adaline-prompts

Adaline Prompts

Concepts

Prompts are the central artifact in Adaline. Each prompt has a draft (the working version you iterate on), a config (provider, model, and inference settings), messages (system/user/assistant turns), tools (function definitions available to the model), and variables (template placeholders injected at inference time).

Key terms:

• Prompt — a named artifact containing a draft with config, messages, tools, and variables
• Draft — the editable working version of a prompt; changes here do not affect deployed versions
• Config — provider selection, model selection, and inference parameters (temperature, maxTokens, topP, etc.)
• Messages — ordered array of system, user, and assistant turns; content is an array of modality objects
• Tools — function definitions the model can call; each has a name, description, and JSON schema for parameters
• Variables — named placeholders used as {{var_name}} in messages; resolved at inference time

Configuration

Set these environment variables when your Adaline credentials are available:

• ADALINE_API_KEY — your workspace API key (from Settings > API Keys at app.adaline.ai)
• projectId — your project ID (from the dashboard sidebar)

You can start integrating before you have credentials. All code examples use placeholder values — replace them with real values when ready.

Quick Triage

Symptom	First Fix
POST /prompts returns 422	Check that messages content uses the array format with modality and value fields
Variables not substituted at inference	Confirm variable names in messages exactly match the name field in the variables array (case-sensitive)
PATCH returns 404	Verify the prompt ID — copy it from the Adaline editor URL
GET /prompts returns empty list	Confirm the projectId query parameter is present and correct
Draft changes not reflected in deployment	Draft edits never auto-deploy — you must explicitly create a deployment after editing

API Endpoints

Base URL: https://api.adaline.ai/v2

List Prompts

curl -X GET "https://api.adaline.ai/v2/prompts?projectId=PROJECT_ID&page=1&pageSize=20" \
  -H "Authorization: Bearer $ADALINE_API_KEY"

Returns a paginated list of prompts in a project. Each item includes id, title, icon, createdAt, and updatedAt.

Create Prompt

curl -X POST "https://api.adaline.ai/v2/prompts" \
  -H "Authorization: Bearer $ADALINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "projectId": "PROJECT_ID",
    "title": "My Prompt",
    "draft": {
      "config": {
        "provider": "openai",
        "model": "gpt-4o",
        "settings": {
          "temperature": 0.7,
          "maxTokens": 1024
        }
      },
      "messages": [
        {
          "role": "system",
          "content": [{ "modality": "text", "value": "You are a helpful assistant." }]
        },
        {
          "role": "user",
          "content": [{ "modality": "text", "value": "Hello {{name}}" }]
        }
      ],
      "variables": [
        { "name": "name", "modality": "text" }
      ]
    }
  }'

Response includes the created prompt id and full draft object.

Get Prompt

# Basic fetch
curl -X GET "https://api.adaline.ai/v2/prompts/PROMPT_ID" \
  -H "Authorization: Bearer $ADALINE_API_KEY"

# With playground data expanded
curl -X GET "https://api.adaline.ai/v2/prompts/PROMPT_ID?expand=playground" \
  -H "Authorization: Bearer $ADALINE_API_KEY"

Update Prompt

curl -X PATCH "https://api.adaline.ai/v2/prompts/PROMPT_ID" \
  -H "Authorization: Bearer $ADALINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Updated Title",
    "draft": {
      "config": {
        "provider": "anthropic",
        "model": "claude-3-5-sonnet-20241022",
        "settings": {
          "temperature": 0.5,
          "maxTokens": 2048
        }
      },
      "messages": [
        {
          "role": "system",
          "content": [{ "modality": "text", "value": "You are an expert assistant." }]
        }
      ]
    }
  }'

Delete Prompt

curl -X DELETE "https://api.adaline.ai/v2/prompts/PROMPT_ID" \
  -H "Authorization: Bearer $ADALINE_API_KEY"

Returns 204 No Content on success.

Get Prompt Draft

curl -X GET "https://api.adaline.ai/v2/prompts/PROMPT_ID/draft" \
  -H "Authorization: Bearer $ADALINE_API_KEY"

Returns the full draft: config, messages, tools, and variables.

Draft Structure

{
  "config": {
    "provider": "openai",
    "model": "gpt-4o",
    "settings": {
      "temperature": 0.7,
      "maxTokens": 1024,
      "topP": 1.0,
      "frequencyPenalty": 0,
      "presencePenalty": 0
    }
  },
  "messages": [
    {
      "role": "system",
      "content": [{ "modality": "text", "value": "You are a helpful assistant." }]
    },
    {
      "role": "user",
      "content": [{ "modality": "text", "value": "Summarize this: {{document}}" }]
    }
  ],
  "tools": [
    {
      "name": "search_web",
      "description": "Search the web for current information",
      "parameters": {
        "type": "object",
        "properties": {
          "query": { "type": "string", "description": "The search query" }
        },
        "required": ["query"]
      }
    }
  ],
  "variables": [
    { "name": "document", "modality": "text" }
  ]
}

Message Format

Each message has a role and a content array. Content items use a modality field to indicate type:

{
  "role": "user",
  "content": [
    { "modality": "text", "value": "Describe this image." },
    { "modality": "image_url", "value": "https://example.com/image.png" }
  ]
}

Supported roles: "system", "user", "assistant" Supported modalities: "text", "image_url"

Variable Format

Variables declare template placeholders used in messages. Reference them in message text as {{var_name}}.

{ "name": "var_name", "modality": "text" }

Variable names must be alphanumeric with underscores. They are case-sensitive and must exactly match their usage in message content.

Best Practices

1. Create prompts programmatically for automation pipelines — use POST /prompts with a full draft to avoid manual setup in the UI
2. Separate prompt creation from deployment — editing a draft never affects live deployments; always trigger a deployment explicitly after finalizing a draft
3. Store projectId and prompt IDs in environment variables, not hardcoded in source
4. Use variables for all dynamic content instead of string interpolation — this keeps the prompt structure stable and enables Adaline's variable substitution at inference time
5. When iterating, use PATCH /prompts/{id} to update only the fields that changed; you do not need to resend the entire draft

References

See references/api.md for the full REST API reference with all request/response schemas, pagination details, and curl examples.