Skip to main content
SpekoLLM is a llm.LLM implementation. It flattens a LiveKit ChatContext into Speko’s messages format and calls the proxy. The router picks the best LLM provider per intent and fails over automatically. 
import { SpekoLLM } from '@spekoai/adapter-livekit';

const spekoLLM = new SpekoLLM({
  speko,
  intent: { language: 'en', vertical: 'general' },
  temperature: 0.7,
  maxTokens: 400,
});
Unlike STT and TTS, SpekoLLM doesn’t need a StreamAdapter — LiveKit’s voice.AgentSession is happy with a non-streaming LLM that emits a single chunk.

​
Constructor

new SpekoLLM(options: SpekoLLMOptions)

​
SpekoLLMOptions

FieldTypeRequiredDescription
spekoSpeko✅@spekoai/sdk client.
intentIntent✅Validated at construction time.
temperaturenumber?Forwarded to /v1/complete.
maxTokensnumber?Forwarded to /v1/complete.
constraintsPipelineConstraints?Allow-list constraints.

​
Properties

  • label() → 'speko.LLM'
  • provider = 'speko'
  • model = 'speko-router'

​
.chat(params)

Standard LiveKit LLM entry point. Returns an LLMStream that emits exactly one ChatChunk carrying the full assistant response, then closes. Signature (from @livekit/agents): 
chat(params: {
  chatCtx: llm.ChatContext;
  toolCtx?: llm.ToolContext;
  connOptions?: APIConnectOptions;
  parallelToolCalls?: boolean;
  toolChoice?: llm.ToolChoice;
  extraKwargs?: Record<string, unknown>;
}): llm.LLMStream;
The emitted chunk includes usage: 
{
  id: '<uuid>',
  delta: { role: 'assistant', content: result.text },
  usage: {
    promptTokens: result.usage.promptTokens,
    completionTokens: result.usage.completionTokens,
    promptCachedTokens: 0,
    totalTokens: result.usage.promptTokens + result.usage.completionTokens,
  },
}

​
Context conversion — chatContextToSpeko

Exported for when you want to reuse the flattening logic (e.g. unit tests, custom pipelines). 
import { chatContextToSpeko } from '@spekoai/adapter-livekit';

const messages = chatContextToSpeko(chatCtx);
Rules:
  • Only llm.ChatMessage items are considered. Function-call and handoff items are skipped.
  • Roles are normalised: developer → system; system / user / assistant pass through; anything else is dropped.
  • Empty textContent messages are skipped.
  • Ordering is preserved.
If the result is empty, .chat() rejects with SpekoAdapterError('INVALID_CONTEXT').

​
v1 — no tool calls

/v1/complete doesn’t expose tool invocation yet. If a non-empty toolCtx is passed, the adapter warns once (deduped per instance) and proceeds as if no tools were set: 
SpekoLLM: tool calls are not supported in v1 — `toolCtx` is being ignored. …
Remove tools from the agent or wait for streaming LLM proxy support.

​
Errors

  • SpekoAdapterError (exported): thrown for adapter-internal problems. code is one of:
    • 'INVALID_CONTEXT' — ChatContext produced no convertible messages.
API-layer errors from the underlying speko.complete() surface unchanged — SpekoApiError, SpekoAuthError, SpekoRateLimitError from @spekoai/sdk.