Emma Wang - Homework3-Submission

[EmmaW215]

Core Features

1. Speech Recognition (ASR)

• Engine: OpenAI Whisper
• Model: Small (configurable)
• Accuracy: >90%
• Latency: 2-3 seconds

2. Language Model (LLM)

• Model: Llama 3.2-1B
• Context: 5-turn conversation memory
• Response: Natural, conversational
• Latency: 3-5 seconds

3. Text-to-Speech (TTS)

• Engine: Google TTS (gTTS)
• Quality: Natural-sounding
• Languages: 100+
• Latency: 1-2 seconds

4. API Server

• Framework: FastAPI
• Protocol: HTTP REST
• Format: Multipart form-data
• Documentation: Auto-generated (Swagger)

5. Session Management

• Type: In-memory
• Capacity: Unlimited concurrent sessions
• History: Last 5 turns per session
• Timeout: 1 hour (configurable)