Deploy your prompt

Ship your prompts from Adaline to your AI applications in real time

Deploying a prompt makes it available to your AI applications via the Adaline API or SDK. You can deploy directly to any deployment environment — and promote prompts across environments as they progress through your pipeline.

Deploy to an environment

The deployment environment view is the central space for managing deployments. It is divided into three zones:

The banner at the bottom shows the deployment direction (e.g., Editor -> Production). This is a visual indicator of what you are about to deploy — it is not clickable.

Zone	What it shows
Left	Deployment history — all previously deployed prompt versions across environments. From here you can rollback to any previous version.
Center	The diff view — blue-highlighted content shows what will change when you deploy. This includes changes to config (model and settings), messages, role, modality (text, image, PDF), and tools.
Right	Environment selector, webhooks, and API/SDK integration code for the deployed prompt.

Confirm the deployment

When you are ready, click Deploy to push the prompt to the selected environment:

After deployment completes, Adaline shows the deployed state:

The right panel now displays:

• Integration IDs — Deployment snapshot ID, environment ID, and prompt bench ID for use in your API/SDK calls. Adaline auto-fills the code snippets in the API Integration section.
• Webhook executions — Status of any webhook deliveries triggered by the deployment.

Deploy to another environment

To promote a prompt from one environment to another (e.g., staging to production), select the prompt and click Deploy to…:

This lets you move prompts forward or backward through your deployment pipeline. For example:

• Staging -> Production — Promote a tested prompt to your live application.
• Production -> Staging — Pull a production prompt back to staging for further testing or iteration.

Before deploying to production, compare the deployments between environments to review exactly what will change.

Access deployed prompts

After deployment, your AI application can access the prompt via:

• API — Use the Adaline REST API with your environment key and deployment IDs.
• SDK — Use the TypeScript or Python SDK to fetch and execute deployed prompts.
• Webhooks — Configure webhooks to receive real-time notifications whenever a deployment is created or rolled back.

What gets deployed

A deployment captures a complete snapshot of your prompt, including:

Component	Description
Config	The selected model and all its parameter settings (temperature, max tokens, response format, etc.).
Messages	All message templates — system, user, assistant, and tool messages.
Variables	All variable definitions and their configured sources.
Tools	Tool definitions, schemas, and request configurations.
MCP servers	Any connected MCP server configurations.

Here is a sample deployment object returned by the API and SDK:

{
  "id": "deploy_abc123",
  "createdAt": 1704067200000,
  "updatedAt": 1704153600000,
  "projectId": "proj_789",
  "promptId": "prompt_abc123",
  "deploymentEnvironmentId": "env_production",
  "prompt": {
    "config": {
      "providerName": "openai",
      "providerId": "provider_abc123",
      "model": "gpt-4o",
      "settings": {
        "temperature": 0.7,
        "maxTokens": 1000,
        "topP": 0.9
      }
    },
    "messages": [
      {
        "role": "system",
        "content": [
          {
            "modality": "text",
            "value": "You are a helpful customer support agent for {{company_name}}. Answer questions about our products and services."
          }
        ]
      },
      {
        "role": "user",
        "content": [
          {
            "modality": "text",
            "value": "{{user_message}}"
          }
        ]
      }
    ],
    "tools": [
      {
        "type": "function",
        "definition": {
          "schema": {
            "name": "search_knowledge_base",
            "description": "Search the company knowledge base for relevant articles",
            "parameters": {
              "type": "object",
              "properties": {
                "query": {
                  "type": "string",
                  "description": "The search query"
                }
              },
              "required": ["query"]
            }
          }
        }
      }
    ],
    "variables": [
      { "name": "company_name", "modality": "text" },
      { "name": "user_message", "modality": "text" }
    ]
  }
}

Deployments return prompt messages with {{variable_name}} placeholders intact. Replace them with runtime values before sending to your AI provider — see Use variables in a prompt for the pattern in TypeScript and Python.

Next steps

Rollback Your Prompt

Revert to a previous deployment instantly.

Compare Your Deployments

Review diffs between deployment versions.