> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ewake.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Bring Your Own Cloud

> How Ewake runs in your own AWS account: what the Terraform creates, where models run, what crosses the network, and who does what.

With Bring Your Own Cloud (BYOC), Ewake runs in an AWS account you own. You deploy it with Ewake's open Terraform, and you can read every resource it creates before you apply it.

Your data, your integration credentials and your Bedrock usage stay in your account. Ewake has no access to the account and receives no telemetry from it.

***

## Architecture

Everything below runs in your account, in `eu-west-3` (Paris), the one supported region.

### Networking

By default the Terraform creates a VPC (`10.10.0.0/16`) across two availability zones, with public and private subnets, a NAT gateway, and private endpoints for SSM, Secrets Manager, ECR and CloudWatch Logs.

You can deploy into a VPC you already run instead. You then provide two private subnets in different availability zones and outbound access, through your own route or a NAT gateway the Terraform adds.

The VPC must resolve private hosted zones through the Amazon-provided DNS resolver, because the agent reaches the dashboard service by a private name. If your VPC uses its own DNS servers or Resolver forwarding rules, read [BYOC troubleshooting](https://docs.ewake.ai/architecture/byoc-troubleshooting.md).

### Compute and storage

The deployment uses these AWS services, all in your account:

| Service | Used for |
| - | - |
| ECS on Fargate, behind an Application Load Balancer | The dashboard and API |
| Lambda, SQS and EventBridge Scheduler | Investigations and scheduled agents |
| RDS for PostgreSQL | The application database. It can be an instance you already run |
| EC2 with an encrypted EBS volume | The knowledge graph |
| Secrets Manager | Application secrets and integration credentials |

Your security policies must allow these services in the account.

### AI models

The agent calls Amazon Bedrock in your account, so prompts and model outputs never leave AWS, and Bedrock usage is billed to you.

* Your account owner enables the models and accepts their Marketplace terms. Ewake gives you the current list.
* You can choose the Bedrock model behind each of the agent's model tiers.
* The default models are EU cross-region inference profiles, which Bedrock can serve from another EU region.

### Access control

* **Users** sign in through your own OIDC identity provider: Okta, Entra ID, Google Workspace, Auth0, or any compliant provider. Until you connect one, the dashboard shows a username and password form, for an admin password the Terraform generates into your Secrets Manager.
* **The first user** from your company's email domain to sign in becomes the administrator, and grants roles to the others.
* **Ewake** has no access to your AWS account. Ewake staff can sign in to the dashboard only if your identity provider lets them. If you want Ewake to help operate the deployment, you grant that access yourself, with the permissions you choose, and you can remove it at any time.

### Telemetry

Nothing is sent to Ewake. Application logs go to CloudWatch Logs in your account.

### Data security

* The database and the knowledge graph are encrypted at rest, and backed up daily in your account.
* The load balancer accepts TLS 1.2 and 1.3 only.
* Terraform state holds credentials, including SSO client secrets. Keep it in an encrypted, access-restricted bucket.

### Network security

#### Inbound

Only these services call the deployment from the internet:

* Slack
* Datadog
* incident.io
* FireHydrant
* Sentry
* GitHub
* Your CI
* Any tool you connect with a generic webhook

The dashboard can stay private.

***

## Deployment

1. Prepare a dedicated AWS account in `eu-west-3`: a delegated Route53 zone, Bedrock model access, and a Lambda memory quota of 10240 MB.
2. Send Ewake the account ID. Ewake grants it read access to the container images.
3. Copy `terraform.tfvars.example`, set your company, domain and version, and run `terraform apply`. The first apply takes 15 to 20 minutes.
4. Configure sign-in, then [create your Slack app](https://docs.ewake.ai/get-started/connect-slack.md) and [your GitHub App](https://docs.ewake.ai/integrations/code/github-app-self-hosted.md).

The repository README has the full procedure.

***

## Operations

Ewake publishes each version of the Terraform as a release of the repository. Each release states the application version it needs.

You upgrade when you choose: set the new version and apply. The apply runs database migrations, then rolls the service.

Ewake keeps a limited number of application versions in its registry, so a deployment that stays on an old version long enough can no longer pull it.

***

## Responsibility matrix

| Area | Responsibility | Ewake | You |
| - | - | :-: | :-: |
| **Infrastructure** | Provide the AWS account and its quotas | | ✓ |
| | Write and release the Terraform | ✓ | |
| | Apply it, and own what it creates | | ✓ |
| | Network: VPC choice, inbound access, outbound filtering | | ✓ |
| **Security** | Encryption at rest and in transit, as the Terraform configures it | ✓ | ✓ |
| | Application secrets, integration credentials and Terraform state | | ✓ |
| | Identity provider and sign-in | | ✓ |
| | Roles in the dashboard | | ✓ |
| | Granting and revoking any access for Ewake | | ✓ |
| | Read access to the container images | ✓ | |
| **Monitoring** | AWS audit logs and account security monitoring | | ✓ |
| | Watching the deployment and responding to failures | | ✓ |
| | Support when something fails | ✓ | |
| **Data** | Governance of your source systems and of Ewake's data | | ✓ |
| | Backups, as the Terraform schedules them | ✓ | ✓ |
| | Restores and recovery requirements | | ✓ |
| **AI models** | Enabling models, accepting their terms, and their costs | | ✓ |
| | Choosing the model behind each tier | | ✓ |
| **Software** | Releasing application and Terraform versions | ✓ | |
| | Applying them | | ✓ |
| | Connecting integrations and configuring the agent | | ✓ |
| **Costs** | AWS and Bedrock usage | | ✓ |


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.