Crystalplane Python Slim Runtime Client (crystalplane)

A hermetic, zero-dependency Python client for connecting any AI agent system (LangChain, LlamaIndex, CrewAI, AutoGen, OpenAI, Claude/MCP) to Crystalplane knowledge and skill crystals.

Features

• Zero External Dependencies: Pure Python standard library (urllib.request, hashlib, json).
• Prompt Hydration: Automatically resolves ~~shardID~~ markers in system prompts into verified authoritative skills.
• Dynamic Skill Assembly: Call /api/assemble to compile composite skills within a specified token budget.
• Memory & Crystal Ingestion: Ingest agent traces and learned workflows into durable SHA-256 microchain shards.
• Dual-Layer Caching: In-memory LRU + optional persistent disk cache with SHA-256 node integrity verification.
• Universal MCP Bridge: Run python -m crystalplane.mcp_bridge to expose Crystalplane tools to Claude Desktop, Cursor, or Zed.

Quickstart

from crystalplane import CrystalClient

# Automatically reads CRYSTAL_SERVER_URL and CRYSTAL_API_KEY from environment
client = CrystalClient()

# 1. Hydrate a prompt template with authoritative skill shards
prompt = client.hydrate("""
You are Picard's Chief Architect. Adhere to our platform invariants:
~~sh_core_directive_mathematical_fidelity~~

Task: {task}
""", task="Verify the Reed-Solomon polynomial codec.")

# 2. Search knowledge crystals
results = client.search(query="distributed database locking", limit=5)

# 3. Dynamic context assembly within token budget
assembled = client.assemble(query="motor safety", max_tokens=2048)
print(assembled.prompt_injection)