ADO Operations→

Scribble currently supports Azure-based devops and our internal custom devops process. In theory this could be expanded to support AWS/GCP DevOps as well.

DevOps Code→

The devops code has been written to help:

Deploy pipeline and LLM code that evolves over time
Update platform libraries and services
Update platform configuration files such as siteconf
Configure common services such as nginx, supervisor
Check the health and clean the system

The Azure-ADO deployment code is a combination of yaml templates and scripts. These scripts inturn call a comprehensive set of management scripts available in $HOME/enrich/manage. These scripts handle the application packaging, dependency management, version metadata management, and any error cases.

The codebase is stored in the customer's repo to enable auditing.

deploy
├── azure-ado
│   ├── common
│   │   ├── cleanup.yaml
│   │   ├── deploy-all-customers.yaml
│   │   ├── deploy-all-llms.yaml
│   │   ├── deploy-build.yaml
│   │   ├── deploy-customer.yaml
│   │   ├── deploy-llm.yaml
│   │   ├── deploy-routine.yaml
│   │   ├── deploy-service.yaml
│   │   ├── health.yaml
│   │   ├── install-env.yaml
│   │   └── install-files.yaml
│   ├── scribble-build
│   │   ├── cleanup.yaml
│   │   ├── deploy-all-customers.yaml
│   │   ├── deploy-all-llms.yaml
│   │   ├── deploy-build.yaml
│   │   ├── deploy-customer.yaml
│   │   ├── deploy-llm.yaml
│   │   ├── deploy-routine.yaml
│   │   ├── deploy-service.yaml
│   │   ├── health.yaml
│   │   ├── install-env.yaml
│   │   └── install-files.yaml
│   ├── Customer
│   │   ├── cleanup.yaml
│   │   ├── deploy-all-customers.yaml
│   │   ├── deploy-all-llms.yaml
│   │   ├── deploy-build.yaml
│   │   ├── deploy-customer.yaml
│   │   ├── deploy-llm.yaml
│   │   ├── deploy-routine.yaml
│   │   ├── deploy-service.yaml
│   │   ├── health.yaml
│   │   ├── install-env.yaml
│   │   └── install-files.yaml
└── scripts
    ├── archive
    ├── deploy
    │   ├── __init__.py
    │   ├── manage.py
    │   └── utils.py
    └── run-deploy.sh

Pipelines→

Once the organization has been setup and permissions to connect a repository are give, pipelines can be added by the customer. Scribble or IT operations team can trigger the pipelines as required. We expect to have 15-20 pipelines that are common across customers and are relatively stable, and any customer-specific pipelines to reflect the needs of a particular setup.

Pipeline	Category	Summary
deploy-build.yaml	CI	Do a full build to check integration of Customer and Scribble pipelines and services
deploy-customer.yaml	Assistants	Update the customer codebase (scribble's contrib repo is also treated as a customer)
deploy-all-customers.yaml	Assistants	Update all customer repos configured on the machine
deploy-all-llms.yaml	Agents	Update all LLM agents and environment
deploy-llm.yaml	Agents	Update a single LLM agent and associated environment
deploy-routine.yaml	System	Regular upgrade of all platform libraries and services
deploy-service.yaml	System	Update the code base of services such as doodle and datastore
install-env.yaml	Configuration	Update environment variables and site configuration
install-files.yaml	Configuration	Update configuration files of services such as nginx and supervisor
health.yaml	Health	Check every aspect of the system including OS, python, libraries, directory structure, processes
cleanup.yaml	Health	Cleanup temp files, very old runs etc

New Organization

Files→

Filename	Summary
Production	Environment variables
<server>.<env name>.config.json	Version management and deployment preferences
<server>.<env name>.siteconf.jsonx	Credentials and site configuration
<server>.<env name>.django-env	Environment variables

Project Management→

Depending on the project management process, the system will use:

Work Items - Issues and Epics including approvals to deploy, status etc
Wiki - Notes about the deployment
Board - Move issues across the various stages

New Organization