Skip to content

HUDHUD v1.0: the entry, as it stands for judging

Latest

Choose a tag to compare

@matx104 matx104 released this 26 Oct 14:25
· 4 commits to main since this release

HUDHUD: AI match commentary that proves every claim. My entry to the Microsoft × Premier League Inside the Game: Developer Hackathon (October 2026), as it stands for judging. Solo build.

Five agents work a live synthetic match. The Scout detects what matters, the Narrator writes the line on Microsoft Foundry, the Critic recomputes every number from the raw events and refuses what it can't prove, the Localizer airs it in English, Urdu or Arabic, and the Orchestrator keeps the broadcast on air when a model fails.

For the judges

Run it yourself

git clone https://github.com/matx104/HUDHUD && cd HUDHUD
make setup && make test        # 278 tests
make studio-offline            # the whole studio at http://localhost:8091, no Azure needed

Attached

  • HUDHUD-read-me.pdf: the entry on one page
  • HUDHUD-deliverables-report.pdf: every hackathon requirement and where it's met
  • HUDHUD-recap-seed-21.pdf: a recap the match wrote itself, every line checked by the Critic

Built on Microsoft Agent Framework, Microsoft Foundry (gpt-4.1-mini, gpt-5-mini, Foundry's evaluators and tracing), Azure Container Apps, Azure AI Speech, Azure AI Content Safety, Azure Cosmos DB and Application Insights, with an MCP server and an A2A agent. Every club, player and match is invented, as the rules require.