slap-sdk

Official Python client for the Semantic Latent-Algebraic Protocol (SLAP).

A modern, high-performance alternative to traditional RAG architectures (Pinecone / Chroma / LangChain) that operates directly in continuous hyperspherical latent space ($S^{d-1}$), achieving 96%+ KV Cache context reduction and sub-millisecond retrieval.

⚡ Installation

pip install slap-sdk

🚀 Quickstart

1. Connect to the Live Global Cluster

from slap_sdk import SlapClient

# Automatically connects to the global production cluster
client = SlapClient(base_url="https://slap-9m29.onrender.com")

# Health check
print(client.health())

2. Ingest Data (Cloud Storage + Compression)

resp = client.ingest(
    text="Kubernetes orchestrates containerized services with declarative manifests.",
    metadata={"service": "infra"}
)
print("Ingested Doc ID:", resp.doc_id)
print("INT8 Compression SNR:", resp.compression_stats["snr_db"])

3. Retrieve Context with 96% Token Reduction

results = client.retrieve(query="container orchestration", top_k=3)

for match in results.matches:
    print(f"Similarity: {match.cosine_similarity:.4f} | Text: {match.text}")

print("KV Cache Savings:", results.token_savings["context_reduction_percent"], "%")

4. Continuous Unit Hypersphere Embeddings

embedding = client.embed(["Microservices architecture and event-driven design"])
print("Dimension:", embedding["dimension"])
print("Vector (first 5 dims):", embedding["data"][0]["embedding"][:5])

🔗 LlamaIndex Integration

from slap_sdk import SlapLlamaIndexRetriever

retriever = SlapLlamaIndexRetriever(base_url="https://slap-9m29.onrender.com")
nodes = retriever.retrieve("distributed systems architecture")

for n in nodes:
    print(n.score, n.node.text)

📄 License

Apache-2.0