Skip to content

CerebrasBackendAdapter

Defined in: backend/src/providers/cerebras.ts:122

Backend adapter for Cerebras Chat Completions API.

Features:

  • World’s fastest inference (969 tokens/s)
  • OpenAI-compatible API
  • Specialized hardware acceleration
  • Seed support for reproducibility
  • Timing information in responses
  • Competitive pricing starting at $0.10 per 1M tokens

new CerebrasBackendAdapter(config): CerebrasBackendAdapter

Defined in: backend/src/providers/cerebras.ts:127

BackendAdapterConfig

CerebrasBackendAdapter

readonly metadata: AdapterMetadata

Defined in: backend/src/providers/cerebras.ts:123

Adapter metadata for identification and capabilities.

BackendAdapter.metadata

estimateCost(request): Promise<number | null>

Defined in: backend/src/providers/cerebras.ts:485

Estimate cost.

IRChatRequest

Promise<number | null>

BackendAdapter.estimateCost


execute(request, signal?): Promise<IRChatResponse>

Defined in: backend/src/providers/cerebras.ts:270

Execute non-streaming request.

IRChatRequest

AbortSignal

Promise<IRChatResponse>

BackendAdapter.execute


executeStream(request, signal?): IRChatStream

Defined in: backend/src/providers/cerebras.ts:312

Execute streaming request.

IRChatRequest

AbortSignal

IRChatStream

BackendAdapter.executeStream


fromIR(request): CerebrasRequest

Defined in: backend/src/providers/cerebras.ts:158

Convert IR to Cerebras format.

IRChatRequest

CerebrasRequest

BackendAdapter.fromIR


healthCheck(): Promise<boolean>

Defined in: backend/src/providers/cerebras.ts:469

Health check.

Promise<boolean>

BackendAdapter.healthCheck


toIR(response, originalRequest, latencyMs): IRChatResponse

Defined in: backend/src/providers/cerebras.ts:211

Convert Cerebras response to IR.

CerebrasResponse

IRChatRequest

number

IRChatResponse

BackendAdapter.toIR