Conversational support, reimagined with OpenAI.
Tactara’s Voice Agent is a plug-and-play AI application that transforms your documentation into a real-time voice assistant for customer support. Built on OpenAI's GPT-4o and Text-to-Speech (TTS) stack, this system crawls through technical docs, generates semantic embeddings, and delivers answers in natural-sounding speech — all through a clean Streamlit UI.
- Crawls documentation using Firecrawl
- Parses and stores structured data in Qdrant
- Embeddings via FastEmbed for lightning-fast semantic search
- Documentation Processor: Translates structured docs into clear, human-readable answers
- TTS Agent: Converts answers into lifelike speech with adjustable voice options
- Voice Customization: Choose from multiple OpenAI voices like
nova,shimmer,onyx, and more
- Built with Streamlit for quick deployments
- Audio player with download support
- Live progress indicators for transparency during processing
git clone https://github.com/Tactara/customer_support_voice_agent.git
cd customer_support_voice_agentpip install -r requirements.txtYou'll need credentials from:
Update the config or input via the sidebar UI.
streamlit run customer_support_voice_agent.py- Enter your API keys via the Streamlit sidebar
- Paste a documentation URL (limit: 5 pages per crawl)
- Select a voice from the dropdown
- Click “Initialize System”
- Ask a question → Get spoken + written answers in real time
| Feature | Description |
|---|---|
| 🔗 Firecrawl | Smart crawler that preserves structure & metadata |
| 🧬 FastEmbed + Qdrant | Fast, semantic vector search for matching content |
| 🗣️ OpenAI TTS | Natural-sounding responses with customizable voices |
- Cloud Service Providers needing automated Tier-1 support
- Internal IT teams seeking self-service documentation tools
- Developer platforms with complex user documentation
Built with ❤️ by Tactara