Giselle
Willi Icon

Multi‑Model Composition

Auto-select the best model

Visual Agent Builder

Create agents in minutes

Knowledge Store

Access external data sources

GitHub Icon

GitHub AI Operations

Automates issues, PRs, and deployments with AI

Use Cases

Deep Researcher

AI-powered research and analysis

PRD Generator

Generate product requirements docs

GitHub Icon

Code Reviewer

Automated code review and feedback

Marketing Teams

Doc Updater

Keep documentation up to date

Users

Engineering Teams

AI-Native Startups

Automate workflows, ship faster

Solopreneurs & Fast Builders

Build and launch AI products, solo

Product-Led Engineers

Build, iterate, and ship faster with AI-powered development tools

Tech Writers & DevRel

Self-updating docs, more strategy time

Innovation Teams at Modern Enterprises

Embed AI workflows, scale innovation

Docs
Pricing
Blog
—
Sign UpArrow Icon
Giselle

Product

  • Multi-Model Composition
  • Visual Agent Builder
  • Knowledge Store
  • GitHub AI Operations

Solutions

  • Deep Researcher
  • PRD Generator
  • Code Reviewer
  • Doc Updater
  • AI-Native Startups
  • Solopreneurs & Fast Builders
  • Product-Led Engineers
  • Tech Writers & DevRel
  • Innovation Teams

Resources

  • Blogs
  • Open Source
  • Dictionary

Legal

  • Term
  • Privacy & Cookies

About

  • About Us
  • Contact Us

Build visually, deploy instantly.

© 2026 Giselle
GitHubLinkedInFacebookBlueskyXInstagramYouTube
Giselle

Build visually,
deploy instantly.

Product

  • Multi-Model Composition
  • Visual Agent Builder
  • Knowledge Store
  • GitHub AI Operations

Solutions

  • Deep Researcher
  • PRD Generator
  • Code Reviewer
  • Doc Updater
  • AI-Native Startups
  • Solopreneurs & Fast Builders
  • Product-Led Engineers
  • Tech Writers & DevRel
  • Innovation Teams

Resources

  • Blogs
  • Open Source
  • Dictionary

Legal

  • Term
  • Privacy & Cookies

About

  • About Us
  • Contact Us
© 2026 Giselle
GitHubLinkedInFacebookBlueskyXInstagramYouTube

We want to be clear about how we collect and use cookies so that you can have control over your browsing data.

If you continue to use Giselle, we will assume you are comfortable with our cookie usage.

Column

From OSS Contributor to Team Member at Giselle

PUBLISHEDDECEMBER 26, 2025

Ryo Washizu,
Engineer
From OSS Contributor to Team Member at Giselle

Table of contents

  • Introduction
  • Contributing as an OSS Contributor
  • Joining as a Team Member
  • What Giselle Values

Introduction

Giselle is developed as an open-source project and is publicly available on GitHub. Anyone can contribute through code or documentation. In this article, we introduce one example of a contributor who started with OSS contributions and is now participating in the project as a team member.

Contributing as an OSS Contributor

My first Pull Request was a simple typo fix. It was my first time contributing to an open-source project, and I still remember how nervous I felt when submitting it. After that, I worked on bug fixes and small feature additions.

My first Pull Request - a typo fix My first PR: fixing a typo in the component filename

What stood out to me was how thoughtful and consistent the Pull Request interactions were. Reviews were rarely delayed or left unanswered, and feedback often included clear explanations about design intent and possible improvements. This review process made it easier to understand the project and encouraged continued contributions.

Review comment with detailed feedback A review comment with suggestions for improvement

Joining as a Team Member

After making ongoing contributions as an OSS contributor, I was invited to join the project as a team member. I am currently working on feature improvements and adding new functionality directly related to the product.

Because of my prior experience with the codebase and development workflow through OSS, onboarding was smooth and required little adjustment. The lack of a clear boundary between OSS contributions and production work is one of the defining characteristics of Giselle.

After working on several smaller tasks and getting familiar with the codebase, I took on a more substantial feature: extending the GitHub RAG system.

Extending GitHub RAG System to Support Issues

In developing this feature, having an existing foundation for GitHub RAG built around code and pull requests was a significant advantage. These resources were already managed with metadata, which made it possible to extend the system without starting from scratch.

Connecting a GitHub Vector Store (Issue) to a Vector Query node Connecting a GitHub Vector Store (Issue) to a Vector Query node

However, working with Issues introduced a different set of challenges. Unlike pull requests, Issues do not have a clear point of completion; their state can change repeatedly through open and close cycles. In the existing system, changes to an Issue's status would trigger a full re-embedding of its content, even when only the metadata had changed.

Ideally, when an Issue's status changes, only the relevant metadata should be updated. To address this, I added a mechanism to detect state changes and update only the necessary information. This allowed us to avoid unnecessary re-embedding while still keeping the data up to date.

During this process, I initially worked by splitting the implementation into multiple pull requests. However, as unexpected design constraints surfaced, the required changes accumulated, and the final pull request became larger than intended. In retrospect, it would have been better to first implement and validate the full flow locally, and then break the work into smaller, more focused pull requests. That approach would likely have made the review process smoother.

What Giselle Values

Even small fixes or suggestions are more than welcome as a starting point. By writing code and participating in discussions, contributors naturally build a stronger connection with the project. While this article presents just one example, Giselle will continue to welcome participation in many forms through open and collaborative development.

Last edited onDECEMBER 26, 2025
  1. Top
  2. Arrow Right
  3. Blog
  4. Arrow Right
  5. Column
  6. Arrow Right
  7. From OSS Contributor to Team Member at Giselle
AI agents
/
Human-AI Collaboration
/
Retrieval-Augmented Generation
Prev Arrow
Prev
Node-Based Interfaces for People Who Don't Think in Flowcharts — Building Giselle
Next Arrow
Next
Managing Coding Agents: Zed + Cursor

Try Giselle Free or Get a Demo

Supercharge your LLM insight journey -- from concept to development launch
Get Started - It's Free

Related Insights

The Making of a Stylish Dog-Centered Lifestyle Media using Giselle
Column

The Making of a Stylish Dog-Centered Lifestyle Media using Giselle

Kaori Nakashima,
Founding Designer
Managing Coding Agents: Zed + Cursor
Column

Managing Coding Agents: Zed + Cursor

Satoshi Toyama,
Founding Engineer
What Giselle is not
Column

What Giselle is not

Kaori Nakashima,
Founding Designer