Home › README
Companion repository: OASIS methodology, primarily Chapter 21: Deployment, Operations and AgentOps, plus Chapters 17, 18, 19, and 26 (see below)
Helm is the AgentOps/LLMOps companion to the OASIS methodology, the practical discipline of running intelligence systems in production once they're built: versioning, tracing, quality monitoring, incident response, rollback, and learning loops, across six operational planes rather than one dashboard. The name reflects what the repo actually governs. Steering, correcting course, and staying at the controls once a system is live, not building or launching it.
In plain terms: OASIS's methodology chapters tell you what production AgentOps requires and why. Helm gives you how: the actual trace schema to log, the actual release manifest checklist, the actual incident runbook to follow when something breaks at 2am.
- 00-foundations: what AgentOps is and isn't, and the six operational planes (Service, Intelligence, Risk, Human, Economic, Outcome)
- 01-observability: the trace_event schema every logged decision must carry, per-plane instrumentation, dashboards and alerting, and the outcome scorecard/metric architecture
- 02-release-management: the immutable release manifest, rollout strategies, rollback vs. forward-fix, and reliability engineering/error budgets
- 03-incident-response: the responsible-layer taxonomy, the evaluation-side failure taxonomy, incident record template, and runbooks
- 04-learning-loop: the production learning loop, and the evaluation hierarchy/dataset strategy behind its regression-testing step
- 05-tooling-landscape: a comparison of LLMOps/AgentOps platforms (LangSmith, Langfuse, Arize Phoenix, Weights & Biases Weave, Helicone, and others)
- 06-security-and-containment: defense-in-depth control layers, the agentic threat model, and containment/emergency control (kill switches, blast radius)
- 07-tool-integration: tool contracts, integration principles, integration patterns, and MCP tool catalogues
- glossary: shared terminology
- templates: copyable starting points (incident record, release manifest, runbook)
Not yet covered: Chapter 20 (Governance, Compliance and Regulatory Engineering) and Chapter 22 (Economics, FinOps and Sustainability), both are largely owned by Responsible-AI and Fulcrum respectively, but this repo's Risk and Economic plane instrumentation is what feeds them.
- OASIS: the parent methodology; Helm implements its deployment, operations, evaluation, security, and measurement chapters specifically, the way Fulcrum implements the FinOps/economics chapters and Responsible-AI implements the governance chapters.
- Nexus: the opportunity catalog (which use case, in which domain/function). Once a use case from Nexus reaches deployment, Helm governs how it's operated.
- Verity: the deep evaluation and reliability methodology (Chapter 18); Verity is upstream of Helm's observability dashboards, release management, and incident response: Helm operationalizes production reliability, Verity defines how quality is measured in the first place.
- Fulcrum: the Economic plane defined in Helm's observability spec is the raw telemetry Fulcrum's unit-economics discipline depends on.
- Responsible-AI: the Risk plane and incident response process here operationalize the controls Responsible-AI defines; a release manifest and incident record are governance evidence generated as a byproduct of running the system well.
This repo is in early build-out: the section structure is in place and content is being filled in progressively, section by section. Check each page to see what's drafted versus still a stub.
Next: 00-foundations/what-is-agentops.md
Licensed under CC BY-SA 4.0. Reuse and adaptation are welcome with credit to KnowledgeTrails-OASIS, a link to the license, an indication of changes, and release of adaptations under the same license.
Shripadraj Mujumdar is an Agentic AI & Automation Strategist, Advisor, and Responsible AI Expert with 28+ years of experience in enterprise architecture and AI-driven transformation, including deep hands-on work in Agentic AI, Generative AI, and enterprise data and knowledge platforms. His practice spans designing multi-agent systems, knowledge-graph and RAG architectures, accelerated delivery capabilities, and Responsible AI governance frameworks aligned to global regulatory standards. This methodology ecosystem distills that practitioner experience; architecture, delivery, evaluation, governance, and economics. Into a single, reusable body of work.
Ankit Mirajkar is a Data & AI Architect and technology consultant specializing in modern data platforms, enterprise data architecture, and Agentic AI. His expertise spans scalable data engineering, AI-ready data platforms, Generative AI, and cloud technologies, with a strong focus on turning complex data challenges into practical, production-ready solutions. He also works at the intersection of architecture, technology strategy, and innovation to help organizations build intelligent, scalable data ecosystems.