fireworks

Documentation

Overview

Package fireworks wraps Fireworks AI's OpenAI-compatible API. Fireworks hosts open-weight models on its FireAttention serving stack and ships latency-optimized custom variants of popular models.

Fireworks-specific knobs reachable through the namespaced OpenAI request extension:

• "context_length_exceeded_behavior" controls truncation policy.
• "prompt_cache_max_len" enables Fireworks' prompt-cache layer.
• The /chat/completions endpoint accepts "response_format" with "type":"grammar" to constrain output via GBNF (alongside the standard "json_schema").

See https://docs.fireworks.ai/ for the full API reference.

Index

• Constants
• type Chat
  • func NewChat(config ChatConfig) (*Chat, error)
• type ChatConfig
  • func (c ChatConfig) Validate() error

Constants

const (
	OpenAIRequestExtensionKey     = "fireworks/openai_request"
	OpenAIResponseExtensionKey    = "fireworks/openai_response"
	OpenAIStreamChunkExtensionKey = "fireworks/openai_stream_chunk"
)

const (
	Provider = "Fireworks"

	BaseURL = "https://api.fireworks.ai/inference/v1"
)

const (
	ModelGPTOSS20B = "accounts/fireworks/models/gpt-oss-20b"
	ModelKimiK26   = "accounts/fireworks/models/kimi-k2p6"
	ModelGLM52     = "accounts/fireworks/models/glm-5p2"
)

Current serverless model ids. See https://fireworks.ai/models for the live catalog — Fireworks prefixes model ids with "accounts/fireworks/models/".

Types

type Chat

type Chat = openai.ChatCompletions

Chat implements Fireworks' chat endpoint.

func NewChat

func NewChat(config ChatConfig) (*Chat, error)

type ChatConfig

type ChatConfig struct {
	APIKey         string
	DefaultOptions corechat.Options
	BaseURL        string
	HTTPClient     *http.Client
}

func (ChatConfig) Validate

func (c ChatConfig) Validate() error