How abhishekgiri1617/ai-receptionist works
This repository implements a configurable, voice-first customer-support receptionist. A caller speaks through a LiveKit session; speech is transcribed, the LangChain agent responds directly or routes the request to knowledge search, inquiry capture, or escalation, and the answer is spoken back. Separate ingestion prepares business documents for retrieval. The runtime also records transcripts and can synchronize conversation updates to an optional browser avatar UI. The README documents the main capabilities; some UI and integration internals were not sampled, so their connections are shown only where the excerpts support them.
Main entry points: main.py Voice Entrypoint · tools.py Agent Graph · rag.py Retrieval and Context · transcript.py Transcript Recorder
Components 21
Voice Runtime
- Speech I/O
voice_agent/main.py - Voice Entrypoint
voice_agent/main.py - Runtime Settings
config.py - Business Persona
agent_config.yaml
Agent and Support
- Agent Graph
langchain_agent/tools.py - Knowledge Search
langchain_agent/tools.py - Inquiry Capture
langchain_agent/tools.py - Human Escalation
langchain_agent/tools.py - Inquiry Log
langchain_agent/tools.py - Escalation Log
langchain_agent/tools.py
Knowledge Base
- Document Ingestion
ingest.py - Embedding Store
db.py - Qdrant Index
db.py - Retrieval and Context
langchain_agent/rag.py
Call Records
- Transcript Recorder
voice_agent/transcript.py - Transcript Files
voice_agent/transcript.py
Avatar UI
- Avatar Sync Server
avatar_agent_ui/sync_server.py - Avatar Browser UI
avatar_agent_ui/app.js
Other components
- Caller
- LiveKit Session
- Groq LLM
Connections 24
- Caller → LiveKit Session: speaks to
README.md - LiveKit Session → Voice Entrypoint: starts session
voice_agent/main.py - Voice Entrypoint → Speech I/O: configures
voice_agent/main.py - Speech I/O → Voice Entrypoint: transcribes
voice_agent/main.py - Voice Entrypoint → Agent Graph: adapts graph
voice_agent/main.py - Voice Entrypoint → Transcript Recorder: records turns
voice_agent/main.py - Voice Entrypoint → Avatar Sync Server: notifies events
voice_agent/main.py - Voice Entrypoint → Business Persona: reads persona
voice_agent/main.py - Voice Entrypoint → Runtime Settings: reads settings
voice_agent/main.py - Agent Graph → Knowledge Search: routes tools
langchain_agent/tools.py - Agent Graph → Inquiry Capture: routes tools
langchain_agent/tools.py - Agent Graph → Human Escalation: routes tools
langchain_agent/tools.py - Knowledge Search → Retrieval and Context: retrieves context
langchain_agent/tools.py - Retrieval and Context → Embedding Store: embeds query
langchain_agent/rag.py - Retrieval and Context → Qdrant Index: searches vectors
langchain_agent/rag.py - Retrieval and Context → Groq LLM: generates answer
langchain_agent/rag.py - Document Ingestion → Qdrant Index: indexes chunks
README.md - Document Ingestion → Embedding Store: generates embeddings
README.md - Inquiry Capture → Inquiry Log: appends inquiry
langchain_agent/tools.py - Human Escalation → Escalation Log: logs ticket
langchain_agent/tools.py - Transcript Recorder → Transcript Files: saves transcript
voice_agent/transcript.py - Transcript Recorder → Avatar Sync Server: writes live turn
voice_agent/transcript.py - Avatar Sync Server → Avatar Browser UI: serves and broadcasts
avatar_agent_ui/sync_server.py - Agent Graph → Caller: returns response