Agent Skills Dashboard
Performance charts comparing GPT-4o vs. Llama 3 across my agents, tool-use accuracy metrics, and a live playground to trigger real agent actions.
A high-signal technical showcase: rather than claiming AI expertise, this dashboard shows the data from running my agents in production over time.
Key Charts
Success Rate vs. Latency Scatter plot comparing GPT-4o and a locally-hosted Llama 3 across agent tasks. Shows the cost/speed/quality tradeoff I’ve benchmarked directly — not theoretical numbers.
Tool Usage Accuracy Bar chart showing how often each agent correctly selected and called the right custom API tool versus falling back or hallucinating a tool call. Broken down by task type.
Live Playground
A “Playground” section lets visitors trigger a real tool-use event:
“Agent, get the weather in my location and recommend a project from my portfolio based on the current mood.”
The agent calls a weather API, reasons about the result, and returns a personalised project recommendation — with the full tool-call trace visible below the response.
Data Source
All metrics come from logged agent runs stored in Postgres, populated by the other three showcases as they run. The dashboard is the observability layer for the entire portfolio.