For the complete documentation index, see llms.txt. Markdown versions of all docs pages are available by appending .md to any docs URL.

Route agentgateway LLM traffic to models available through Perplexity.

Configure Perplexity as an LLM provider in agentgateway. Perplexity serves search-augmented models through an OpenAI-compatible API.

You can connect to Perplexity with an AgentgatewayModel or an AgentgatewayBackend. The two approaches accept different request formats. Choose the approach that matches the requests that your clients send.

Before you begin

1. Install and set up an agentgateway proxy.
2. To use the AgentgatewayModel approach, allow the AgentgatewayModel route kind on the http listener of the agentgateway-proxy Gateway. For the steps, see Enable LLM serving on a listener.

Set up provider credentials

Store your Perplexity API key in a Secret. The model and backend examples both reference this Secret.

1. Get an API key from Perplexity.

2. Save the API key in an environment variable.

   export PERPLEXITY_API_KEY='<your-api-key>'

3. Create a Kubernetes Secret to store your API key.

   kubectl apply -f- <<EOF
   apiVersion: v1
   kind: Secret
   metadata:
     name: perplexity-secret
     namespace: agentgateway-system
   type: Opaque
   stringData:
     Authorization: $PERPLEXITY_API_KEY
   EOF

Responses requests with an AgentgatewayModel

The model attaches directly to the Gateway listener, so you do not create an AgentgatewayBackend or an HTTPRoute.

1. Create an AgentgatewayModel that uses the Perplexity preset and references the Secret.

   kubectl apply -f- <<EOF
   apiVersion: agentgateway.dev/v1alpha1
   kind: AgentgatewayModel
   metadata:
     name: perplexity
     namespace: agentgateway-system
   spec:
     parentRefs:
     - group: gateway.networking.k8s.io
       kind: Gateway
       name: agentgateway-proxy
       sectionName: http
     match:
       model: perplexity/sonar
     provider: Perplexity
     policies:
       auth:
         secretRef:
           name: perplexity-secret
   EOF

   Review the following table to understand this configuration. For more information, see the API docs.

   Setting	Description
   spec.match.model	The model name that clients send in requests. Perplexity’s Responses API uses model names with a vendor prefix, such as perplexity/sonar.
   spec.provider	The Perplexity preset. The preset supplies the default URL, https://api.perplexity.ai/v1, and the Responses request format.
   spec.policies.auth.secretRef	The Secret that holds your Perplexity API key.

2. Send a Responses request to the model. Chat Completions requests to this model fail.

   Cloud Provider LoadBalancerPort-forward for local testing

   curl "http://$INGRESS_GW_ADDRESS/v1/responses" \
     -H 'Content-Type: application/json' \
     -d '{"model": "perplexity/sonar", "input": "Explain retrieval-augmented generation in one sentence."}' | jq

   curl "localhost:8080/v1/responses" \
     -H 'Content-Type: application/json' \
     -d '{"model": "perplexity/sonar", "input": "Explain retrieval-augmented generation in one sentence."}' | jq

Chat Completions requests with an AgentgatewayBackend

Use this approach to select the backend with an HTTPRoute.

1. Create an AgentgatewayBackend that points the openai provider at Perplexity. Perplexity’s chat completions path is /chat/completions, without the /v1 prefix that most providers use. The Chat Completions API uses model names without the vendor prefix, such as sonar instead of the perplexity/sonar name that the Responses API uses.

   kubectl apply -f- <<EOF
   apiVersion: agentgateway.dev/v1alpha1
   kind: AgentgatewayBackend
   metadata:
     name: perplexity
     namespace: agentgateway-system
   spec:
     ai:
       provider:
         openai:
           model: sonar
         host: api.perplexity.ai
         port: 443
         path: /chat/completions
     policies:
       auth:
         secretRef:
           name: perplexity-secret
       tls:
         sni: api.perplexity.ai
   EOF

2. Create an HTTPRoute that routes requests on the /perplexity path to the AgentgatewayBackend.

   kubectl apply -f- <<EOF
   apiVersion: gateway.networking.k8s.io/v1
   kind: HTTPRoute
   metadata:
     name: perplexity
     namespace: agentgateway-system
   spec:
     parentRefs:
       - name: agentgateway-proxy
         namespace: agentgateway-system
     rules:
     - matches:
       - path:
           type: PathPrefix
           value: /perplexity
       backendRefs:
       - name: perplexity
         namespace: agentgateway-system
         group: agentgateway.dev
         kind: AgentgatewayBackend
   EOF

3. Send a Chat Completions request to the route.

   Cloud Provider LoadBalancerPort-forward for local testing

   curl "http://$INGRESS_GW_ADDRESS/perplexity" \
     -H 'Content-Type: application/json' \
     -d '{
       "model": "sonar",
       "messages": [{"role": "user", "content": "Explain retrieval-augmented generation in one sentence."}]
     }' | jq

   curl "localhost:8080/perplexity" \
     -H 'Content-Type: application/json' \
     -d '{
       "model": "sonar",
       "messages": [{"role": "user", "content": "Explain retrieval-augmented generation in one sentence."}]
     }' | jq

Next steps

To connect other providers that expose an OpenAI-compatible API, see OpenAI-compatible providers.

Cleanup

kubectl delete AgentgatewayModel perplexity -n agentgateway-system
kubectl delete HTTPRoute perplexity -n agentgateway-system
kubectl delete AgentgatewayBackend perplexity -n agentgateway-system
kubectl delete secret perplexity-secret -n agentgateway-system