GCP Expert

Expert knowledge base for Google Cloud Platform service behavior, defaults, and operational pitfalls. Contains 863 justified beliefs covering GCE, GCS, Cloud Run, GKE, Cloud SQL, Pub/Sub, Secret Manager, VPC networking, IAM, KMS, and cross-service interactions.

What is this?

This is an External Epistemic Memory (EEM) โ€” a model-agnostic knowledge base that any LLM can use via the reasons CLI or tool calling. Unlike a LoRA or fine-tune, this knowledge is not baked into model weights. It is external, inspectable, correctable, and works with any model.

Stats

Metric Value
Total beliefs 863
Status 863 IN / 0 OUT
Premises (observations) 709
Derived (justified conclusions) 154
Nogoods (contradictions) 0
Retraction rate 0%
Max derivation depth 8

Top Topics

Topic Beliefs
gce 77
vpc 73
gcs 70
cloudrun 67
gke 64
cloud 60
gcp 58
cloudsql 50
iam 43
pubsub 43
secretmanager 43
interconnect 36
kms 35
cloudbuild 30
private 27
dns 24

Domain Coverage

  • GCE: instance lifecycle, machine types, persistent disks, preemptible/spot VMs, metadata, live migration (75 beliefs)
  • GCS: bucket policies, lifecycle rules, versioning, access control, retention, cross-region replication (68 beliefs)
  • Cloud Run: container deployment, autoscaling, concurrency limits, networking, service mesh, cold starts (66 beliefs)
  • GKE: cluster management, node pools, networking, Workload Identity, autopilot vs standard, upgrades (60 beliefs)
  • GCP Cross-Cutting: default configurations, security posture, architectural commitments, cross-service patterns (56 beliefs)
  • Cloud SQL: HA configuration, private networking via VPC peering, backup/PITR, production architecture costs (49 beliefs)
  • VPC & Networking: peering constraints, Private Service Connect, Cloud Interconnect, firewall rules, DNS, subnets (35 beliefs)
  • Secret Manager: rotation patterns (notification-only), access patterns, versioning, production usage (43 beliefs)
  • Pub/Sub: message ordering, dead lettering, exactly-once delivery, subscriptions, push vs pull (42 beliefs)
  • Artifact Registry: container images, language packages, vulnerability scanning, repository management (31 beliefs)
  • KMS/CMEK: key rotation decoupled from re-encryption, data governance control plane, cross-service encryption (32 beliefs)
  • Cloud Build: CI/CD pipelines, triggers, build steps, artifact management (29 beliefs)
  • IAM: policy hierarchy, service accounts, Workload Identity Federation, conditions, org policies (28 beliefs)
  • Cloud Interconnect: dedicated vs partner, VLAN attachments, redundancy requirements (20 beliefs)
  • Memorystore: Redis/Memcached managed instances, HA, networking constraints (20 beliefs)
  • Monitoring & Logging: Cloud Monitoring metrics, Cloud Logging, alerting policies, SLO tracking (34 beliefs)
  • Security: Cloud Armor edge filtering, automatic DDoS protection, CMEK governance, default hardening (10 beliefs)

How to Use

Import into a reasons database

reasons init
reasons import-json network.json

Query beliefs

reasons search "Cloud SQL private networking"
reasons explain cmek-single-control-plane-for-data-governance
reasons show gcp-security-requires-upfront-architectural-commitment

Use as an MCP tool or CLI

Any LLM agent that can call reasons search, reasons show, and reasons explain can use this knowledge base. The agent does not need to be told it is an expert โ€” the knowledge base speaks for itself.

Key Beliefs

Node Summary
cmek-single-control-plane-for-data-governance CMEK key lifecycle serves as the single control plane for data governance across GCP
gcp-security-requires-upfront-architectural-commitment GCP's dual security governance (IAM + CMEK) compounds with cross-layer interactions
secretmanager-rotation-notification-only Secret Manager rotation is notification-only: it sends a Pub/Sub message rather than rotating
vpc-peering-limited-connectivity-model VPC peering is non-transitive, never exchanges IAM policies, and has a 25 peering limit
cloudsql-private-networking-doubly-constrained-by-peering Cloud SQL private IP inherits VPC peering constraints (non-transitivity, 25 peering limit)
cloudsql-production-architecture-requires-triple-investment Production Cloud SQL requires concurrent HA, backup, and private networking investment
cloud-armor-operates-at-edge Cloud Armor filters traffic at the Google Cloud edge before it reaches backends
cloud-armor-auto-ddos-global-external-alb DDoS protection is automatic for global external Application Load Balancers
kms-rotation-decoupled-from-reencryption KMS key rotation creates new versions without re-encrypting existing data
secretmanager-production-access-pattern Production secret access should use the API directly, pin to specific versions

Sources

Built from exploration of GCP documentation, API behavior, and operational experience across GCE, GCS, Cloud Run, GKE, Cloud SQL, Pub/Sub, Secret Manager, VPC, IAM, KMS, and 15+ additional GCP services.

Files

File Description
network.json Full belief network (machine-readable, portable)
reasons.db SQLite database (gitignored, regenerate with reasons import-json network.json)
CLAUDE.md Agent instructions for using this knowledge base
entries/ 114 exploration entries โ€” raw observations behind the premises

Quality

  • All 863 beliefs are IN (none retracted)
  • 709 premises grounded in direct observations of GCP service behavior
  • 154 derived beliefs justified from premises via SL justifications
  • 0 nogoods โ€” no contradictions detected
  • Max derivation depth of 8, indicating multi-step reasoning chains
  • Built and reviewed using ftl-reasons derive and review-beliefs pipeline

Limitations

  • Focused on GCP service behavior and defaults as of mid-2026
  • GCP services evolve; some beliefs may become stale as features change
  • Heavier coverage of GCE, GCS, and Cloud Run than other services
  • Does not cover pricing in detail beyond capacity/commitment mechanics
  • No ATMS or assumption-based beliefs (single-context TMS only)

Authors

License

mit

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support