Learn how the GitHub Vector Store Node in Giselle vectorizes your GitHub repositories, creating a searchable knowledge base that can be queried within your AI workflows.

​
GitHub Vector Store Node in Giselle

The GitHub Vector Store Node is a specialized “Source” node that creates a powerful, searchable database from the contents of a GitHub repository. It works by vectorizing the code and documents within the repository, allowing AI models to perform semantic searches and retrieve relevant context. This process uses OpenAI’s powerful and efficient text-embedding-3-small model to convert your repository’s content into vector embeddings. These vectorized data are then stored securely in Giselle’s dedicated Vector Store. To use this feature, the Giselle GitHub App must be installed on the target GitHub repository.

​
Setting up a GitHub Vector Store Node:

  1. Add the Node:
    • From the toolbar at the bottom of the canvas, click the Source icon (a folder with a link symbol).
    • Select GitHub Vector Store from the pop-up menu to add the node to your workspace.
  2. Configure the Repository:
    • When first added, the node will display a “REQUIRES SETUP” status.
    • Select the node to open its configuration panel on the right.
    • Click the Select a repository dropdown menu. This list will only show repositories that have the Giselle GitHub App installed.
  3. Finalize Setup:
    • Choose the repository you want to use as a knowledge source (e.g., giselles-ai/docs).
    • Once selected, the node on the canvas will update to show the name of the configured repository. The vectorization process for this repository will be initiated or updated in the background.

​
Managing Your Vector Stores

You can add or remove the GitHub repositories that are processed by the GitHub Vector Store Node from your team’s settings page.
  • Navigate to Settings > Team > Vector Stores in your Giselle account.
  • Alternatively, you can click the Set Up GitHub Vector Store link directly from the node’s configuration panel.
  • This page allows you to manage which repositories are indexed and available for your workflows. You can access it here: https://studio.giselles.ai/settings/team/vector-stores.

​
Rate Limits and Considerations

When using the GitHub Vector Store Node, please be aware of GitHub API rate limits. Giselle’s GitHub Vector Store uses GitHub App Installation for GitHub API access, which has a rate limit of approximately 5,000 requests per hour. You may encounter errors when working with large-scale projects. If this occurs, please wait some time before trying again. For large repositories or frequent updates, you may experience rate limiting during the vectorization process. If you encounter errors, simply wait and retry after the rate limit window resets.

​
Usage in Workflows

The GitHub Vector Store Node is designed to be used in combination with a Query Node. It acts as the knowledge base that the Query Node searches through.
  • Connect the Output: The “Output” of the GitHub Vector Store Node should be connected to the input of a Query Node.
  • Perform a Search: The Query Node can then take a user’s question or a dynamic input, search the vectorized repository content for the most relevant information, and pass that information to subsequent nodes (like a Generator Node) for processing, analysis, or summarization.

​
Output of the Node

The GitHub Vector Store Node’s output is a reference to the vectorized data of the selected repository. This output provides the necessary connection for other nodes, like the Query Node, to access and search the repository’s content.