https://github.com/google/agents-cli

Hi @Sri Mathi ! Great to see your interest in Agentic AI — you’re focusing on the right areas already. For learning practical AI agents, I’d strongly recommend checking out AI with Brandon YouTube Channel. He has some really good hands-on content around Google ADK, AI agents, workflows, orchestration, and real-world implementation patterns. For your roadmap, I’d suggest learning in this order: 1. AI Agent Fundamentals Prompting Tool calling Memory Planning & reasoning Function execution 2. Single-Agent Frameworks LangChain LangGraph Google ADK CrewAI AutoGen 3. RAG Systems Embeddings Vector DBs (FAISS, Chroma, Pinecone) Retrieval pipelines Document processing 4. Multi-Agent Systems Agent communication Task delegation Workflow orchestration Human-in-the-loop systems 5. Deployment & Production FastAPI Docker Cloud deployment (GCP/AWS) Monitoring & evaluation Some additional resources: - LangChain Docs - LangGraph Docs - CrewAI Docs - Microsoft AutoGen Docs - Google ADK Docs The best way to learn is by building projects continuously. Start with a simple RAG chatbot, then move into tool-using agents and multi-agent workflows. Consistency matters more than speed.

As an AI Engineer, I use the Gemini CLI mainly as a deep research and implementation agent in my workflow. It’s really useful for planning features, researching the latest docs, and helping with implementation directly from the terminal. Since I work a lot with Google-related frameworks and GCP, I’ve noticed it gives more accurate results for things like Vertex AI, Firebase, TensorFlow, Cloud Run, and other Google ecosystem tools. I also use it to access and manage GCP through code instead of always using the console. It makes deployment, debugging, and maintenance much faster. For example, I can generate deployment scripts, analyze errors, or automate repetitive cloud tasks directly from the CLI. For me, the biggest advantage is reducing context switching research, coding, implementation, and cloud operations all happen in one place.

Happy New Year My friends 🙌

Tnq @Tom Welsh

Yes, absolutely,Firestore is not special or required. Any persistent store (SQL or NoSQL) that can save and retrieve messages can be used with LangChain. Supabase is actually a great choice if you want strong querying, auth, and long-term conversation storage in one place.