Architecture
HARP owns evidence, Nero-Camp owns workflow state, OpenRAG owns semantic retrieval, MCP owns agent access, LiteLLM owns model routing, and n8n owns workflow glue.
Prometheus, Alertmanager, Uptime Kuma, n8n webhook, or human question.
Payload becomes a query, alert context, or task input.
HARP searches BM25 and optional OpenRAG semantic index.
RRF merge produces cited sources, snippets, confidence, and gaps.
Nero-Camp receives task/evidence state and asks for human review.
Outcome becomes RCA, runbook, or session log for future retrieval.
Module Map
Indexes runbooks, RCAs, session logs, monitoring docs, and generated evidence packs. Exposes FastAPI endpoints and read-only MCP tools.
Owns: Search, evidence packs, evals, policy labels, MCP surface.
Does not own: Task approval, live mutation, autonomous remediation.
Stores synchronized HARP knowledge documents in a semantic/vector backend to improve symptom-language discovery beyond exact keywords.
Owns: Vector search, embeddings, semantic ranking.
Does not own: HARP policy, Nero workflow state, final decisions.
Lets AI clients ask HARP for tools, resources, and prompt templates through a standard interface.
Owns: The read-only access contract for AI clients.
Does not own: Shell, Git writes, kubectl mutation, remediation.
Consumes HARP evidence as citations and advisory metadata. Keeps source records, timelines, approvals, and human decisions.
Owns: Task state, approval, audit timeline, operator decision.
Does not own: Assuming HARP advice is executable action.
Provides an OpenAI-compatible gateway for local or cloud models, including a gated AWS Bedrock path when policy allows. Useful for classification and summaries, not control.
Owns: Model routing and provider abstraction.
Does not own: Retrieval, approvals, task state, source citations.
Connects events around the core reliability flow without becoming the source of operational truth.
Owns: Webhook fan-out, schedules, notification routing, low-risk event glue.
Does not own: Evidence scoring, policy, approval state.
Safety Boundary
| Layer | Allowed | Blocked | Reason |
|---|---|---|---|
| MCP | Search, hybrid search, source preview, triage rendering. | Shell, Git write, kubectl mutation, evidence writes. | Agents need context, not production power. |
| HARP | Evidence rendering, citation ranking, policy labels, evals. | Autonomous remediation and source-of-truth task decisions. | HARP is memory and evidence, not the operator. |
| Nero-Camp | Task state, source records, timeline, approval, closeout. | Treating HARP output as executable authority. | Nero records the human decision. |
| OpenRAG | Semantic indexing and retrieval. | Direct agent access that bypasses HARP policy. | HARP mediates retrieval and fallback behavior. |
| LiteLLM | Classification and summarization when policy allows. | Replacing citations or final decisions. | Model output must remain grounded and governed. |