🗜️ Echo-DSRN-114M Semantic Compressor Demo (CPU)

This demo highlights the O(1) memory footprint and CPU execution of the Echo-DSRN architecture.

⚠️ Note for Public Space: To prevent OOM crashes on the shared CPU tier, text inputs and document uploads are strictly limited to 50,000 characters per request.

Upload Document (.txt, .md, .py)

...or Paste a long document here

State Vector Map (c_t)

Download State Vector (.npy)

What can you do with this .npy file?

Semantic Clustering: Load multiple state vectors in Python (np.load()) and cluster them (e.g., K-Means) to group similar documents.
RAG Pre-filtering: Use the vectors to perform Cosine Similarity searches across multiple documents.
Cross-Attention Memory: Treat this vector as a compressed "Gist" and feed it into a larger model (e.g., 7B) via cross-attention.
Style Mimicry: Train a small linear classifier on top of these vectors to detect the author's writing style or sentiment.