Perplexity LLM Provider

The Perplexity provider connects Beluga AI to Perplexity’s search-augmented language models. Perplexity models combine LLM reasoning with real-time web search, making them well-suited for tasks requiring up-to-date information. Perplexity exposes an OpenAI-compatible API, so this provider supports all standard features including streaming.

Choose Perplexity when your application needs answers grounded in current web data without building a separate search pipeline. This is particularly useful for research assistants, news analysis, and any task where the model’s training data cutoff is a limitation.

Installation

go get github.com/lookatitude/beluga-ai/llm/providers/perplexity

Configuration

Field	Required	Default	Description
Model	Yes	—	Model ID (e.g. "sonar-pro")
APIKey	Yes	—	Perplexity API key (pplx-...)
BaseURL	No	https://api.perplexity.ai	Override API endpoint
Timeout	No	30s	Request timeout

Environment variables:

Variable	Maps to
PERPLEXITY_API_KEY	APIKey

Basic Usage

package main

import (
    "context"
    "fmt"
    "log"
    "os"

    "github.com/lookatitude/beluga-ai/config"
    "github.com/lookatitude/beluga-ai/llm"
    "github.com/lookatitude/beluga-ai/schema"
    _ "github.com/lookatitude/beluga-ai/llm/providers/perplexity"
)

func main() {
    model, err := llm.New("perplexity", config.ProviderConfig{
        Model:  "sonar-pro",
        APIKey: os.Getenv("PERPLEXITY_API_KEY"),
    })
    if err != nil {
        log.Fatal(err)
    }

    msgs := []schema.Message{
        schema.NewSystemMessage("You are a helpful research assistant."),
        schema.NewHumanMessage("What are the latest developments in Go 1.24?"),
    }

    resp, err := model.Generate(context.Background(), msgs)
    if err != nil {
        log.Fatal(err)
    }

    fmt.Println(resp.Text())
}

Streaming

for chunk, err := range model.Stream(context.Background(), msgs) {
    if err != nil {
        log.Fatal(err)
    }
    fmt.Print(chunk.Delta)
}
fmt.Println()

Advanced Features

Generation Options

resp, err := model.Generate(ctx, msgs,
    llm.WithTemperature(0.2),
    llm.WithMaxTokens(2048),
    llm.WithTopP(0.9),
)

Lower temperatures are often preferred for factual search queries.

Error Handling

resp, err := model.Generate(ctx, msgs)
if err != nil {
    log.Fatal(err)
}

Direct Construction

import "github.com/lookatitude/beluga-ai/llm/providers/perplexity"

model, err := perplexity.New(config.ProviderConfig{
    Model:  "sonar-pro",
    APIKey: os.Getenv("PERPLEXITY_API_KEY"),
})

Available Models

Model ID	Description
sonar-pro	Advanced search-augmented model
sonar	Standard search-augmented model

Refer to Perplexity’s documentation for the latest model list.