This application stores short-term data for use in Prebid Server and Prebid.js, primarily targeting video, native, and AMP formats.

First install Go version 1.18 or newer.

Note that prebid-cache uses Go modules. We officially support the most recent two major versions of the Go runtime. However, if you'd like to use a version <1.13 and are inside GOPATH GO111MODULE needs to be set to GO111MODULE=on.

Download and prepare Prebid Cache:

cd YOUR_DIRECTORY
git clone https://github.com/prebid/prebid-cache src/github.com/prebid/prebid-cache
cd src/github.com/prebid/prebid-cache

Run the automated tests:

./validate.sh

Or just run the server locally:

go build .
./prebid-cache

Adds one or more values to the cache. Values can be given as either JSON or XML. A sample request is below.

{
  "puts": [
    {
      "type": "xml",
      "ttlseconds": 60,
      "value": "<tag>Your XML content goes here.</tag>"
    },
    {
      "type": "json",
      "ttlseconds": 300,
      "value": [1, true, "JSON value of any type can go here."]
    }
  ]
}

If any of the puts are invalid, then it responds with a 400 and none of the values will be retrievable. Assuming that all of the values are well-formed, then the server will respond with IDs which can be used to fetch the values later.

Note: ttlseconds is optional, and will only be honored on a best effort basis. Callers should never assume that the data will stay in the cache for that long.

{
  "responses": [
    {"uuid": "279971e4-70f0-4b18-bd65-5c6e7aa75d40"},
    {"uuid": "147c9934-894b-4c1f-9a32-e7bb9cd15376"}
  ]
}

If ttlseconds is included, its value must be non-negative and no larger than the request_limits.max_ttl_seconds configuration, which defaults to 3600 seconds. The key parameter will be ignored unless the boolean configuration flag request_limits.allow_setting_keys is set to true. The following is a sample config.yaml configuration file that sets the ttl to 100 seconds and allows Prebid Cache to set custom keys:

request_limits:
  max_ttl_seconds: 100
  allow_setting_keys: true

in the example below, A Prebid Cache instance that supports key by setting allow_setting_keys to true, includes key fields in some of the "puts" array elements:

{
  "puts": [
    {
      "type": "xml",
      "ttlseconds": 60,
      "value": "<tag>Your XML content goes here.</tag>",
      "key": "CustomKeyValueHere"
    },
    {
      "type": "json",
      "ttlseconds": 300,
      "value": [1, true, "JSON value of any type can go here."]
    },
    {
      "type": "json",
      "value": "{\"someJsonField\":\"some JSON string value.\"}",
      "key": "AnotherCustomKeyValue"
    }
  ]
}

This would result in a response like this:

{
  "responses": [
    {"uuid": "CustomKeyValueHere"},
    {"uuid": "147c9934-894b-4c1f-9a32-e7bb9cd15376"},
    {"uuid": "AnotherCustomKeyValue"}
  ]
}

Prebid Cache will use the custom keys for those elements that include them and will create system-generated uuids for the ones that don't. Note that if configuration flag allow_setting_keys is set to false or simply not set inside the config.yaml file, Prebid Cache would generate random uuids for all elements and not use the custom keys "CustomKeyValueHere" nor "AnotherCustomKeyValue" at all.

Prebid Cache does not allow overwritting any value, for either autogenerated or custom keys. If an entry already exists for a given key, it will not be overwitten, and an empty string will be returned as the uuid value of that entry. Suppose we wanted to overwrite the entries under "CustomKeyValueHere" and the system-generated key "147c9934-894b-4c1f-9a32-e7bb9cd15376".

{
  "puts": [
    {
      "type": "xml",
      "ttlseconds": 60,
      "value": "<tag>Some other XML content under a key already in use.</tag>",
      "key": "CustomKeyValueHere"
    },
    {
      "type": "json",
      "ttlseconds": 300,
      "value": [2, false, "Trying to store other JSON under a key already in use."],
      "key": "147c9934-894b-4c1f-9a32-e7bb9cd15376"
    },
    {
      "type": "xml",
      "ttlseconds": 30,
      "value": "<tag>Different XML content will be stored under a system-generated key.</tag>",
    }
  ]
}

Then we get a response with empty uuid fields for those "puts" elements that tried to overwrite entries:

{
  "responses": [
    {"uuid": ""},
    {"uuid": ""},
    {"uuid": "efc6ca1d-3409-4b8b-96e5-aec508a57639"}
  ]
}

Those items in the cache did not get overwritten as a subsequent GET call would assert:

$ curl http://someprebidcachehost.com/cache\?uuid\=CustomKeyValueHere

HTTP/1.1 200 OK
Content-Type: application/xml

<tag>Your XML content goes here.</tag>

This is to prevent bad actors from trying to overwrite legitimate caches with malicious content, or a poorly coded app overwriting its own cache with new values, generating uncertainty of what is actually stored under a particular key. Note that cases like these are the only time where a subset of caches would not get stored. Under any other scenario, we expect the entire request to fail and no elements will be stored.

Trying to overwrite the value under an existing key is also the only instance where an unsuccessful Put is not considered an error. As such, Prebid Cache will not respond with an error message or return an error code on these particular instances.

Retrieves a single value from the cache. If the id isn't recognized, then it will return an HTTP 404. The following are sample requests and responses based on the POST call examples above.

GET /cache?uuid=279971e4-70f0-4b18-bd65-5c6e7aa75d40

HTTP/1.1 200 OK
Content-Type: application/xml

<tag>Your XML content goes here.</tag>

GET /cache?uuid=147c9934-894b-4c1f-9a32-e7bb9cd15376

HTTP/1.1 200 OK
Content-Type: application/json

[1, true, "JSON value of any type can go here."]

This section does not describe permanent API contracts; it just describes limitations on the current implementation.

• This application does not validate XML. If users POST malformed XML, they'll GET a bad response too.
• The host company can set a max length on payload size limits in the application config. This limit will vary from vendor to vendor.

Prebid Cache requires a backend data store which enforces TTL expiration. The following storage options are supported: Aerospike, Cassandra, Memcache, and Redis. You're welcomed to contribute a new backend adapter if needed.

There is also an option (enabled by default) for a basic in-memory data store intended only for development. This backend does not support TTL expiration and is not built for production use.

To configure, select the storage service for your Prebid Cache server by setting the backend.type property in the config.yaml file:

backend:
  type: "aerospike"

Prebid Cache makes use of an Aerospike Go client that requires Aerospike server version 4.9 and will not work properly with older versions. Full documentation of the Aerospike Go client can be found here.

Prebid Cache makes use of a Cassandra client that supports latest 3 major releases of Cassandra (2.1.x, 2.2.x, and 3.x.x). Full documentation of the Cassandra Go client can be found here.

Prebid Cache makes use of a Redis Go client compatible with Redis 6. Full documentation of the Redis Go client Prebid Cache uses can be found here.

Sample configuration file config/configtest/sample_full_config.yaml shown below:

port: 9500
admin_port: 2525
index_response: "Any index response"
log:
  level: "info"
rate_limiter:
  enabled: false
  num_requests: 150
request_limits:
  max_size_bytes: 10240
  max_num_values: 10
  max_ttl_seconds: 5000
  allow_setting_keys: true
backend:
  type: "memory"
  aerospike:
    default_ttl_seconds: 3600
    host: "aerospike.prebid.com"
    hosts: ["aerospike2.prebid.com", "aerospike3.prebid.com"]
    port: 3000
    namespace: "whatever"
    user: "foo"
    password: "bar"
  cassandra:
    hosts: "127.0.0.1"
    keyspace: "prebid"
  memcache:
    hosts: ["10.0.0.1:11211","127.0.0.1"]
  redis:
    host: "127.0.0.1"
    port: 6379
    password: "redis-password"
    db: 1
    expiration: 1
    tls:
      enabled: false
      insecure_skip_verify: false
compression:
  type: "snappy"
metrics:
  type: "none"
  influx:
    host: "metrics-host"
    database: "metrics-database"
    username: "metrics-username"
    password: "metrics-password"
    enabled: true
  prometheus:
    port: 8080
    namespace: "prebid"
    subsystem: "cache"
    timeout_ms: 100
    enabled: true
routes:
  allow_public_write: true

Golang 1.18.x or newer.

./validate.sh runs the unit tests and reformats your code with gofmt. ./validate.sh --nofmt runs the unit tests, but will not reformat your code.

Run prebid-cache locally with:

go build .
./prebid-cache

The service will respond to requests on localhost:2424, and the admin data will be available on localhost:2525. The port numbers can be configured to different values in your production deployment.

Configuration is handled by Viper. The easiest way to set config during development is by editing the config.yaml file. You can also set the config through environment variables. For instance:

export PBC_COMPRESSION_TYPE="none"

Prebid Cache's rate limiting feature, that has the downside of considerable memory consumption, is enabled by default for a maximum of 100 requests per second. From the config.yaml file, use the rate_limiter.enabled and rate_limiter.num_requests options to either disable the rate limiter or modify its request capacity. For instance adding the following in the config.yaml file:

rate_limiter:
  enabled: false

disables the rate limiter. We could also disable it by setting the following environment variable:

export PBC_RATE_LIMITER_ENABLED="false"

In contrast, we could keep the rate limiter running and set its maximum number of requests to a value other than 100. For instance, to set them to 150, we could modify the num_requests field inside config.yaml:

rate_limiter:
  num_requests: 150

Or via the following environment variable:

export PBC_RATE_LIMITER_NUM_REQUESTS=150

Prebid Cache works in Docker out of the box. It comes with a Dockerfile that creates a container, downloads all dependencies, and instantly installs a working image for us to run Prebid Cache right away. Using the docker build command we specify an image name and the location of the folder where we cloned or downloaded Prebid Cache to create an image ready to run. If we cloned Prebid Cache in ~/go/src/github.com/prebid/prebid-cache, then we could use the command that follows to create the image prebid-cache.

docker build -t prebid-cache ~/go/src/github.com/prebid/prebid-cache

We can run Prebid Cache using the newly created image:

docker run -p 8000:8000 -t prebid-cache

pprof stats can be accessed from a running app on the admin port localhost:2525.