mcp_knowledge_ops 0.4.0

Knowledge operations including pipelines, workflows, scheduling, and automated processing for curation, summarization, and pattern mining in the MCP ecosystem.

MCP Knowledge Ops

Positioning: mcp_knowledge_ops is an internal component of the MakeMind knowledge stack exposed through the mcp_knowledge facade. Application code should import package:mcp_knowledge/mcp_knowledge.dart — the symbols declared here are re-exported from there. Direct package:mcp_knowledge_ops/ imports remain valid for advanced or integration scenarios but are discouraged in product code.

A stateless orchestration layer for MakeMind. Coordinates pipelines, workflows, scheduling, runbooks, and observability across the ecosystem; persistence is delegated to mcp_fact_graph.

Components

• Execution Engine — runs pipeline / workflow definitions: stages with retry and failure policies, gates, checkpoints with resume, review / approval stages through the ApprovalPort, parallel execution.
• Pipelines and workflows — declarative specifications (ingest, curation, summary refresh, pattern mining, index rebuild; skill, profile and bundle build); their run() is a placeholder, the host builds engine definitions and registers the stage handlers.
• Scheduler — cron (* or exact numbers per field), event, deadline, manual and composite triggers with timeout, concurrency and de-duplication.
• Runbooks — operational procedures with variables, checks and approvals.
• Behavior engine — guard / outcome based step runner with suspend and resume.
• Events, observability — event toolkit (publisher, subscriber, router, store, replay, dead letters, reliability) and logging, metrics, tracing, audit, alerts, health checks as independent tools.
• Standard port adapters — OpsRuntime provides WorkflowPort, PipelinePort, ScheduleTriggerPort, AuditPort and RunbookPort from the mcp_bundle contract layer.

Quick Start

import 'package:mcp_knowledge_ops/mcp_knowledge_ops.dart';

// 1. Run a definition on the engine.
final handlers = HandlerRegistry()
  ..register('myHandler', MyHandler());

final engine = ExecutionEngine(
  config: const ExecutionEngineConfig(),
  handlers: handlers,
);

final result = await engine.executePipeline(definition, input, context);

// 2. Or assemble the whole runtime from your ports. The built-in pipelines
// are specifications: runPipeline calls their placeholder run(), so use the
// runtime's engine with your own definition to execute stages.
final runtime = OpsRuntime.fromConsumedPorts(consumedPorts);
runtime.handlers.register('myHandler', MyHandler());
final result2 = await runtime.engine.executePipeline(definition, input, context);

See docs/ for the requirements, design, test lists and operations guides.

Support

• Issue Tracker
• Discussions

License

MIT — see LICENSE.