I will build a browser ai voice assistant with rag and tool calling


About this gig
Let users ask questions out loud and take useful actions inside your web app.
I build browser AI voice assistants with document retrieval and controlled tool calling. My work includes a textbook voice tutor with page context, retrieval tools, transcripts and PDF navigation. Gallery diagrams show the implementation; they are not a latency benchmark.
BASIC $100: one English browser voice assistant, one simple mock-data function, local demo, source and setup notes. No RAG, login or deployment.
STANDARD: 5 text-readable PDFs/50 pages total, one retrieval tool, one agreed API tool, transcript, web UI and one deployment.
PREMIUM: Standard plus login, two API tools total and handover. One app and knowledge base.
Python/FastAPI with React/Next.js and OpenAI Realtime. Credentials stay on the server. We agree sample conversations, allowed actions and acceptance checks first.
API, voice, database and hosting fees are separate. Phone/SIP calling, WhatsApp, complex scans, multi-tenant SaaS and additional languages need custom scope. No guaranteed response time or perfect answers.
Message me with your use case, sample documents and the action your assistant should perform.
Get to know Mangesh W
RAG and Voice AI Developer
- FromIndia
- Member sinceSep 2024
Languages
English
My Portfolio
FAQ
What does the $100 package include?
One English browser voice prototype, one mock-data function, local demo, source and setup notes. It does not include document retrieval, login, deployment or a production SaaS.
Can the assistant use my PDFs and APIs?
Standard includes 5 text PDFs/50 total pages, one retrieval tool and one API tool. Premium includes two API tools total and login. We agree schemas, permitted actions and test conversations before starting.
Does this include phone calls or WhatsApp?
These packages are for a browser app. Phone/SIP calling, WhatsApp, multi-tenant SaaS and additional languages require a separate agreed scope.
Are usage fees and performance guarantees included?
API, voice, database and hosting costs are paid separately by you. Latency and answer quality depend on the model, network and documents. We use agreed acceptance checks; I do not guarantee perfect answers or a universal response time.

