DEV Community

Cover image for How I added memory export and import to my open-source AI library published: true
Aftab Bashir
Aftab Bashir

Posted on

How I added memory export and import to my open-source AI library published: true

#dotnet #blazor #ai #openai

When I built BlazorMemory, I knew from the start that storing memories in the browser had one obvious problem. What happens when you clear your browser data? Or switch devices? Everything is gone.

v0.3.0 fixes that with two new methods: ExportAsync and ImportAsync.

What it does

Export serialises all memories for a user to JSON:

var json = await memory.ExportAsync(userId);
Enter fullscreen mode Exit fullscreen mode

The output looks like this:

{
  "userId": "demo_user",
  "exportedAt": "2025-03-15T10:00:00Z",
  "version": "1.0",
  "memories": [
    {
      "id": "abc123",
      "content": "User is a senior .NET engineer",
      "learnedAt": "2025-03-14T09:30:00Z"
    }
  ]
}
Enter fullscreen mode Exit fullscreen mode

Notice embeddings are not in the export. They are large (1,536 floats per memory) and they are model-specific. If you exported with text-embedding-3-small and later imported using a different model, the similarity search would break. So the export skips them, and import re-generates them fresh.

Import reads the JSON, skips any memory whose content already exists (to avoid duplicates), and stores the rest:

await memory.ImportAsync(userId, json);
Enter fullscreen mode Exit fullscreen mode

It is built into the MemoryPanel component

If you use BlazorMemory.Components, you get Export and Import buttons in the panel footer with no extra code:

<MemoryPanel UserId="@userId" IsOpen="true" />
Enter fullscreen mode Exit fullscreen mode

Export triggers a browser file download. Import opens a file picker that accepts .json files.

Both buttons are opt-in via parameters if you want to hide them:

<MemoryPanel UserId="@userId"
             AllowExport="false"
             AllowImport="false" />
Enter fullscreen mode Exit fullscreen mode

Why not include the embeddings

I thought about this. Including embeddings would make the import faster since you skip the re-embedding API call. But it creates two problems.

First, file size. A user with 50 memories using text-embedding-3-small would have a 300KB export file. Not terrible, but not great either.

Second, model coupling. If you export from an app using OpenAI embeddings and import into an app using a different provider, the vectors are incompatible and similarity search silently breaks. Re-generating on import keeps things clean regardless of which embedding provider the target app uses.

Getting started 

dotnet add package BlazorMemory
dotnet add package BlazorMemory.Storage.IndexedDb
dotnet add package BlazorMemory.Embeddings.OpenAi
dotnet add package BlazorMemory.Extractor.OpenAi
dotnet add package BlazorMemory.Components
Enter fullscreen mode Exit fullscreen mode 
builder.Services
    .AddBlazorMemory()
    .UseIndexedDbStorage()
    .UseOpenAiEmbeddings(apiKey)
    .UseOpenAiExtractor(apiKey);
Enter fullscreen mode Exit fullscreen mode 
<MemoryPanel UserId="@userId" IsOpen="true" />
Enter fullscreen mode Exit fullscreen mode

That is all you need. The panel handles display, delete, clear, export, and import.

GitHub: https://github.com/aftabkh4n/BlazorMemory

Top comments (0)