agentkit-sdk 0.1.0

Developer SDK for building personalized voice AI assistants

AgentKit

Developer SDK for building personalized voice AI assistants with mobile APK generation.

Quick Start

# Install
pip install agentkit

# Initialize a new agent project
agentkit init my-agent

# Add API keys to .env
cd my-agent
cp .env.example .env

# Start the server
agentkit serve

Configuration

Edit agent.config.yaml:

agent:
  name: "Assistant"
  persona: "You are a helpful personal assistant..."
  language: hinglish

voice:
  stt:
    provider: sarvam  # or deepgram
    api_key: ${SARVAM_API_KEY}
  tts:
    provider: sarvam  # or elevenlabs
    voice: "meera"
    api_key: ${SARVAM_API_KEY}

llm:
  provider: gemini  # or openai
  model: gemini-2.0-flash
  api_key: ${GEMINI_API_KEY}
  temperature: 0.7

memory:
  type: markdown  # or vector
  backend: local  # or qdrant

learning:
  enabled: true
  correction_detection: true
  implicit_feedback: true

Commands

• agentkit init - Initialize a new project
• agentkit serve - Start the FastAPI server
• agentkit build android - Build Android APK
• agentkit deploy --platform railway - Deploy to Railway or Render

Architecture

• CLI - Python Click commands for init, serve, build, deploy
• Pipeline - Streaming STT → LLM → TTS orchestration with <500ms latency
• Memory - Markdown (MVP) or Qdrant vector storage with semantic retrieval
• Learning - Explicit/implicit correction detection, proactive recommendations
• Mobile - React Native + Expo VoiceOrb interface

Environment Variables

SARVAM_API_KEY=your_sarvam_key
GEMINI_API_KEY=your_gemini_key
DEEPGRAM_API_KEY=your_deepgram_key
ELEVENLABS_API_KEY=your_elevenlabs_key
OPENAI_API_KEY=your_openai_key