Machine Learning & AI

Compare local and cloud decision models, see how wording changes a guardrail verdict, and build support triage with explicit priority and fallback rules.

Build an evaluation harness that connects agent routes, tool evidence, traces, and calibrated judges to explain why an application succeeds or fails.

The final Akgents chapter: turn a catalog-defined team into a service with a browser client, durable history, and human approvals over HTTP.

See how a Qdrant contribution turned Akgents vector search into a pluggable backend architecture, then follow the sample agent team.

Move Akgents team configuration from Python into a validated, referenceable catalog without changing the runtime.

Combine BaseAgent, typed messages, read-only tools, deterministic routing, and human approval in one durable Akgents workflow.

Build reusable Akgents tools with ToolCard, ToolFactory, observers, channels, MCP, and one observable event stream.

Give an Akgents application an LLM, tools, budgets, context, capabilities, and a visible ReAct loop.

Build a typed Pydantic AI case agent with dependencies, outputs, handoffs, Capabilities, MCP, and observability.

Learn how akgentic-team turns an actor tree into a persistent, resumable team with cards, metadata, events, and lifecycle.