Hákon Freyr

Grimoire (SETS)

Enterprise Knowledge Management Platform. The enterprise evolution of Memory Palace, deployed on Azure at Wise. Inherits the same specification-pattern core and Neo4j graph backend, but adds full Azure infrastructure ownership (Bicep IaC, Container Apps, auto-scaling), three-RFC OAuth 2.0 authentication, and a pluggable connector architecture for Confluence, GitHub, and code complexity analysis — all queryable through a unified JSON DSL.

General notes

  1. Deployed on Azure Container Apps with Bicep IaC: Neo4j AuraDB, Azure Container Registry, auto-scaling (scale-to-zero dev, 1–3 replicas prod)
  2. Full OAuth 2.0 authentication spanning three RFCs (7591, 8414, 9728) — Claude.ai MCP connector auto-negotiates auth, zero-config enterprise deployment
  3. Pluggable connector architecture: Confluence space indexing, GitHub repo analysis, code complexity metrics — all queryable through unified specification-based JSON DSL
  4. Enterprise features: JWT identity tracking, background job orchestration with APScheduler, liveness/readiness health probes

Construction notes

  1. Azure Container Apps with Bicep IaC: Neo4j AuraDB for graph + vector search, Azure Container Registry, Log Analytics workspace, CI/CD via GitHub Actions with smoke test (MCP discovery endpoint validation)
  2. Dynamic Client Registration (RFC 7591), Authorization Server Metadata (RFC 8414), Protected Resource Metadata (RFC 9728). Domain-restricted email (@wise.is) — zero-config enterprise deployment
  3. Pluggable connector architecture: Confluence (space indexing, semantic page search with label/space filtering), GitHub repositories (clone → language detection → embedding generation → graph storage), code complexity analysis (Halstead, cyclomatic, cognitive)
  4. Same specification core evolved: 12+ discriminated union types generating type-safe Cypher, composable queries with similarity search, relationship traversal, and pagination in a single endpoint
  5. Multi-stage Docker builds with non-root users, liveness/readiness health probes, JWT identity tracking for audit trails

3 RFCs Full OAuth 2.0 stack (7591, 8414, 9728)

0→3 Auto-scaling replicas with scale-to-zero dev

IaC Full Azure infrastructure as Bicep code

Materials

  • Python/FastAPI
  • Neo4j AuraDB
  • Azure Container Apps
  • Bicep IaC
  • OAuth 2.0
  • MCP

Back to the cover sheet

Project
Hákon Freyr Gunnarsson
Sheet
Detail H
Title
Grimoire (SETS)
Scale
N.T.S.
Rev
2026-10