integrate-hiveloop

What problem does it solve?

Agentic applications often lack structured observability for agent identity, task lifecycle, LLM usage, costs, and errors; this Skill guides integrating the HiveLoop SDK to capture and stream those telemetry events to HiveBoard for actionable monitoring and troubleshooting.

Core Features & Use Cases

• SDK initialization: Ensure hiveloop.init runs at process startup so telemetry is available from the first agent action.
• Agent registration: Register each agent with hb.agent to attach identity, framework, and heartbeat state.
• Task lifecycle wrapping: Wrap units of work with agent.task to emit task_started/task_completed and attach payloads.
• LLM call tracking: Record tokens, latency, and cost via task.llm_call or agent.llm_call to power cost and latency dashboards.
• Shutdown and log bridging: Ensure hiveloop.shutdown runs on exit and optionally add HiveBoardLogHandler to forward warnings/errors as issues.
• Use case example: Instrument a sales lead scoring pipeline or a LangChain agent to see per-task LLM costs, action trees, and error trends in HiveBoard.

Quick Start

Initialize hiveloop with your API key at process startup, register each agent with hb.agent, wrap each work unit with agent.task, record LLM responses with task.llm_call, and call hiveloop.shutdown on exit.