Instructions to use imjj/touchdesigner-rag-wiki-index with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- sentence-transformers
How to use imjj/touchdesigner-rag-wiki-index with sentence-transformers:
from sentence_transformers import SentenceTransformer model = SentenceTransformer("imjj/touchdesigner-rag-wiki-index") sentences = [ "The weather is lovely today.", "It's so sunny outside!", "He drove to the stadium." ] embeddings = model.encode(sentences) similarities = model.similarity(embeddings, embeddings) print(similarities.shape) # [3, 3] - Notebooks
- Google Colab
- Kaggle
TouchDesigner RAG Wiki Index
This repository contains precompiled FAISS vector indexes and metadata files built from the offline TouchDesigner wiki (Samples/Learn/OfflineHelp).
It is intended for use in local Retrieval-Augmented Generation (RAG) systems to assist in answering TouchDesigner-related questions.
Files
td_index_mini_flat_l2.faiss- FAISS index generated using all-MiniLM-L6-v2 embeddings (Flat L2).
td_metadata_mini_flat_l2.json- Metadata corresponding to
td_index_mini_flat_l2.faiss.
- Metadata corresponding to
td_index_mini_hnsw_l2.faiss- FAISS index generated using all-MiniLM-L6-v2 embeddings (HNSW L2).
td_metadata_mini_hnsw_l2.json- Metadata corresponding to
td_index_mini_hnsw.faiss.
- Metadata corresponding to
td_index_mpnet_flat_l2.faiss- FAISS index generated using all-mpnet-base-v2 embeddings (Flat L2).
td_metadata_mpnet_flat_l2.json- Metadata corresponding to
td_index_mpnet_flat_l2.faiss.
- Metadata corresponding to
Usage
You can load the FAISS indexes with faiss.read_index and match queries using your own embedding model or retrieval pipeline.
The preprocessing, embedding, and search scripts are available at:
GitHub - TouchDesigner RAG
The retrieval system is typically used alongside a local LLM (e.g., via Ollama).
You can also try the online semantic search demo (FAISS retrieval only) here:
TouchDesigner Wiki Search (Hugging Face Space)
Notes
- The offline wiki documentation was automatically chunked without extensive manual cleaning.
Some redundant, outdated, or imperfect text segments may exist. - You are encouraged to refine the data, add external sources, or re-train indexes as needed.
- A more curated version may be published in future updates.
License
This project is licensed under the MIT License.
Disclaimer:
The offline TouchDesigner documentation remains the property of Derivative Inc. and is used here purely for educational and research purposes.