How arpit2811/arpit-s_ai_persona works
Arpit AI Persona serves questions and interview-booking requests across chat and voice-oriented API entry points. The principal flow is channel input → FastAPI session routing → either RAG retrieval and answer generation or a seven-stage booking workflow → channel response. Knowledge is built from the resume and repository READMEs into a FAISS index. Booking delegates availability and calendar creation to Make.com webhooks. The tree confirms the voice endpoint and web chat; Twilio/Vapi routing is documented but not implemented in the sampled files.
Components 22
User Interfaces
- Web Chat
frontend/app.py
API Orchestration
- FastAPI API
backend/main.py - Session State
backend/main.py
Conversation Flows
- Booking Router
backend/booking_workflow.py - Booking Stages
backend/booking_workflow.py - RAG Router
backend/rag_retrieve.py
Knowledge System
- Query Routing
backend/rag_retrieve.py - Semantic Retrieval
backend/rag_retrieve.py - Context Assembly
backend/rag_retrieve.py - Answer Generation
backend/rag_retrieve.py - Resume Source
data/Resume_Arpit_v2026.docx - Repository READMEs
data/repos - Ingestion Pipeline
backend/ingest.py - FAISS Index
index/resume.index - Chunk Metadata
index/chunks.json
External Integrations
- Make Service
backend/make_service.py - Make.com Webhooks
- GPT-4o-mini
Other components
- User
- Voice Channel
- Twilio
- Vapi Voice
Connections 24
- User → Web Chat: enters chat
- Web Chat → FastAPI API: posts message
- Voice Channel → Vapi Voice: speaks
- Vapi Voice → FastAPI API: posts voice
- Twilio → FastAPI API: routes calls
- FastAPI API → Session State: reads sessions
- FastAPI API → Booking Router: handles booking
- FastAPI API → RAG Router: asks knowledge
- FastAPI API → Make Service: serves calendar API
- Booking Router → Booking Stages: advances stages
- Booking Router → Make Service: checks and books
- RAG Router → Query Routing: routes query
- RAG Router → Semantic Retrieval: retrieves context
- RAG Router → Context Assembly: assembles context
- RAG Router → Answer Generation: generates answer
- Semantic Retrieval → FAISS Index: searches vectors
- Semantic Retrieval → Chunk Metadata: reads chunks
- Answer Generation → GPT-4o-mini: requests completion
- Ingestion Pipeline → Resume Source: parses resume
- Ingestion Pipeline → Repository READMEs: parses READMEs
- Ingestion Pipeline → FAISS Index: writes vectors
- Ingestion Pipeline → Chunk Metadata: writes metadata
- Make Service → Make.com Webhooks: calls webhooks
- FastAPI API → Session State: writes sessions