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

# AI application lifecycle

> How the Cloud APIs support building, evaluating, deploying, running, operating, and improving AI applications.

An AI application evolves through changes to source, agent instructions, model choices, tools, and knowledge. Veryfront Cloud provides shared resources and execution infrastructure for developing those changes and operating the resulting application.

**Build → Evaluate → Deploy → Run → Operate → Improve**

These stages form an iteration loop. Development and evaluation also involve runs; the Run stage here describes using the released application and its agents or background work.

## Build

A support application starts with project source and an agent definition. Instructions describe its behavior, tools provide actions, and knowledge supplies information to retrieve.

The [Projects API](/docs/cloud/apis/projects) manages source and branches. The [Agents API](/docs/cloud/apis/agents) manages definitions and capabilities, while the [Knowledge API](/docs/cloud/apis/knowledge), [Integrations API](/docs/cloud/apis/integrations), and [AI Gateway API](/docs/cloud/apis/ai-gateway) provide retrieval, external access, and model requests.

## Evaluate

An evaluation exercises an agent against defined cases and metrics. For the support agent, cases might represent the questions and failure conditions the application needs to handle.

The [Evaluations API](/docs/cloud/apis/evaluations) separates definitions from evaluation runs and their reports. Results help you assess a change before release. Your release process determines which results are acceptable and how evaluation participates in approval.

## Deploy

A release identifies project content. A deployment assigns that release to an environment with its own configuration, such as variables and domains.

The [Deployment API](/docs/cloud/apis/deployment) manages this transition. A deployment record identifies the selected version and destination; runtime readiness establishes whether the application is available.

## Run

The application now handles requests and starts agent, task, or workflow executions. A support question can start an agent run in a conversation, while a schedule can start a daily-summary task without a chat interaction.

The [Execution API](/docs/cloud/apis/execution) records status, events, output, and related runs. [Sandbox API](/docs/cloud/apis/sandbox) sessions provide isolated command and file environments when work needs a workspace. The [Cache API](/docs/cloud/apis/cache) supports reuse of values across requests.

## Operate

Operation includes diagnosing failures, monitoring performance, controlling access, and understanding resource consumption.

The [Observability API](/docs/cloud/apis/observability) supplies logs, metrics, and traces. The [Identity and Access API](/docs/cloud/apis/identity-and-access) manages memberships and credentials, while the [Billing and Usage API](/docs/cloud/apis/billing-and-usage) exposes consumption and spending controls.

## Improve

Suppose the support agent gives an incomplete answer. Its run history and traces can help identify the relevant instructions, retrieval, or tool behavior. You can revise the implementation and add an evaluation case that exercises the failure, then assess the change before deploying another release.

Cloud supplies the source, evaluation, execution, and operational records for this loop. You connect those capabilities through your development process, Studio, scripts, or CI tools; observing a production failure does not automatically create an evaluation or change the agent.

[Cloud architecture](/docs/cloud/getting-started/cloud-architecture) explains which responsibilities belong to the control plane and runtimes. [Configure a project](/docs/cloud/projects-and-configuration) is a starting point for implementing the lifecycle.


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