> ## Documentation Index
> Fetch the complete documentation index at: https://docs.oddspapi.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Go-Live Checklist - From Trial Key to Production

> A production readiness checklist for an OddsPapi integration: state handling, reconnect and resume, staleOdds, rate limits, monitoring, and what to send support.

A trial proves the data. Production needs the integration to hold when a connection drops, a bookmaker goes quiet, or we ship a release. This page is the list to work through before you switch on real volume. Every item links to the page with the detail.

***

## 1. State: snapshot plus stream

* [ ] Connect the WebSocket first, then fetch the REST snapshot and merge it in. Per key, the newer `changedAt` wins.
* [ ] Store odds keyed by `oddsId`. The key is the same on REST, WebSocket, history and settlement, so one store serves all four.
* [ ] Log in with `sportIds`, `tournamentIds` and `bookmakers` filters for what you trade. Smaller streams are faster to process and less likely to hit backpressure.
* [ ] Pick a `receiveType`. `zstd` cuts `odds` traffic to roughly a fifth or sixth of JSON with a one-line decode.

Detail: [Core concepts](https://docs.oddspapi.io/api-reference/concepts.md), [Login and filters](https://docs.oddspapi.io/websocket/auth.md), [Compression](https://docs.oddspapi.io/websocket/compression.md).

***

## 2. Reconnect and resume

* [ ] Persist `serverEpoch` and the last processed `entryId` per channel.
* [ ] On reconnect, log in with `serverEpoch` and `lastSeenId`, and wait for `resume_complete`.
* [ ] Handle `snapshot_required`: re-fetch the REST snapshot for the listed channels, clear their cursors, keep consuming.
* [ ] Act on the `reconnect` control frame as soon as it arrives. We send it before every release, so do not wait for the socket to close.
* [ ] Reconnect on close codes `1006` and `1011` with a short backoff. Treat `4002` (backpressure) and `4003` (too many connections) as problems to fix, not to retry blindly.

Before go-live, run both drills once: a disconnect of a few seconds that resumes without a gap, and one longer than `resumeWindowMs` that ends in a snapshot. Time the second one. It is your worst-case recovery.

Detail: [Resume & Replay](https://docs.oddspapi.io/websocket/resume-replay.md), [Troubleshooting](https://docs.oddspapi.io/websocket/troubleshooting.md).

***

## 3. Bookmaker staleness

* [ ] Subscribe to the `bookmakers` channel alongside `odds`.
* [ ] Keep `staleOdds` per `fixtureId:bookmaker` pair and check it before you act on a price.
* [ ] When it is `true`, pause that bookmaker on that fixture, or fall back to another book. Resume when it returns to `false`.
* [ ] Keep it apart from `suspended`, `active` and `marketActive`. Those are the bookmaker's own stops. `staleOdds` is about our view of the bookmaker.
* [ ] Expect a book-wide incident as one message per fixture, not as a single message.

Detail: [Bookmaker staleness](https://docs.oddspapi.io/api-reference/reliability.md#bookmaker-staleness-staleodds).

***

## 4. Rate limits and connections

* [ ] Odds endpoints allow 10 requests per second per `apiKey`. All other endpoints allow 200 per minute.
* [ ] Read `X-RateLimit-Remaining` and back off on `429` for the time in `Retry-After`.
* [ ] Use REST for snapshots and backfills, and the stream for changes. Polling odds in a loop is the usual cause of rate-limit trouble.
* [ ] Stay within 8 concurrent WebSocket connections per `apiKey` group. Ask us before you design for more.

Detail: [Rate limits](https://docs.oddspapi.io/api-reference/rate-limits.md), [WebSocket overview](https://docs.oddspapi.io/websocket/overview.md).

***

## 5. Monitoring

* [ ] Track the delivery hop: your receipt time minus the message `ts`. Run NTP, or sub-100 ms readings mean little.
* [ ] Track observation delay per bookmaker: `changedAt − bookmakerChangedAt`, and compare it to `maxDelayLiveInSec` and `maxDelayPregameInSec` from `GET /bookmakers`.
* [ ] Alert on reconnect count, `snapshot_required` count, and the share of fixtures flagged `staleOdds` per bookmaker.
* [ ] Subscribe to the [status page](https://oddspapi-v5.instatus.com) for incidents and maintenance.
* [ ] Deploy close to us if latency matters: central Germany or US East.

Detail: [Reliability & Operations](https://docs.oddspapi.io/api-reference/reliability.md), [Server location](https://docs.oddspapi.io/api-reference/concepts.md#server-location-matters), [Latency methodology](https://docs.oddspapi.io/guides/latency-methodology.md).

***

## 6. Changes and support

* [ ] Follow the [changelog](https://docs.oddspapi.io/changelog.md). Entries tagged **Breaking** need action on your side.
* [ ] Ignore unknown fields in responses. New fields can appear in any release.
* [ ] Send the key in the `X-API-Key` header rather than the query string, so it stays out of URLs and logs.
* [ ] When you report an issue to [contact@55-tech.com](mailto:contact@55-tech.com), include the affected channels or endpoints, UTC timestamps, and your `serverEpoch` and `entryId` values.

***

## Ready when

* A forced disconnect resumes with no gap, and a long one recovers through a snapshot in a time you have measured.
* A `staleOdds` flag stops trading on that bookmaker and fixture without anyone stepping in.
* A `429` slows your client down instead of breaking it.
* Your own dashboards show delivery hop, observation delay and stale share per bookmaker.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.