PyckLM is a code-aware semantic embedding model that returns 512-dimensional L2-normalized float vectors. Ideal for semantic code search, clustering, and similarity — up to 64 texts per batch.
embeddings
semantic-search
code-search
+5
|PyckLM Embeddings gives you direct access to Pyckle's proprietary code-aware embedding model via a single MCP tool. Unlike general-purpose embedding models, PyckLM is trained specifically on code and developer content, making it significantly more accurate for tasks like semantic code search, duplicate detection, clustering similar functions, and building code similarity pipelines.
What you get:
- 512-dimensional L2-normalized float vectors, ready for cosine similarity
- Up to 64 texts per request — batch your inputs for efficiency
- Texts truncated at 512 tokens automatically
- No setup, no model hosting — just call embed_texts and get vectors back
Built for developers who need:
- Semantic search over codebases
- Code clustering and deduplication
- Similarity scoring between functions, files, or snippets
- A fast, lightweight embedding layer for AI-powered dev tools