Building Cross-Platform, Multi-Modal Agents with Flutter
summary
Built a cross‑platform agent platform that ships voice bots, document authoring, media gen and real‑time collaboration using Flutter web + mobile, a Python backend, and a custom websocket protocol. Agents manipulate local docs with CRDTs to sync discrete edits (Google Docs style), generate classroom worksheets via OpenAI GPD1 image model, and run voice through WebRTC & LiveKit. Production stack includes Superbase for auth, ClickHouse for analytics and LanceDB for vector search; examples include an allergy‑aware teacher agent and live browser automation that successfully found session times.
AI-written summary · about
chapters (21)
- 0:00Multi-modal AI agents with Flutter
- 1:06Proactive agents with context
- 2:54Multiplayer document editing
- 5:04Dynamic UI for agent tools
- 6:30Custom classroom assets
- 8:39Localization with agents
- 10:23Voice bots with Flutter & WebRTC
- 11:49Browser automation without APIs
- 13:15Flutter’s cross-platform value
- 14:38Open source AI agent stack
- 16:24Kubernetes for agent workflows
- 18:51Flutter packages for agent UIs
- 19:32Agents for real-time collaboration
- 21:20Cross-platform agents with Flutter
- 22:05Agents as new UI paradigm
- 23:32LLMs as agent frameworks
- 26:21Agents need dedicated backend
- 27:28Stateful agents & persistence
- 29:34Limitations of MCP protocol
- 31:19Dart limits for AI agents
- 32:43Q&A
related talks
32mHow I Build Web Applications in 2026
40mAI Agents in Practice
49mAI Agents: Collaboration in Real-Life Applications
41mHow AI Will Change Mobile Development
01h 48mBuilding Conversational Agents
19mThe Future of AI
29mClient side Web AI Agents Building smarter user experiences for a future age
01h 26mBuilding voice agents with OpenAI