EdgeForge AI: How Artificial Intelligence Autonomously Deploys Cloud Infrastructure
- Jun 19
- 5 min read
Updated: 1 day ago
EdgeForge AI: How Artificial Intelligence Autonomously Deploys Cloud Infrastructure
Imagine being able to verbally describe the infrastructure you need—and having it up and running in the space of a meeting.
This isn't science fiction. This is what EdgeForge AI does today, in production, on real OpenStack infrastructure.
The problem that EdgeForge AI solves
Anyone who has managed cloud infrastructure knows the cycle that repeats every time a new environment is needed: open tickets, wait for approvals, write Ansible playbooks, test, fix errors, repeat.
For a full monitoring environment with Grafana, Prometheus, Loki, and two monitored application VMs, an experienced team typically spends 1-2 days of work including design, playbook writing, networking setup, and testing.
With EdgeForge AI, the same environment is deployed by describing what's needed in natural language. The system interprets the request, designs the architecture, generates playbooks, and automatically performs the deployment.
How EdgeForge AI Works
EdgeForge AI is the provisioning module of SentinelForge AI — the on-premises AI platform developed by Epic Edge for open-source cloud infrastructure automation.
The system is based on two AI models that work in parallel according to a dynamic pipeline configured by Epic Edge: one model dedicated to natural language interpretation and one specialized in code generation. The two models operate in coordination, following precise instructions that ensure output consistency and quality.
Deterministic, not random, choices. Unlike general-purpose AI systems that produce variable output, EdgeForge AI is trained to reason deterministically—every architectural choice follows precise and reproducible logic. If the user requests Grafana with external access and Prometheus on the internal network, the system designs the correct networking with dual NICs, not improvises.
Continuous research and learning. If EdgeForge AI isn't familiar with a specific software requested by the user, it searches for it online, analyzes the documentation, and generates the appropriate playbook. The acquired information is integrated into future operations.
Memory and optimization. Successfully generated Ansible playbooks are stored as history. The next time the same software is requested, EdgeForge AI reuses the validated playbook instead of regenerating it from scratch—drastically reducing deployment times. An environment identical to one already running completes in a fraction of the original time, since all playbooks are already in memory and ready to use.
Dynamic templates. Deployment templates are not static but are dynamically generated based on specific user interactions. The same software can be deployed in different configurations depending on the context—standalone, clustered, with or without high availability, with specific networking—and EdgeForge AI adapts the deployment accordingly.
A real-world example: complete monitoring environment
To concretely illustrate the capabilities of EdgeForge AI, here is a real-world deployment running in production.
The natural language request: "I need a complete monitoring environment with Grafana, Prometheus, and Loki, with two application VMs to monitor with node_exporter."
The plan generated by EdgeForge AI — 11 operations on 5 VMs:
grafana-server VM — small, exposed to the internet with dual NIC (external access + internal network to reach other services)
VM prometheus-server — small, on separate internal network
Installing Prometheus on prometheus-server
VM loki-server — small, on separate internal network
Loki installation on loki-server
Installing node_exporter on grafana-server
Installing node_exporter on prometheus-server
VM app1 — small, on internal network
Installing node_exporter on app1
VM app2 — small, on internal network
Installing node_exporter on app2
Resources: 5 VMs — Run time: less than 25 minutes
The execution:
EdgeForge AI performs all 11 operations sequentially — creating VMs on OpenStack, dynamically generating Ansible playbooks for each component, setting up networking, and installing and configuring each service.
The final result:
grafana-server: IP 10.5.50.70
Grafana service configured
URL: https://10.5.50.70 (SSL included)
Door: 443
prometheus-server: IP 192.168.10.67 — completed
loki-server: IP 192.168.10.239 — completednode_exporter: completed on all nodes
app1: IP 192.168.10.81
app2: IP 192.168.10.233
Complete, working environment with SSL configured and login credentials—without a single manually written command.
Supported environments
EdgeForge AI is designed to run on any OpenStack infrastructure. The installation is tailored to each specific environment—tenant, networking, VM flavor, storage—adapting to existing configurations rather than requiring forced standardization.
This means that EdgeForge AI can be deployed on a customer's existing OpenStack infrastructure, managed by Epic Edge or any other provider, without requiring any changes to the underlying infrastructure.
Pre-validated templates available
In addition to custom natural language deployments, EdgeForge AI includes a catalog of pre-validated environments ready for immediate deployment:
Monitoring stack — Grafana + Prometheus + Loki + node_exporter
Kubernetes cluster — multi-master control plane + worker nodes + Calico networking
WordPress HA — High Availability Cluster with Replicated Database
GitLab — full installation with integrated CI/CD runners
AWX — Ansible automation platform
MySQL Galera — multi-master database cluster
JupyterHub — multi-user AI/ML environment
PostgreSQL HA — Highly available database with replication
Each template is dynamically adapted to the specific context of the environment — it is not a static deployment but a starting point that EdgeForge AI customizes based on user requirements.
Because this changes the way we manage infrastructure
Manually provisioning cloud infrastructure has three structural problems that EdgeForge AI eliminates.
Time. Writing correct Ansible playbooks for a complex environment takes hours or days. EdgeForge AI generates and runs the same playbook in less than 25 minutes, even for complex environments—and even faster if the playbooks are already in memory from previous deployments.
Human errors. Manual configuration is prone to errors—an incorrect variable, a missing parameter, an unexpected incompatibility. EdgeForge AI follows a deterministic pipeline that minimizes errors and validates each step before proceeding.
Scalability. An engineering team has limited capacity for manual provisioning. With EdgeForge AI, that same capacity multiplies—more environments in parallel, faster, with fewer human resources dedicated to repetitive tasks.
On-premise LLM: zero data outside the perimeter
A critical aspect for European companies with GDPR and NIS2 compliance requirements: EdgeForge AI operates entirely on-premises. AI models run on local infrastructure, user requests never leave the corporate perimeter, and generated playbooks remain in the system's local memory.
No data is sent to external cloud services. No dependency on third-party APIs for code generation. Full control over the automation process.
The Future: CloudSentinel AI
The next SentinelForge AI module in development is CloudSentinel AI—a predictive monitoring system that analyzes OpenStack, Ceph, and Kubernetes infrastructure in real time to detect anomalies, perform automated root cause analysis, and optimize resource utilization.
EdgeForge AI and CloudSentinel AI will together form a complete AIOps platform—from provisioning automation to intelligent infrastructure monitoring—entirely on-premises and open-source.
Conclusion: Automation that respects complexity
EdgeForge AI isn't a wizard that oversimplifies the complexity of cloud infrastructure. It's a system that understands that complexity—networking, security, high availability, component dependencies—and manages it autonomously.
The result is a tangible change in the way engineering teams work: less time on repetitive configurations, more time on architecture, optimization, and product development.
Comments