Inngest Agent Example — Utah

Universally Triggered Agent Harness

A durable AI agent built with Inngest and pi-ai. No framework. Just a think/act/observe loop — Inngest provides durability, retries, and observability, while pi-ai provides a unified LLM interface across providers.

Simple TypeScript that gives you:

• 🔄 Durable agent loop — every LLM call and tool execution is an Inngest step
• 🔁 Automatic retries — LLM API timeouts are handled by Inngest, not your code
• 🔒 Singleton concurrency — one conversation at a time per chat, no race conditions
• ⚡ Cancel on new message — user sends again? Current run cancels, new one starts
• 📡 Multi-channel — Slack, Telegram, and more via a simple channel interface
• 🏠 Local development — runs on your machine via connect(), no server needed

Architecture

Channel (e.g. Telegram) → Inngest Cloud (webhook + transform) → WebSocket → Local Worker → LLM (Anthropic/OpenAI/Google) → Reply Event → Channel API

The worker connects to Inngest Cloud via WebSocket. No public endpoint. No ngrok. No VPS. Messages flow through Inngest as events, and the agent processes them locally with full filesystem access.

Sidecar: Orchestration-Aware Agent Loops

The core agent handles conversations. But conversations are ephemeral — the agent forgets, the process restarts, the context window rolls over. The sidecar is what makes Utah's output durable.

A separate process (utah-sidecar) dynamically loads Inngest functions from disk, connects to Inngest Cloud via WebSocket, and runs them independently. The agent can write a new .ts file to the functions directory and the sidecar hot-reloads it automatically — no restart, no deploy, no human intervention.

The key idea: the agent doesn't just run inside loops — it authors new loops and deploys them to the orchestration engine. Each deployed function is a durable skill that runs on its own schedule, with its own retry logic, completely independent of whether the agent is in a conversation.

┌─────────────────────┐       ┌───────────────────────┐
│   Core Agent        │       │   Sidecar             │
│   app: "ai-agent"   │       │   app: "utah-sidecar" │
│                     │       │                       │
│   handleMessage     │       │   workspace/functions/│
│   sendReply         │       │     *.ts (dynamic)    │
│   subAgent          │       │   + heartbeat (auto)  │
│   etc.              │       │   + file watcher      │
└────────┬────────────┘       └────────┬──────────────┘
         │                             │
         │    connect() via WebSocket  │
         └──────────┬──────────────────┘
                    │
           ┌────────▼────────┐
           │  Inngest Cloud  │
           │  events, crons, │
           │  retries, state │
           └─────────────────┘

Both processes connect to Inngest independently. They share nothing except the event bus.

How it works

1. The sidecar reads workspace/functions/*.ts, dynamically imports each file, and registers the exported Inngest functions
2. A fs.watch() monitors the directory — on any change, a 2-second debounce fires, the existing WebSocket closes, functions are re-imported with cache-busting, and a new connection opens
3. A heartbeat function is auto-injected (runs every 30 minutes) so the sidecar always has at least one registered function
4. No process restart needed — the agent writes a file, the sidecar picks it up

The agent writes its own skills

The agent can author new Inngest functions — cron jobs, event handlers, multi-step workflows — by writing a .ts file to workspace/functions/. The sidecar deploys them automatically.

Some example functions that the main agent might write to extend itself: morning-triage, daily-meeting-digest, nightly-workspace-commit, weekly-review. You can also create "loops" with review functions that use LLMs to review and iterate on functions, for example: inbox-triage-review, cold-email-learner.

Each function is durable — retried on failure, observable in the Inngest dashboard, independently scheduled. Skills compound. The agent builds infrastructure for itself.

Agent skills as persistent knowledge

Agent skills are markdown reference docs (with name/description frontmatter) that appear in the agent's system prompt. The agent can create its own skills to persist knowledge across conversations.

This creates a self-referential system:

• The Inngest Functions skill teaches the agent how to write sidecar functions (templates, triggers, step API, best practices)
• The Sidecar Management skill teaches file operations for managing the functions directory
• When the agent learns a new pattern, it can write a new skill and a new function — persisting both the knowledge and the automation

The agent is ephemeral. Its output is durable.

Communication

Sidecar functions talk back to the main agent by sending agent.message.received events:

await step.sendEvent("alert-agent", {
  name: "agent.message.received",
  data: {
    channel: "system",
    sessionKey: "system-alerts",
    message: "Alert: something needs attention",
  },
});

This means a cron job can monitor something, detect a problem, and start a conversation with the agent — which can then use its tools to investigate and respond. The loops feed each other.

Prerequisites

• Node.js 23+ (uses native TypeScript strip-types)
• LLM API key (e.g. Anthropic API key (console.anthropic.com))
• Inngest account (app.inngest.com)
• At least one channel configured (see Channels below)

Setup

1. Create an Inngest Account

1. Sign up at app.inngest.com
2. Go to Settings → Keys and copy your:
   • Event Key (for sending events)
   • Signing Key (for authenticating your worker)

2. Configure and Run

git clone https://github.com/inngest/utah
cd utah
npm install # or pnpm
cp .env.example .env

Edit .env with your keys:

ANTHROPIC_API_KEY=sk-ant-...
INNGEST_EVENT_KEY=...
INNGEST_SIGNING_KEY=signkey-prod-...

Then add the environment variables for your channel(s) — see setup guides below.

Start the worker:

# Production mode (connects to Inngest Cloud via WebSocket)
npm start

# Development mode (uses local Inngest dev server)
npx inngest-cli@latest dev &
npm run dev

On startup, the worker automatically sets up webhooks and transforms for each configured channel.

Channels

The agent supports multiple messaging channels. Each channel has its own setup guide:

• Telegram — Fully automated setup. Just add your bot token and run.
• Slack — Requires creating a Slack app and configuring Event Subscriptions.

Project Structure

src/
├── worker.ts                  # Entry point — connect() or serve()
├── client.ts                  # Inngest client
├── config.ts                  # Configuration from env vars
├── agent-loop.ts              # Core think → act → observe cycle
├── setup.ts                   # Channel setup orchestration
├── lib/
│   ├── llm.ts                 # pi-ai wrapper (multi-provider: Anthropic, OpenAI, Google)
│   ├── tools.ts               # Tool definitions (TypeBox schemas) + execution
│   ├── context.ts             # System prompt builder with workspace file injection
│   ├── session.ts             # JSONL session persistence
│   ├── memory.ts              # File-based memory system (daily logs + distillation)
│   └── compaction.ts          # LLM-powered conversation summarization
├── functions/
│   ├── message.ts             # Main agent function (singleton + cancelOn)
│   ├── send-reply.ts          # Channel-agnostic reply dispatch
│   ├── acknowledge-message.ts # Message acknowledgment (typing indicator, etc.)
│   ├── heartbeat.ts           # Cron-based memory maintenance
│   └── failure-handler.ts     # Global error handler with notifications
└── channels/
    ├── types.ts               # ChannelHandler interface
    ├── index.ts               # Channel registry
    ├── setup-helpers.ts       # Inngest REST API helpers for webhook setup
    └── <channel-name>/        # A channel implementation (see README for setup)
        ├── handler.ts         # ChannelHandler implementation
        ├── api.ts             # API client
        ├── setup.ts           # Webhook setup automation
        ├── transform.ts       # Webhook transform
        └── format.ts          # Formatting for channel messages
workspace/                       # Agent workspace (persisted across runs)
├── SOUL.md                    # Agent personality and behavioral guidelines
├── USER.md                    # User information
├── MEMORY.md                  # Long-term memory (agent-writable)
├── memory/                    # Daily logs (YYYY-MM-DD.md, auto-managed)
└── sessions/                  # JSONL conversation files (gitignored)

How It Works

The Agent Loop

The core is a while loop where each iteration is an Inngest step:

1. Think — step.run("think") calls the LLM via pi-ai's complete()
2. Act — if the LLM wants tools, each tool runs as step.run("tool-read")
3. Observe — tool results are fed back into the conversation
4. Repeat — until the LLM responds with text (no tools) or max iterations

Inngest auto-indexes duplicate step IDs in loops (think:0, think:1, etc.), so you don't need to track iteration numbers in step names.

Event-Driven Composition

One incoming message triggers multiple independent functions:

Workspace Context Injection

The agent reads markdown files from the workspace directory and injects them into the system prompt:

Edit these files to customize your agent's personality and knowledge. The agent can also update MEMORY.md using the write tool to remember things across conversations.

Memory System

The agent has a two-tier memory system:

• Daily logs (workspace/memory/YYYY-MM-DD.md) — append-only notes written via the remember tool during conversations
• Long-term memory (workspace/MEMORY.md) — curated summary distilled from daily logs by the heartbeat function

The agent-heartbeat function runs on a cron schedule (default: every 30 minutes). It checks if daily logs have accumulated enough content, then uses the LLM to distill them into MEMORY.md. Old daily logs are pruned after a configurable retention period (default: 30 days).

Conversation Compaction

Long conversations get summarized automatically so the agent doesn't lose context or hit token limits:

• Token estimation: Uses a chars/4 heuristic to estimate conversation size
• Threshold: Compaction triggers when estimated tokens exceed 80% of the configured max (150K)
• LLM summarization: Old messages are summarized into a structured checkpoint (goals, progress, decisions, next steps)
• Recent messages preserved: The most recent ~20K tokens of conversation are kept verbatim
• Persisted: The compacted session replaces the JSONL file, so it survives restarts

Compaction runs as an Inngest step (step.run("compact")), so it's durable and retryable.

Context Pruning

Long tool results bloat the conversation context and cause the LLM to lose focus. The agent uses two-tier pruning:

• Soft trim: Tool results over 4K chars get head+tail trimmed (first 1,500 + last 1,500 chars)
• Hard clear: When total old tool content exceeds 50K chars, old results are replaced entirely
• Budget warnings: System messages are injected when iterations are running low

Adding New Channels

The agent is channel-agnostic. Each channel implements a ChannelHandler interface (src/channels/types.ts) with methods for sending replies, acknowledging messages, and setup. Each channel directory follows the same structure:

src/channels/<name>/
├── handler.ts      # ChannelHandler implementation (sendReply, acknowledge)
├── api.ts          # API client for the channel's platform
├── setup.ts        # Webhook setup automation
├── transform.ts    # Plain JS transform for Inngest webhook
└── format.ts       # Markdown → channel-specific format conversion

To add Discord, WhatsApp, or any other channel:

1. Create a new directory under src/channels/ following the structure above
2. Implement the ChannelHandler interface in handler.ts
3. Write a webhook transform that converts the channel's payload to agent.message.received
4. Register the channel in src/channels/index.ts

The agent loop, reply dispatch, and acknowledgment functions are all channel-agnostic — no changes needed outside src/channels/.

Key Inngest Features Used

• connect() — WebSocket-based worker
• Singleton execution — one run per chat at a time
• Step retries — automatic retry on LLM API failures
• Event-driven functions — compose behavior from small focused functions
• Webhook transforms — convert external payloads to typed events
• Checkpointing — near-zero inter-step latency

Acknowledgments

This project uses pi-ai (@mariozechner/pi-ai) by Mario Zechner for its unified LLM interface and @mariozechner/pi-coding-agent for it's. standard tools. pi-ai provides a single complete() function that works across Anthropic, OpenAI, Google, and other providers — making it easy to swap models without changing any agent code. It's a great library.

License

Apache-2.0