This project started with a simple question: How can we give LLMs full access to the Revit API so they become specialized Revit software developers and interact flawlessly with Revit?
Disclaimer: This project is shared "as-is" as a proof-of-concept and is not actively maintained. It demonstrates a method for enabling Large Language Models (LLMs) like Google Gemini to interact with the Revit API using a Retrieval-Augmented Generation (RAG) system. Feel free to fork, modify, and build upon it, but expect potential bugs or limitations. Use at your own risk.
RevitGeminiRAG is a Revit plugin that allows users to interact with their Revit models using natural language queries processed by Google's Gemini models. It leverages a RAG system to provide the LLM with relevant context from the Revit API documentation, enabling it to understand and respond to Revit-specific requests more effectively.
- User Input: The user enters a prompt via a dedicated Revit panel (implemented in
PromptForm.cs). - RAG Processing: A Python script (
python/generate_rag_prompt.py) preprocesses the prompt and retrieves relevant Revit API information from a pre-built knowledge base. - LLM Interaction: The original prompt and the retrieved context are sent to the Google Gemini API using the
Google.Cloud.AIPlatform.V1library within the C# plugin. - Response Generation: Gemini generates a response based on the prompt and the provided API context.
- Display: The plugin (
RunRAGCommand.cs) receives the response and displays it to the user in the prompt window.
- Current Version: The included RAG knowledge base is built upon the Revit 2025 API documentation. Therefore, the plugin's understanding is currently optimized for Revit 2025.
- Updating for Other Revit Versions: To adapt this plugin for a different Revit version, you will need to rebuild the RAG knowledge base. The general process involves:
- Downloading the Revit SDK for the target version from the Autodesk Developer Network (ADN).
- Locating the API documentation, often provided as a Compiled HTML Help (
.chm) file (e.g.,RevitAPI.chm). - Extracting and cleaning the content from the
.chmfile, potentially converting it to Markdown (.md) or plain text format suitable for processing. - Indexing the cleaned documentation (e.g., splitting into meaningful chunks based on classes, methods, properties).
- Generating vector embeddings for the indexed documentation using an appropriate embedding model (compatible with the retrieval mechanism in the Python script).
- Storing these embeddings in a vector database or a file format that the
python/generate_rag_prompt.pyscript can efficiently query. Note: The specifics of steps 3-6 depend on the implementation within the Python script and the chosen embedding/vector store technologies.
- Natural language interaction with Revit models via a dedicated panel.
- Integration with Google Gemini LLM for response generation.
- Retrieval-Augmented Generation (RAG) system providing Revit 2025 API context to the LLM.
- Autodesk Revit 2025 (Required for full compatibility with the current RAG knowledge base)
- .NET Framework 4.8 (Standard for Revit 2025 plugins)
- Visual Studio (e.g., 2019, 2022) with the ".NET desktop development" workload installed.
- Google Cloud Platform Account and Project: You need an active GCP project to generate an API key.
- A valid Google API Key: This is essential for authenticating requests to the Gemini API.
- (Optional) Python Environment: If modifications to the
python/generate_rag_prompt.pyscript or the RAG generation process are needed, a Python environment with relevant libraries (e.g., for embedding generation, vector stores) will be required.
- Clone the Repository:
git clone https://github.com/ismail-seleit/RevitGeminiRAG cd RevitGeminiRAG - Set Google API Key:
- CRITICAL: You MUST set your Google API Key as a Windows Environment Variable for the plugin to authenticate with Google Cloud.
- Search for "Edit the system environment variables" in the Windows search bar and open it.
- Click the "Environment Variables..." button.
- Under "User variables" (or "System variables" if you prefer), click "New...".
- Variable name:
GOOGLE_API_KEY - Variable value:
<Your_Actual_Google_API_Key> - Click OK on all dialogs.
- Important: Restart Visual Studio and/or Revit if they were open to ensure they pick up the new environment variable.
- Build the Project:
- Open
RevitGeminiRAG.slnin Visual Studio. - Check References: Verify that the Revit API DLLs (
RevitAPI.dll,RevitAPIUI.dll) are correctly referenced. By default, the.csprojmight point to a standard Revit installation path (e.g.,C:\Program Files\Autodesk\Revit 2025\). If your installation is different, you'll need to update the reference paths in Visual Studio (Solution Explorer -> References -> Right-click -> Add Reference -> Browse). - Restore NuGet Packages: Right-click the solution in Solution Explorer and select "Restore NuGet Packages".
- Build: Build the solution (Build > Build Solution). This will compile the code and create the necessary DLLs in the
bin\Debug\orbin\Release\folder (relative to the project directory).
- Open
- Install the Plugin in Revit:
- Locate Add-ins Folder: Navigate to your Revit Add-ins folder for the current user:
%APPDATA%\Autodesk\Revit\Addins\2025(Create the2025folder if it doesn't exist) - Copy Addin Manifest: Copy the
RevitGeminiRAG.addinfile from the root of your cloned repository into this2025folder. - Edit Addin Manifest: Open the copied
RevitGeminiRAG.addinfile in a text editor (like Notepad). - Update Assembly Path: Modify the
<Assembly>tag to point to the full path of theRevitGeminiRAG.dllfile that was generated when you built the project. For example:<Assembly>C:\path\to\your\cloned\repo\RevitGeminiRAG\bin\Debug\RevitGeminiRAG.dll</Assembly>(ReplaceC:\path\to\your\cloned\repo\with the actual path where you cloned the repository and ensure you point to the correct build output folder, e.g.,bin\Debugorbin\Release). - Copy Dependencies: Ensure that all necessary dependency DLLs (like the Google Cloud libraries from the build output folder) are located either in the same directory as
RevitGeminiRAG.dllor in a location where Revit can find them. Copying them to the same folder as the main DLL is usually the simplest approach.
- Locate Add-ins Folder: Navigate to your Revit Add-ins folder for the current user:
- Start Autodesk Revit 2025.
- If the plugin installed correctly, navigate to the "Add-Ins" tab on the Revit ribbon.
- You should find a button related to "RevitGeminiRAG" (likely named based on the
RunRAGCommandclass). Click it. - A window (
PromptForm) will appear. Enter your natural language query about the Revit model or API into the text box. - Click the submit/run button.
- The plugin will communicate with the Gemini API (using your API key and the RAG context) and display the generated response in the window.
Example Prompts:
- "Color all walls thicker than 150mm red in the current view."
- "Apply the 'Architectural Plan' view template to the current view."
- "Export the 'Room Schedule' to an Excel file on my Desktop."
- "Select all doors located on 'Level 1'."
- "Hide all furniture elements in the active 3D view."
- "Create a text note saying 'Please verify dimensions' near the currently selected element."
Demo Video:
- Google.Cloud.AIPlatform.V1
- Google.Apis / Google.Apis.Auth / Google.Apis.Core
- Grpc.* (Core, Net.Client, Auth, etc.)
- Newtonsoft.Json
- (Review
packages.configor the.csprojfile for a complete list)
This project is licensed under the MIT License.