Mind the Gap - A Developer's Roadmap To Building AI Agents
Talkat Open Source Summit Europe 2025 · on Monday, 2025-08-25· at 14:25· Elicium 1 (Elicium Level 1)·
summary
Shows a practical pipeline: local models with Ollama, LM Studio and LLM Runner versus OpenAI/Anthropic in production, a homegrown Gordon agent that learns from bad Dockerfiles, a Docker model runner for local inference, and an agentic compose/Docker agent builder for wiring models, tools, secrets and MCP servers through an MCP gateway. Covers nondeterministic error handling, a “tool budget” limit from context windows, integration testing with QA agents, and file-backed secret injection that prevents token exfiltration.
AI-written summary · about
chapters (11)
- 0:00Building AI agents
- 2:10Docker’s AI agents journey
- 5:02From GenAI to Gordon the agent
- 7:10Local vs hosted models
- 10:04Agents versus workflows
- 11:28MCP and tool integration
- 13:12MCP server design
- 16:20Testing AI systems
- 19:54Agent frameworks
- 20:36MCP server config
- 24:10Summary and security
related talks
25mBuilding AI Workflows: From Local Experiments To Serving Users
01h 42mBuilding Responsible AI Agents in Decentralized Open-Source Environments
40mAgents in Action: Advancing Open Source AI Missions
39mBuilding Agentic AI Apps with Docker MCP & Cagent
07h 25mAI Agents Conference 2026 - Day 2
43mDocker Agent — How to Make Building AI Agents Even Simpler?
22mHOW TO BUILD YOUR OWN AI AGENT: THE COMPLETE ARCHITECTURE FOR ENGINEERING TEAMS
40mHow To Tame Your Agents?