Memory API: Iterate and Refine Existing Brand Generations

Every successful generation job in Nomiq produces a Memory Node — a persisted context object that captures the prompt, the derived style seed, and the structural decisions the AI Engine made during that run. By passing a Memory Node’s ID into a subsequent generation call, you instruct the engine to load the prior stylistic context and treat your new prompt as a refinement instruction rather than a fresh creative brief. This is the primary mechanism for iterative brand work: adjusting colors while keeping the logo form, tightening voice while preserving visual identity, or generating multiple client-facing variations from a single approved root.

How Memory IDs Are Returned

Memory Nodes are created automatically at the end of every successful generation job. You do not call a separate endpoint to create them. The memory_id for a completed job is returned inside the generation.success webhook payload under data.memory_id:

{
  "event_id": "evt_5k6l7m8n",
  "type": "generation.success",
  "created_at": 1677649428,
  "data": {
    "job_id": "job_9x8y7z6w",
    "project_id": "proj_1A2b3C4d5E",
    "memory_id": "mem_X1y2Z3w4",
    "assets": {
      "logo_svg": "<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'>...</svg>",
      "primary_color": "#4338ca",
      "secondary_color": "#faf7f0",
      "font_family": "Inter"
    }
  }
}

Store the memory_id value (mem_X1y2Z3w4 in this example) in your database alongside the job record. You will pass it back to the Generation API in any subsequent refinement calls.

Iterating with a Memory ID

To refine an existing generation, include the memory_id parameter in the body of a new POST /v1/brands/generate request. The new prompt should describe only the change you want — you do not need to repeat the entire original brief.

The `memory_id` Parameter on Generate

memory_id

string

The unique identifier of an existing Memory Node (prefix mem_) from a previous successful generation. When provided, the AI Engine loads the referenced stylistic context and applies the new prompt as an incremental refinement. The resulting job produces a new Memory Node, allowing you to chain refinements in sequence. If omitted, the generation starts from scratch using only the provided prompt.

The following example shows a two-step workflow. The first call creates the initial brand identity for a coffee shop. The second call references the resulting Memory Node to adjust only the typography, while preserving the indigo color palette and logo form established in step one. Step 1 — Initial generation:

curl -X POST "https://api.nomiq.com/v1/brands/generate" \
  -H "Authorization: Bearer sk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "project_id": "proj_1A2b3C4d5E",
    "prompt": "A modern coffee shop in Brooklyn that specializes in cold brew. Minimalistic cold brew glass logo, deep indigo and warm cream palette, clean geometric sans-serif typography.",
    "webhook_url": "https://yourapp.com/api/nomiq/webhook"
  }'

// Step 1: dispatch the initial generation
const initialResponse = await fetch("https://api.nomiq.com/v1/brands/generate", {
  method: "POST",
  headers: {
    "Authorization": "Bearer sk_live_xxx",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    project_id: "proj_1A2b3C4d5E",
    prompt:
      "A modern coffee shop in Brooklyn that specializes in cold brew. " +
      "Minimalistic cold brew glass logo, deep indigo and warm cream palette, " +
      "clean geometric sans-serif typography.",
    webhook_url: "https://yourapp.com/api/nomiq/webhook",
  }),
});

const initialJob = await initialResponse.json();
// Save initialJob.id — you will receive memory_id via webhook

import requests

# Step 1: dispatch the initial generation
initial_response = requests.post(
    "https://api.nomiq.com/v1/brands/generate",
    headers={
        "Authorization": "Bearer sk_live_xxx",
        "Content-Type": "application/json",
    },
    json={
        "project_id": "proj_1A2b3C4d5E",
        "prompt": (
            "A modern coffee shop in Brooklyn that specializes in cold brew. "
            "Minimalistic cold brew glass logo, deep indigo and warm cream palette, "
            "clean geometric sans-serif typography."
        ),
        "webhook_url": "https://yourapp.com/api/nomiq/webhook",
    },
)

initial_job = initial_response.json()
# Save initial_job["id"] — you will receive memory_id via webhook

After the generation.success webhook fires, you receive memory_id: "mem_X1y2Z3w4". Now use that ID to refine the typography in a second call: Step 2 — Refine typography using the Memory Node:

curl -X POST "https://api.nomiq.com/v1/brands/generate" \
  -H "Authorization: Bearer sk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "project_id": "proj_1A2b3C4d5E",
    "prompt": "Change the typography to a humanist serif. Keep the logo and color palette exactly as they are.",
    "memory_id": "mem_X1y2Z3w4",
    "webhook_url": "https://yourapp.com/api/nomiq/webhook"
  }'

// Step 2: refine typography — preserve logo and palette via memory_id
const refinedResponse = await fetch("https://api.nomiq.com/v1/brands/generate", {
  method: "POST",
  headers: {
    "Authorization": "Bearer sk_live_xxx",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    project_id: "proj_1A2b3C4d5E",
    prompt:
      "Change the typography to a humanist serif. " +
      "Keep the logo and color palette exactly as they are.",
    memory_id: "mem_X1y2Z3w4", // ID from the generation.success webhook
    webhook_url: "https://yourapp.com/api/nomiq/webhook",
  }),
});

const refinedJob = await refinedResponse.json();
// refinedJob.status === "queued"
// A new memory_id will be returned in the next generation.success webhook

import requests

# Step 2: refine typography — preserve logo and palette via memory_id
refined_response = requests.post(
    "https://api.nomiq.com/v1/brands/generate",
    headers={
        "Authorization": "Bearer sk_live_xxx",
        "Content-Type": "application/json",
    },
    json={
        "project_id": "proj_1A2b3C4d5E",
        "prompt": (
            "Change the typography to a humanist serif. "
            "Keep the logo and color palette exactly as they are."
        ),
        "memory_id": "mem_X1y2Z3w4",  # ID from the generation.success webhook
        "webhook_url": "https://yourapp.com/api/nomiq/webhook",
    },
)

refined_job = refined_response.json()
# refined_job["status"] == "queued"
# A new memory_id will be returned in the next generation.success webhook

Each generation call that includes a memory_id produces its own new Memory Node. This means you can chain refinements in a linear sequence — or branch multiple times from the same parent node to produce parallel variations.

Use Cases

Memory is most valuable when you need to make targeted adjustments without disrupting everything the AI Engine established in a prior run. Common patterns include:

Goal	Prompt Strategy
Refine colors while keeping the logo	`"Shift the primary color from indigo to forest green. Preserve the logo form."`
Adjust brand voice while keeping visuals	`"Rewrite the tagline and font pairing to feel more playful. Keep all visual assets."`
Generate light and dark mode variants	`"Produce a dark-mode version of the palette. Logo and typography stay the same."`
Create size-specific logo variants	`"Generate a compact wordmark variant suitable for 32×32 px favicons."`

Memory is especially powerful in client revision workflows. When a client approves a logo but requests color changes, branch a new generation from the approved Memory Node rather than re-running from scratch. This preserves billable approval milestones as discrete nodes in the graph and gives you a clean audit trail of every revision. Because Nomiq’s Memory graph is acyclic, you can also dispatch multiple parallel branches from the same approved node — for example, three color palette variations — and present them simultaneously to the client for a single round of feedback.

Listing Memory Nodes for a Project

GET https://api.nomiq.com/v1/projects/{project_id}/memory Returns a chronologically ordered list of all Memory Nodes belonging to a project, most recent first. Use this endpoint to inspect the full revision history of a project or to look up a memory_id you did not store at webhook delivery time.

project_id

string

required

The unique identifier of the project whose Memory Nodes you want to list.

limit

integer

Maximum number of Memory Nodes to return. Defaults to 10, maximum 100.

starting_after

string

A Memory Node ID cursor for pagination. Returns the page of results after this node.

curl -X GET "https://api.nomiq.com/v1/projects/proj_1A2b3C4d5E/memory?limit=5" \
  -H "Authorization: Bearer sk_live_xxx"

const response = await fetch(
  "https://api.nomiq.com/v1/projects/proj_1A2b3C4d5E/memory?limit=5",
  {
    headers: { Authorization: "Bearer sk_live_xxx" },
  }
);

const history = await response.json();
history.data.forEach((node) => {
  console.log(node.id, node.prompt);
});

import requests

response = requests.get(
    "https://api.nomiq.com/v1/projects/proj_1A2b3C4d5E/memory",
    headers={"Authorization": "Bearer sk_live_xxx"},
    params={"limit": 5},
)

history = response.json()
for node in history["data"]:
    print(node["id"], node["prompt"])

{
  "object": "list",
  "data": [
    {
      "id": "mem_L9m8N7o6",
      "object": "memory_node",
      "prompt": "Change the typography to a humanist serif. Keep the logo and color palette exactly as they are.",
      "parent_id": "mem_X1y2Z3w4",
      "created_at": 1677649500
    },
    {
      "id": "mem_X1y2Z3w4",
      "object": "memory_node",
      "prompt": "A modern coffee shop in Brooklyn that specializes in cold brew. Minimalistic cold brew glass logo, deep indigo and warm cream palette, clean geometric sans-serif typography.",
      "parent_id": null,
      "created_at": 1677649420
    }
  ],
  "has_more": false,
  "next_cursor": null
}

Next Steps

Generation API

Dispatch brand generation jobs and pass memory_id to iterate on existing outputs.

Prompt Engineering

Learn how to write effective refinement prompts that work well with Memory Nodes.

​How Memory IDs Are Returned

​Iterating with a Memory ID

​The memory_id Parameter on Generate

​Example: Iterative Refinement

​Use Cases

​Listing Memory Nodes for a Project

​Next Steps

Generation API

Prompt Engineering

How Memory IDs Are Returned

Iterating with a Memory ID

The `memory_id` Parameter on Generate

Example: Iterative Refinement

Use Cases

Listing Memory Nodes for a Project

Next Steps