ncheta

your webhooks just got memory.

ncheta (igbo: memory / remembrance) is a self-hostable webhook monitoring tool. it intercepts incoming webhooks, stores the raw request before your handler runs, then forwards it. if your handler crashes — the data is safe. fix the bug. replay. done.

no more "stripe sent it once and we missed it."

how it works

[Stripe] ──POST──> [ncheta :7000] ──stores──> [SQLite]
                        │
                        ├── returns 200 to Stripe immediately
                        │
                        └── forwards to your handler in background
                                │
                                ├── handler returns 200? done.
                                └── handler returns 500? retry with backoff.

two ports. one binary.

• port 7000 — ingestion. webhook senders POST here.
• port 7001 — control API. list events, inspect payloads, trigger replays.

quickstart

rust binary

cargo build --release
./target/release/ncheta

node sdk

npm install ncheta

const { Ncheta } = require("ncheta");

const ncheta = new Ncheta();
await ncheta.start();

// express middleware — captures everything, blocks nothing
app.post("/webhooks/stripe",
  ncheta.watch({ endpointName: "stripe", targetUrl: "http://localhost:3000/webhooks/stripe" }),
  (req, res) => {
    // your handler. if this blows up, ncheta has the raw request.
    res.json({ ok: true });
  }
);

config

all env vars. all optional. sane defaults.

| var | default | what it does | |-----|---------|-------------| | NCHETA_DB_URL | sqlite://./ncheta.db | database connection | | NCHETA_INGESTION_PORT | 7000 | where webhooks land | | NCHETA_CONTROL_PORT | 7001 | REST API + dashboard | | NCHETA_MAX_BODY_SIZE | 524288 (512KB) | max request/response body | | NCHETA_MAX_RETRIES | 5 | retries before giving up | | NCHETA_EVENT_TTL_SECONDS | 604800 (7 days) | auto-cleanup after this |

api

ingestion (port 7000)

POST /in/{endpoint_name}  →  200 { "received": true, "event_id": "..." }

control (port 7001)

GET  /health              →  200 { "status": "ok" }
GET  /endpoints           →  list all endpoints
POST /endpoints           →  register a new endpoint
GET  /events              →  list events (filters: ?status=&endpoint_id=&limit=)
GET  /events/{id}         →  full event detail with headers + body
GET  /events/{id}/attempts → delivery attempt history
POST /events/{id}/replay  →  202 re-deliver the event

replay

# event failed? fix your handler, then:
curl -X POST http://localhost:7001/events/{id}/replay

# ncheta re-sends the exact same headers + body
# adds X-Ncheta-Replay: true so you know it's a replay

retry backoff

failed deliveries retry automatically:

| attempt | delay | |---------|-------| | 1 | 30 seconds | | 2 | 5 minutes | | 3 | 30 minutes | | 4+ | 2 hours |

after max_retries failures → event goes to dead status. replay manually whenever you're ready.

architecture

• rust binary — axum + tokio + sqlx
• sqlite with WAL mode — fast writes, no external db needed
• zero mutex — all mutable state lives in the db
• single reqwest client — connection pooling, no fd exhaustion
• store-before-ack — db write completes before 200 is returned to sender

license

MIT