Gemini for developers

Gemini is Google’s family of multimodal AI models. In this talk, you’ll learn about the models, witness their powerful capabilities such as deep thinking, Google Search and Map grounding, computer use tool, spatial understanding, live API, and native image and audio output. You’ll also see demos of Gemini’s coding agent tools Antigravity and Gemini CLI, and also learn how to integrate Gemini into your applications using the Google Gen AI SDK.

Transcript

1. Google Cloud 03 Proprietary & Confidential Last update: Jan 2026 Gemini for Developers
2. Mete Atamel Developer Advocate Google Cloud atamel.dev speakerdeck.com/meteatamel @meteatamel
4. Gemini is an umbrella brand used in Google products that utilize the Gemini model
6. Gemini app — gemini.google.com
7. Gemini app — gemini.google.com
8. Gemini app — gemini.google.com
9. Google AI Studio — ai.dev
10. Vertex AI
11. Vertex AI Studio console.cloud.google.com/vertex-ai/studio DEMO
12. Gemini CLI — geminicli.com
13. Google Antigravity — antigravity.google
15. docs.cloud.google.com/vertex-ai/generative-ai/docs/models Latest Gemini Models
16. 3 Pro Largest model tier, for complex tasks 3 Flash Best model for general performance across a wide range of tasks 2.5 Flash-Lite Lightweight model, optimized for speed and cost efficiency
17. Google — Confidential & Proprietary What makes Gemini unique? Multimodal Live API Natively Multimodal Long context Google — Confidential & Proprietary Advanced coding
18. Long context up to 10M in research Google — Confidential & Proprietary 2M Gemini 3.0 Pro & Flash (Preview) 1M GPT-4.5 128k Gemini 1.0 Pro 32k Claude 3.7 200k
19. Google Gen AI SDK
20. Google Gen AI SDK* *SDK is unified but features are not always available in both
21. Python Go Java Node.js .NET Google Gen AI SDK
22. Google AI API (Google AI Studio) client = genai.Client( api_key=your-gemini-api-key) response = client.models.generate_content( model="gemini-3-flash-preview", contents="Why is the sky blue?")
23. Google Cloud API (Vertex AI Studio) client = genai.Client( vertexai=True, project=your-google-cloud-project, location="us-central1") response = client.models.generate_content( model="gemini-3-flash-preview", contents="Why is the sky blue?")
24. Using environment variables Google AI API (Google AI Studio) export GOOGLE_API_KEY='your-api-key' Google Cloud API (Vertex AI Studio) export GOOGLE_GENAI_USE_VERTEXAI=true export GOOGLE_CLOUD_PROJECT='your-project-id' export GOOGLE_CLOUD_LOCATION='us-central1' client = genai.Client() Common client initialization
25. 🆕 Interactions API (beta)* The Interactions API is a unified interface for interacting with Gemini models and agents It simplifies state management, tool orchestration, and long-running tasks *Currently only supported on Google AI API (Google AI Studio)
26. Basic interaction interaction = client.interactions.create( model="gemini-3-flash-preview", input="Tell me a short joke." ) print(interaction.outputs[-1].text)
27. Stateful conversation interaction1 = client.interactions.create( model="gemini-3-flash-preview", input="Hi, my name is Phil." ) print(f"Model: {interaction1.outputs[-1].text}") interaction2 = client.interactions.create( model="gemini-3-flash-preview", input="What is my name?", previous_interaction_id=interaction1.id ) print(f"Model: {interaction2.outputs[-1].text}")
28. Agents interaction = client.interactions.create( input="Research the history of the Google TPUs.", agent="deep-research-pro-preview-12-2025", background=True ) while True: if interaction.status == "completed": print("\nFinal Report:\n", interaction.outputs[-1].text) break
29. Performance Gemini Live API Deep Think Mode Tools & Agents Image & Audio Output Spatial Understanding
30. Performance
32. Deep Think Mode
33. The Gemini 3 and 2.5 models use an internal thinking process that significantly improves their reasoning for complex tasks Thinking mode in Gemini
34. Thinking levels (Gemini 3) and budgets (Gemini 2.5) to control thinking behaviour Enable include_thoughts=True to see model's raw thoughts Thinking mode in Gemini
35. Thinking configuration response = client.models.generate_content( model="gemini-3-pro-preview", contents="How does AI work?", config=types.GenerateContentConfig( thinking_config=types.ThinkingConfig( thinking_level="low", include_thoughts=True ) ), ) DEMO
36. Tools & Agents
37. Google Search Tool Ground model responses in Google Search results For more accurate, up-to-date, and relevant responses
38. Google Search Tool google_search_tool = Tool(google_search=GoogleSearch()) response = client.models.generate_content( model="gemini-3-flash-preview", contents="How’s the weather like today in London?", config=GenerateContentConfig(tools=[google_search_tool]) ) DEMO
39. Google Maps Tool Ground model responses with Google Maps, which has access to information on over 250 million places
40. Google Maps Tool google_maps_tool = Tool(google_maps=GoogleMaps()) response = client.models.generate_content( model="gemini-3-flash-preview", contents=""What are the best restaurants near here?", config=GenerateContentConfig(tools=[google_maps_tool]), # Optional: Provide location context (this is in Los Angeles) tool_config=ToolConfig(retrieval_config=types.RetrievalConfig( lat_lng=types.LatLng( latitude=34.050481, longitude=-118.248526))), ) DEMO
41. Code Execution Tool Model generates and runs Python code with a list of supported libraries (pandas, numpy, PyPDF2, etc.) Useful for applications that benefit from code-based reasoning (e.g. solving equations)
42. Code Execution Tool code_execution_tool = Tool(code_execution=ToolCodeExecution()) response = client.models.generate_content( model="gemini-3-flash-preview", contents="What is the sum of the first 50 prime numbers?", config=GenerateContentConfig( tools=[code_execution_tool], temperature=0)) DEMO
43. Computer Use Tool 💻 The Gemini Computer Use model (preview) enable you to build browser control agents to automate tasks
44. Computer Use Tool github.com/google-gemini/computer-use-preview DEMO
45. File Search Tool – effortless RAG
46. File Search Tool file_search_store = client.file_search_stores.create( config={'display_name': 'your-fileSearchStore-name'}) client.file_search_stores.upload_to_file_search_store( file='sample.txt', file_search_store_name=file_search_store.name, config={ 'display_name' : 'display-file-name',})
47. File Search Tool response = client.models.generate_content( model="gemini-3-flash-preview", contents="""Can you tell me about [insert question]""", config=types.GenerateContentConfig( tools=[ types.Tool( file_search=types.FileSearch( file_search_store_names=[file_search_store.name] ) ) ] ) )
48. Gemini Deep Research Agent Autonomously plans, executes, and synthesizes multi-step research tasks It navigates complex information landscapes using web search or your own data to produce detailed, cited reports
49. Gemini Deep Research Agent interaction = client.interactions.create( input="Research the history of the Google TPUs.", agent="deep-research-pro-preview-12-2025", background=True ) while True: if interaction.status == "completed": print("\nFinal Report:\n", interaction.outputs[-1].text) break
50. Custom Tools with Function Calling No Manual Auto MCP
51. Function Calling def get_current_weather(location: str) -> str: """Example method. Returns the current weather. Args: location: The city and state, e.g. San Francisco, CA """ weather_map: dict[str, str] = { "Boston, MA": "snowing", "San Francisco, CA": "foggy", "Seattle, WA": "raining", "Austin, TX": "hot", "London, UK": "rainy and dark", } return weather_map.get(location, "unknown")
52. Before: Manual Function Calling
53. Now: Automatic Function Calling Submit a Python function for automatic function calling (instead of submitting an OpenAPI specification of the function)
54. Automatic Function Calling response = client.models.generate_content( model="gemini-3-flash-preview", contents="What is the weather like in Austin?", config=GenerateContentConfig( tools=[get_current_weather], temperature=0))
55. MCP from google.adk.agents import LlmAgent from google.adk.tools.mcp_tool.mcp_toolset import MCPToolset root_agent = LlmAgent( model='gemini-3-flash-preview', name='maps_assistant_agent', instruction='Help the user with mapping, directions, and finding places using Google Maps tools.', tools=[ MCPToolset( connection_params=StdioConnectionParams( server_params = StdioServerParameters( command='npx', args=[ "-y", "@modelcontextprotocol/server-google-maps", ],
56. Spatial Understanding
57. Improved accuracy on 2D and 3D spatial understanding Spatial Understanding DEMO
58. Gemini Live API
59. Enables low-latency, two-way interactions → Input: text, audio, and video ← Output: audio and text gemini-2.5-flash-native-audio-preview-12-2025 Gemini Live API
60. Multimodal – model can see, hear, speak Low-latency – for realtime interaction Memory – model remembers the session Tools – Function calling, code execution, and Google search Gemini Live API–key capabilities DEMO
61. Native Image & Audio Output
62. Native image generation: 🍌 gemini-3-pro-image-preview 🍌 gemini-2.5-flash-image Native text-to-speech: gemini-2.5-flash-preview-tts gemini-2.5-pro-preview-tts Native Image & Audio Output DEMO
63. Thank you Mete Atamel Developer Advocate Google Cloud atamel.dev speakerdeck.com/meteatamel @meteatamel