ForwardLane Embedding Similarity Search

Choose this endpoint when you need to find semantically similar code functions or classes using natural language queries and vector similarity, rather than keyword or BM25 matching. Prefer it over the BM25+vector hybrid search sibling when you want pure cosine similarity over function embeddings specifically. Best when the user's query is conceptual or descriptive rather than exact-name based.

• Missing required 'q' query parameter returns a 400 validation error
• Invalid model name may cause embedding failure or fallback
• top_k values outside 1-200 range are rejected
• Domain filter that matches no indexed repositories returns empty results
• Payment not provided or insufficient USDC triggers 402 Payment Required