Fireworks AI

Use Fireworks AI with Blazen — construct the FireworksProvider and call complete().

Fireworks AI is an OpenAI-compatible LLM provider. Construct it like any other Blazen provider — build a list of ChatMessages, call complete() (or stream()), and read back a typed ModelResponse.

At a glance

Provider id	fireworks
Base URL	https://api.fireworks.ai/inference/v1
Default model	accounts/fireworks/models/llama-v3p3-70b-instruct
API key env var	FIREWORKS_API_KEY
Auth	Authorization: Bearer <key>

Set FIREWORKS_API_KEY in the environment and Blazen reads it automatically, or pass the key explicitly when you construct the provider.

Capabilities

Capability	Supported
Streaming	Yes
Tool calling	Yes
Structured output	Yes
Vision	Yes
Model listing	Yes
Embeddings	Yes

Usage

Construct the provider and call complete(). The default model is accounts/fireworks/models/llama-v3p3-70b-instruct; override it with with_model / model when you need a different one.

use blazen_llm::{Model, ModelRequest, ChatMessage};
use blazen_provider_fireworks::FireworksProvider;

// Reads FIREWORKS_API_KEY from the environment, or pass the key to `new`.
let model = FireworksProvider::new(std::env::var("FIREWORKS_API_KEY")?);

let resp = model
    .complete(ModelRequest::new(vec![ChatMessage::user("Hello")]))
    .await?;
println!("{}", resp.content.unwrap_or_default());

from blazen import FireworksProvider, ProviderOptions, ChatMessage

# Omit the api_key to read FIREWORKS_API_KEY from the environment.
model = FireworksProvider(options=ProviderOptions(api_key="..."))

resp = await model.complete([ChatMessage.user("Hello")])
print(resp.content)

import { FireworksProvider, ChatMessage } from "blazen";

// Omit apiKey to read FIREWORKS_API_KEY from the environment.
const model = FireworksProvider.create({ apiKey: "..." });

const resp = await model.complete([ChatMessage.user("Hello")]);
console.log(resp.content);

Streaming

async for chunk in model.stream([ChatMessage.user("Count to five")]):
    print(chunk.delta, end="")

await model.stream([ChatMessage.user("Count to five")], (chunk) => {
  if (chunk.delta) process.stdout.write(chunk.delta);
});

See also

• Models & Providers — the canonical provider + ModelManager guide