     log.error({ requestId, err }, "Dispatch failed"),
   );

server/plugins/queue-consumer.ts

@@ -0,0 +1,8 @@
+import type { Env } from "../lib/types";
+import { handleQueueBatch } from "../lib/queue/consumer";
+
+export default defineNitroPlugin((nitroApp) => {
+  nitroApp.hooks.hook("cloudflare:queue", async ({ batch, env }) => {
+    await handleQueueBatch(batch, env as Env);
+  });
+});

tests/delivery.test.ts

@@ -0,0 +1,211 @@
+import { describe, it, expect } from "bun:test";
+import {
+  classifyDelivery,
+  deliveryStateKey,
+  discardPayload,
+  enqueueWebhook,
+  getDeliveryState,
+  isRetryableError,
+  resolvePayload,
+  retryDelay,
+  setDeliveryState,
+  type DeliveryMessage,
+  type DispatchSummary,
+} from "../server/lib/queue/delivery";
+import type { Env } from "../server/lib/types";
+
+function createMockKV(): { kv: KVNamespace; store: Map<string, string> } {
+  const store = new Map<string, string>();
+  const kv = {
+    get: async (key: string) => (store.has(key) ? store.get(key)! : null),
+    put: async (key: string, value: string) => {
+      store.set(key, value);
+    },
+    delete: async (key: string) => {
+      store.delete(key);
+    },
+    list: async () => ({ keys: [], list_complete: true, cacheStatus: null }),
+  } as unknown as KVNamespace;
+  return { kv, store };
+}
+
+function createMockQueue(): { queue: Queue; sent: DeliveryMessage[] } {
+  const sent: DeliveryMessage[] = [];
+  const queue = {
+    send: async (message: DeliveryMessage) => {
+      sent.push(message);
+    },
+    sendBatch: async () => {},
+  } as unknown as Queue;
+  return { queue, sent };
+}
+
+function createEnv(kv: KVNamespace, queue?: Queue): Env {
+  return {
+    GITHUB_WEBHOOK_SECRET: "secret",
+    KV: kv,
+    DB: {} as D1Database,
+    QUEUE: queue,
+  };
+}
+
+function message(overrides: Partial<DeliveryMessage> = {}): DeliveryMessage {
+  return {
+    deliveryId: "d1",
+    provider: "github",
+    event: "push",
+    payload: { ref: "refs/heads/main" },
+    receivedAt: Date.now(),
+    ...overrides,
+  };
+}
+
+describe("isRetryableError", () => {
+  it("treats undefined as retryable", () => {
+    expect(isRetryableError(undefined)).toBe(true);
+  });
+  it("treats 5xx/network/retries as retryable", () => {
+    expect(isRetryableError("DISCORD_5XX")).toBe(true);
+    expect(isRetryableError("TELEGRAM_5XX")).toBe(true);
+    expect(isRetryableError("NETWORK")).toBe(true);
+    expect(isRetryableError("RETRIES")).toBe(true);
+  });
+  it("treats 4xx and config errors as permanent", () => {
+    expect(isRetryableError("DISCORD_ERROR")).toBe(false);
+    expect(isRetryableError("TELEGRAM_ERROR")).toBe(false);
+    expect(isRetryableError("NO_TOKEN")).toBe(false);
+    expect(isRetryableError("NO_TARGET")).toBe(false);
+  });
+});
+
+describe("retryDelay", () => {
+  it("follows the exponential backoff schedule", () => {
+    expect(retryDelay(1)).toBe(5);
+    expect(retryDelay(2)).toBe(30);
+    expect(retryDelay(3)).toBe(120);
+    expect(retryDelay(4)).toBe(600);
+    expect(retryDelay(10)).toBe(600);
+  });
+  it("clamps sub-one attempts to the first delay", () => {
+    expect(retryDelay(0)).toBe(5);
+  });
+});
+
+describe("classifyDelivery", () => {
+  it("reports success when no failures", () => {
+    expect(classifyDelivery({ attempts: 1, failures: [] })).toEqual({
+      failed: false,
+      retryable: false,
+    });
+  });
+  it("is retryable when every failure is retryable", () => {
+    const summary: DispatchSummary = {
+      attempts: 2,
+      failures: [{ target: "c1", errorCode: "DISCORD_5XX" }],
+    };
+    expect(classifyDelivery(summary)).toEqual({ failed: true, retryable: true });
+  });
+  it("is not retryable when any failure is permanent", () => {
+    const summary: DispatchSummary = {
+      attempts: 2,
+      failures: [
+        { target: "c1", errorCode: "DISCORD_5XX" },
+        { target: "c2", errorCode: "DISCORD_ERROR" },
+      ],
+    };
+    expect(classifyDelivery(summary)).toEqual({ failed: true, retryable: false });
+  });
+});
+
+describe("delivery state", () => {
+  it("round-trips status via KV", async () => {
+    const { kv } = createMockKV();
+    const env = createEnv(kv);
+    const key = deliveryStateKey("github", "g1", "d1");
+    await setDeliveryState(env, key, "processing");
+    expect(await getDeliveryState(env, key)).toBe("processing");
+    await setDeliveryState(env, key, "delivered");
+    expect(await getDeliveryState(env, key)).toBe("delivered");
+  });
+
+  it("returns null for an unknown key", async () => {
+    const { kv } = createMockKV();
+    expect(await getDeliveryState(createEnv(kv), "delivery-state:nope")).toBeNull();
+  });
+
+  it("scopes state keys by provider, group and delivery id", () => {
+    expect(deliveryStateKey("github", "g1", "d1")).toBe("delivery-state:github:g1:d1");
+    expect(deliveryStateKey("github", undefined, "d1")).toBe("delivery-state:github:global:d1");
+  });
+});
+
+describe("enqueueWebhook", () => {
+  it("sends the message directly when it fits", async () => {
+    const { kv } = createMockKV();
+    const { queue, sent } = createMockQueue();
+    const env = createEnv(kv, queue);
+    await enqueueWebhook(env, message());
+    expect(sent).toHaveLength(1);
+    expect(sent[0].deliveryId).toBe("d1");
+    expect(sent[0].payload).toEqual({ ref: "refs/heads/main" });
+    expect(sent[0].payloadRef).toBeUndefined();
+  });
+
+  it("stores the payload in KV when the message overflows", async () => {
+    const { kv, store } = createMockKV();
+    const { queue, sent } = createMockQueue();
+    const env = createEnv(kv, queue);
+    const large = "x".repeat(150_000);
+    await enqueueWebhook(env, message({ payload: { data: large } }));
+    expect(sent).toHaveLength(1);
+    expect(sent[0].payload).toBeUndefined();
+    expect(sent[0].payloadRef).toBe("queue:payload:github:global:d1");
+    const stored = store.get("queue:payload:github:global:d1");
+    expect(stored).toBeDefined();
+    expect(JSON.parse(stored!).data).toBe(large);
+  });
+
+  it("throws when the QUEUE binding is missing", async () => {
+    const { kv } = createMockKV();
+    await expect(enqueueWebhook(createEnv(kv), message())).rejects.toThrow(
+      "QUEUE binding is not configured",
+    );
+  });
+});
+
+describe("resolvePayload / discardPayload", () => {
+  it("reads the payload from the message", async () => {
+    const { kv } = createMockKV();
+    const env = createEnv(kv);
+    expect(await resolvePayload(env, message())).toEqual({ ref: "refs/heads/main" });
+  });
+
+  it("reads the payload from KV when a payloadRef is set", async () => {
+    const { kv, store } = createMockKV();
+    store.set("queue:payload:github:global:d1", JSON.stringify({ a: 1 }));
+    const env = createEnv(kv);
+    expect(
+      await resolvePayload(
+        env,
+        message({ payload: undefined, payloadRef: "queue:payload:github:global:d1" }),
+      ),
+    ).toEqual({ a: 1 });
+  });
+
+  it("returns an empty object when the payloadRef is missing", async () => {
+    const { kv } = createMockKV();
+    expect(
+      await resolvePayload(
+        createEnv(kv),
+        message({ payload: undefined, payloadRef: "queue:payload:missing" }),
+      ),
+    ).toEqual({});
+  });
+
+  it("deletes the payloadRef on discard", async () => {
+    const { kv, store } = createMockKV();
+    store.set("queue:payload:github:global:d1", "{}");
+    await discardPayload(createEnv(kv), message({ payloadRef: "queue:payload:github:global:d1" }));
+    expect(store.has("queue:payload:github:global:d1")).toBe(false);
+  });
+});

tests/queue-consumer.test.ts

@@ -0,0 +1,139 @@
+import { describe, it, expect, beforeEach, mock } from "bun:test";
+import { handleQueueBatch } from "../server/lib/queue/consumer";
+import { invalidateConfigCache } from "../server/lib/config";
+import type { Env } from "../server/lib/types";
+import type { DeliveryMessage, DispatchSummary } from "../server/lib/queue/delivery";
+
+let summary: DispatchSummary;
+let dispatchCalls: number;
+
+mock.module("../server/lib/core/dispatch", () => ({
+  dispatchEvent: async (): Promise<DispatchSummary> => {
+    dispatchCalls += 1;
+    return summary;
+  },
+}));
+
+function createMockKV(): { kv: KVNamespace; store: Map<string, string> } {
+  const store = new Map<string, string>();
+  const kv = {
+    get: async (key: string) => (store.has(key) ? store.get(key)! : null),
+    put: async (key: string, value: string) => {
+      store.set(key, value);
+    },
+    delete: async (key: string) => {
+      store.delete(key);
+    },
+    list: async () => ({ keys: [], list_complete: true, cacheStatus: null }),
+  } as unknown as KVNamespace;
+  return { kv, store };
+}
+
+function createEnv(kv: KVNamespace): Env {
+  return {
+    GITHUB_WEBHOOK_SECRET: "secret",
+    KV: kv,
+    DB: {} as D1Database,
+  };
+}
+
+interface FakeMessage {
+  body: DeliveryMessage;
+  attempts: number;
+  acked: boolean;
+  retried: { delaySeconds?: number } | null;
+  ack: () => void;
+  retry: (opts?: { delaySeconds?: number }) => void;
+}
+
+function makeMessage(body: DeliveryMessage, attempts = 1): FakeMessage {
+  return {
+    body,
+    attempts,
+    acked: false,
+    retried: null,
+    ack(): void {
+      this.acked = true;
+    },
+    retry(opts): void {
+      this.retried = opts ?? {};
+    },
+  };
+}
+
+function makeBatch(queue: string, messages: FakeMessage[]): MessageBatch<DeliveryMessage> {
+  return {
+    queue,
+    messages: messages as never,
+    ackAll: () => {},
+    retryAll: () => {},
+  } as unknown as MessageBatch<DeliveryMessage>;
+}
+
+function body(overrides: Partial<DeliveryMessage> = {}): DeliveryMessage {
+  return {
+    deliveryId: "d1",
+    provider: "github",
+    event: "push",
+    payload: { ref: "refs/heads/main" },
+    receivedAt: Date.now(),
+    ...overrides,
+  };
+}
+
+const STATE_KEY = "delivery-state:github:global:d1";
+
+describe("handleQueueBatch", () => {
+  beforeEach(() => {
+    invalidateConfigCache();
+    summary = { attempts: 0, failures: [] };
+    dispatchCalls = 0;
+  });
+
+  it("marks DLQ messages dead and acks them", async () => {
+    const { kv, store } = createMockKV();
+    const msg = makeMessage(body());
+    await handleQueueBatch(makeBatch("webhooker-delivery-dlq", [msg]), createEnv(kv));
+    expect(msg.acked).toBe(true);
+    expect(JSON.parse(store.get(STATE_KEY)!)).toMatchObject({ status: "dead" });
+  });
+
+  it("acks and marks delivered on success", async () => {
+    const { kv, store } = createMockKV();
+    summary = { attempts: 1, failures: [] };
+    const msg = makeMessage(body());
+    await handleQueueBatch(makeBatch("webhooker-delivery", [msg]), createEnv(kv));
+    expect(msg.acked).toBe(true);
+    expect(msg.retried).toBeNull();
+    expect(JSON.parse(store.get(STATE_KEY)!)).toMatchObject({ status: "delivered" });
+  });
+
+  it("retries with backoff on a retryable failure", async () => {
+    const { kv, store } = createMockKV();
+    summary = { attempts: 2, failures: [{ target: "c1", errorCode: "DISCORD_5XX" }] };
+    const msg = makeMessage(body(), 1);
+    await handleQueueBatch(makeBatch("webhooker-delivery", [msg]), createEnv(kv));
+    expect(msg.acked).toBe(false);
+    expect(msg.retried).toEqual({ delaySeconds: 5 });
+    expect(JSON.parse(store.get(STATE_KEY)!)).toMatchObject({ status: "retrying" });
+  });
+
+  it("acks and marks failed on a permanent failure", async () => {
+    const { kv, store } = createMockKV();
+    summary = { attempts: 2, failures: [{ target: "c1", errorCode: "DISCORD_ERROR" }] };
+    const msg = makeMessage(body());
+    await handleQueueBatch(makeBatch("webhooker-delivery", [msg]), createEnv(kv));
+    expect(msg.acked).toBe(true);
+    expect(msg.retried).toBeNull();
+    expect(JSON.parse(store.get(STATE_KEY)!)).toMatchObject({ status: "failed" });
+  });
+
+  it("skips already-delivered deliveries without re-dispatching", async () => {
+    const { kv, store } = createMockKV();
+    store.set(STATE_KEY, JSON.stringify({ status: "delivered", at: Date.now() }));
+    const msg = makeMessage(body());
+    await handleQueueBatch(makeBatch("webhooker-delivery", [msg]), createEnv(kv));
+    expect(msg.acked).toBe(true);
+    expect(dispatchCalls).toBe(0);
+  });
+});

wrangler.jsonc

@@ -29,4 +29,27 @@
 			"database_id": "214a0104-3235-47c0-b7bf-ddda95f3c8ac"
 		},
 	],
+	"queues": {
+		"producers": [
+			{
+				"binding": "QUEUE",
+				"queue": "webhooker-delivery"
+			}
+		],
+		"consumers": [
+			{
+				"queue": "webhooker-delivery",
+				"max_batch_size": 10,
+				"max_batch_timeout": 30,
+				"max_retries": 5,
+				"dead_letter_queue": "webhooker-delivery-dlq"
+			},
+			{
+				"queue": "webhooker-delivery-dlq",
+				"max_batch_size": 10,
+				"max_batch_timeout": 30,
+				"max_retries": 1
+			}
+		]
+	},
 }