LokerDollar Voice

Created by team LokerDollar on September 27, 2026

Remote jobs paying US dollars can earn an Indonesian worker about seven times Jakarta's minimum wage, but they're hard to find. Most "remote" listings quietly exclude Asia, salaries are quoted per year in dollars, and searching means English filters, while many here think in Bahasa Indonesia. LokerDollar Voice lets you simply talk. Say "find me remote customer support jobs that pay in dollars" and the agent searches LokerDollar's live database of ~4,800 open remote jobs filtered for Indonesian applicants, showing numbered job cards and reading the top three with pay. Ask "tell me more about the second one" and it fetches that job, gives pay in rupiah per month and whether Indonesians can apply, and highlights the Apply button. You can cut in any time — the agent stops mid-sentence and follows your new request. How it uses AssemblyAI: the whole conversation (Universal-3.5 Pro streaming speech-to-text, the LLM, text-to-speech, turn detection, and barge-in) runs over one AssemblyAI Voice Agent API WebSocket. A Cloudflare Worker mints single-use temporary tokens so the API key never reaches the browser, and the agent is configured inline with one session.update. Job search and job details run as client-side function tools against the Worker, rendering cards and returning results exactly when reply.done is the latest event. A transcription prompt biases recognition toward job titles, tech stacks, and mixed Indonesian-English phrases, and the agent's word-level transcript (transcript.agent.delta) lights up the card it is talking about. Every result is real: jobs come from LokerDollar's production database through its public MCP server, and every Apply link goes to a real listing. Honest limitation: AssemblyAI doesn't support Indonesian speech yet. The agent understands mixed Indonesian-English and typed Indonesian, replying in simple English. Next: a phone line over AssemblyAI SIP for workers without laptops, mock interview practice, and native Bahasa once supported.

Category tags: