Documentation · Roadmap
Vision & roadmap
IntentCenter targets provider and distributor scale: ISPs, backbone operators, and large infrastructure teams that need throughput, resilience, and operational maturity.
Vision
Build an API-first, multi-region, multi-cloud-deployable platform that is the authoritative intent for network and facility inventory, orchestrates change through safe automation, integrates with BSS/OSS-adjacent and cloud ecosystems where needed, and scales to high cardinality, high read/write API rates, and always-on operations—with plugins, integrations, and policy as first-class concerns.
How clean-room research maps to delivery
| Clean-room theme | Delivery focus |
|---|---|
| Source A — extensibility, jobs, APIs, events | Plugin host, REST/GraphQL/event contracts, automation spine |
| Sources B–D — DCIM/IPAM shapes | Domain model refinements, import/export ergonomics |
| Source E — service/request workflows | Approval and request objects as first-class |
| Source F — monitoring adjacency | Telemetry ingest, correlation, observed vs intended state |
| Sources G–H — facility / SNMP angles | Reporting apps, discovery adapters as optional packs |
| Additional Source 1 — inventory + orchestration + assurance | Closed-loop narratives in later phases (without copying proprietary designs) |
Full notes live under
cleanroom/.
Roadmap phases (summary)
Phase 0 — Foundation (repository)
Traceability to clean-room sections, ADRs for stack and tenancy, contributing and security baseline.
Phase 1 — Core platform skeleton
Identity & tenancy, data model v1 (locations, racks, devices, IPAM core, minimal circuits), public APIs (REST,
GraphQL read, events), OpenAPI artifacts, the in-app AI copilot (tool-grounded chat, page- and
chat–aware next steps, org skills, import and triage assist), LLM
configuration for admins, optional Model Context Protocol (MCP) for external clients at
/mcp, and a growing extensibility story: org-scoped plugins and
UI placements, connectors, optional async job worker
(nims-worker), and sign-in options (local + LDAP / Entra / OIDC) via admin
configuration—see Web UI & platform and the repository README.
Phase 2 — Automation & scale
Deeper job engine (schedules, cross-region brokers), Git-backed and signed automation artifacts, horizontal scale and HA—building on the Phase 1 job and worker hooks already in the tree.
Phase 3 — Provider-grade operations
Bulk at provider volumes, observability and SLOs, multi-cloud reference deploys, integrations without mandating a monolith.
Phase 4 — Maturity & ecosystem
Third-party app certification, DR runbooks, compliance and data residency as apps—not forks.
Non-functional targets (directional)
| Area | Direction |
|---|---|
| Capacity | Millions of inventory objects; sustained API traffic; batch and streaming ingestion. |
| Availability | HA control plane; multi-AZ data tier; clear RPO/RTO per profile. |
| Scalability | Scale-out stateless services; queue workers; partition-friendly keys. |
| Security | Strong authn/z, externalized secrets, encryption in transit and at rest, audit on mutating paths. |
| Deployability | Helm/Kustomize-style examples, IaC samples, air-gapped options. |
| Operability | Metrics, runbooks, feature flags, safe rollouts. |
| Extensibility | Plugins with stable contracts; webhooks; events; custom fields and policy hooks. |
Wishlist (backlog)
For a concrete, design-aligned list of deferred features, partial implementations, and architectural opportunities (auth, extensibility, copilot, MCP, validation, scale), see Wishlist & future work. The same items are tracked as checkboxes in the repository README.md.