dchub 0.2.0

Python SDK + LangChain/smolagents tools for the DC Hub data-center & energy intelligence API.

dchub — Python SDK for the DC Hub API

Live data-center + energy intelligence for AI agents. Query DC Hub instead of guessing from stale training data: 21,000+ facilities across 170+ countries, 300+ markets (DC Hub Power Index / DCPI), real-time ISO grid telemetry, fiber routes, gas pipelines, per-facility tenants, and 2,000+ tracked M&A deals.

The free tier needs no key (10 calls/day). A free X-API-Key raises limits and unlocks key-gated endpoints — mint one at https://dchub.cloud/playground or via the MCP claim_free_key tool.

Install

pip install dchub                 # core (requests only)
pip install "dchub[langchain]"    # + LangChain StructuredTools
pip install "dchub[agents]"       # + smolagents Tools

Quickstart

from dchub import DCHub

dc = DCHub()                              # free tier, no key
# dc = DCHub(api_key="dck_...")           # higher limits / key-gated endpoints

rank = dc.get_market_dcpi_rank("phoenix")
print(rank["composite_score"], rank["verdict"])   # e.g. 21.3 AVOID  (free per-market)

dc.search_facilities(country="US", limit=5)       # GET /api/v1/facilities
dc.stats()                                        # facility counts + sources
dc.ai_capacity_index(horizon=90)                  # AI Capacity Index markets
dc.get_grid_intelligence("PJM")                   # real-time ISO grid intel
dc.get_changes(since="24h")                        # incremental diffs

Every method returns parsed JSON (a dict). Non-2xx responses raise DCHubError, which carries .status_code and the API's self-correcting .detail (e.g. how to claim a free key).

Generic, manifest-driven call

Any of the 45 REST-backed tools can be called by name; path placeholders are filled from kwargs, the rest become the query string:

dc.call("get_market_dcpi_rank", slug="phoenix")
dc.call("get_grid_intelligence", iso="ERCOT")
dc.call("search_facilities", provider="Equinix", limit=3)

The tool→endpoint map is embedded (dchub.TOOL_MANIFEST), so the SDK has no runtime dependency on the live manifest endpoint.

Flagship methods

Method	REST endpoint
stats()	/api/v1/stats
search_facilities(**filters)	/api/v1/facilities
get_facility(slug)	/api/v1/facilities/<slug>
fetch_facility(id)	/api/v1/facility/<id> (numeric id)
get_market_dcpi_rank(market_slug)	/api/v1/dcpi/scores/<slug>
rank_markets(by=...)	/api/v1/mcp/tools/rank_markets
ai_capacity_index(horizon=90)	/api/v1/ai-capacity-index
get_changes(since="24h")	/api/v1/changes/since
get_grid_intelligence(iso)	/api/v1/grid/intelligence/<iso>
list_transactions(**f)	/api/v1/deals
get_news(**f)	/api/news
get_pipeline(**f)	/api/v1/pipeline (needs a free key)

get_facility uses the plural route, which expects a facility slug. For a numeric facility id use fetch_facility(id) (the singular /api/v1/facility/<id> route, exposed as the fetch tool in the manifest).

LangChain

from dchub import dchub_langchain_tools

tools = dchub_langchain_tools()          # list[StructuredTool]
# tools = dchub_langchain_tools(api_key="dck_...")

# e.g. with a LangChain agent:
# from langchain.agents import create_react_agent
# agent = create_react_agent(llm, tools, prompt)

smolagents

from dchub import dchub_smolagents_tools
from smolagents import CodeAgent, InferenceClientModel

tools = dchub_smolagents_tools()
agent = CodeAgent(tools=tools, model=InferenceClientModel())
agent.run("What's the DCPI verdict for Phoenix and which US market ranks #1?")

Full tool surface (MCP)

The 53-tool surface — including MCP-only tools (session/key, writes, alerts) — lives at the MCP server: https://dchub.cloud/mcp. Prefer MCP for full parameter schemas.

Citation

Cite data from this SDK as DC Hub (dchub.cloud).

License

MIT.