How to Access NotebookLM Via API?

Gemini API
,
23

Jo secher! Was dänksch denn du

24

I developed one, here is the link for introduce the notebooklm api. https://youtu.be/MdFZIsOqtBo

1 Like
25

Here is the sample json, the api is using servicekey instead of oauth2, but I can enable it as well:
{
“Query”: "Greeting People We greet people when we see them. We always greet people we know. We sometimes greet people we don’t know. To greet someone is to say hello in a friendly way. There are many ways to greet someone. What expressions do you use? ",
“UserID”:“Modified by moderator”
“Language”:“English”
}

26

I want connect you for this solution

27

Hi Duy,

Sure. what can I help you with?

Thanks,
Tony

28

We desperately need a NotebookLM API!!

1 Like
29

I would just to use personally

30

Feature Request

I think there should be an API to organize and aggregate blog posts into a knowledge map.

1 Like
32

I had an idea for a project to generate a weekly podcast about technology, and an API for NotebookLM would be helpful. I could make a script or a program to get some articles (perhaps from some RSS feeds), then send those articles to NotebookLM and have it generate a podcast talking about those articles, and send a link to the podcast somewhere. Maybe I’d need to have it upload/transfer the podcast somewhere else that would be accessible to anyone..

33

The issue here with making “normal” calls to it instead of a real API (like Vertex AI with a RAG Engine) is that there is no access to the settings (AFAIK). For example, with a RAG engine I could set the temperature, I could set penalties for using a word too many time in a row, I can set Top-P (the number of choices it chooses from when writing the next word). There are a LOT of things you can set. I would love to just make REST calls against it, if I could set some parameters. The word penalty is a very important one to me. I have found that it finds some word it loves and it just keeps using it. I had over 250 uses of the word “profound” (not a common word), and there was another word that was even worse (I forget what now).

The bottom line is that if you were allowed to, NotebookLM has a lot of settings and as convenient as it is, you could make it even better for your purposes just by tuning a couple numbers. But, it doesn’t seem like they have any intention of letting us do this (without paying for, and setting up a whole Vertex AI + RAG solution). I’ve been told they really gouge you on the Vertex AI + RAG calls as opposed to a plain Gemini call, but I can’t put any numbers to it.

34

I tried this, but no one listened to it. Maybe it’s my fault, I don’t know. It’s called “Startup Multipliers” and I released it on Spotify and YouTube/Youtube Music.

35

Kerem, I would not worry too much about this. Social media stuff takes time.

What I am finding (I have an Ultra subscription to Gemini Pro), is that replicating NotebookLM using the Gemini API (plus your own set of tools at home like Python with langchain and ChromaDB) is quite possible.

But, it is NOT easy. You have have to tune temperature, top_p, top_k, get your source documents correct, get your prompts correct, choose the AI model to use, choose the best text embedding model… and everything affects everything else, so you change one thing and you might find you’ve improved 1 thing, and made 3 other issues worse. So…

If someone has a screen-scraping-based API for NotebookLM, I’d really like to see it.

For me, using NotebookLM but being able to automate the steps and do a little conventional programming in between steps (e.g., format things with regular expressions), would be the best of both worlds.

And, since all the settings and extra features of NotebookLM are not documented, it would be easier to just use it rather than duplicate it. Can anyone help with that?

36

Sincerely, I would love to do a 15 minute overview of a scientific paper together with some references. And automating it so I can hear while I walk to college would be fantastic. I could also decide if I want to read the paper based on the overview. But I would like to reduce friction by selecting the papers from a list of orcidIDs and them doing the overview. An notebookLM API would be perfect.

37

Hi Ana and all the contributors to this discussion:

That’s excellent news and a crucial update! Thank you for sharing.

You are absolutely correct. As of recent developments, Google has indeed released an official NotebookLM Enterprise API within Google Cloud. This is a significant step, as it moves the advanced functionalities of NotebookLM from a standalone application into a programmable, scalable service.

This means that developers and organizations can now leverage the power of NotebookLM’s features—such as content summarization, synthesis, and deep Q&A on source documents—directly within their own applications, workflows, and business processes. This is a game-changer for a variety of use cases, including:

  • Podcasts API (standalone): Companies can build their own podcasts for teams to interact with vast libraries of documents.
  • Custom Knowledge Management Systems: Companies can build their own internal tools for teams to interact with vast libraries of documents.
  • Automated Research & Analysis: Researchers can programmatically analyze large datasets of papers, reports, and articles to extract insights.
  • Educational Platforms: Universities and online learning platforms can integrate NotebookLM’s capabilities to provide students with powerful study aids for their course materials.
  • Legal & Medical Document Review: Professionals can build applications to quickly summarize and find critical information within complex legal cases or patient records.

The existence of this official API is a definitive answer for those asking about programmatic access to NotebookLM. It clarifies that while the public, consumer-facing version of NotebookLM remains a user-interface-driven tool, the enterprise-level functionality is now fully integrated into the Google Cloud ecosystem, offering a robust and supported solution for developers.

Regards,
Vox Hunter

38

It’s indeed a big lack, I’ve now been implementing lookio app when helping companies automate workflows that require high quality knowledge retrieval such as NotebookLM does, but via API which allows to use it in automation tools like n8n, Make, Zapier.

Hope this helps!

39

There are some services like Lookio RAG, Dust, or Super that allow to get knowledge retrieval via API. Maybe that will help?

1 Like
40

Thanks for sharing, was looking for something exactly like Lookio to automate answers to my PDFs as I’m using Make to automate - great discovery!

1 Like