Review
Inspect candidate packages without losing provenance.
Source, repository, and ownership stay visible so teams can decide whether to use the upstream package or maintain their own published version.
rawctx Hub
Official semantic packs for customers and AI
Discover candidate packages, review provenance and structure, then publish the version you own for code and AI use.
Claim
Claim upstream candidates only when you maintain them.
Indexed packages stay review-first, and claim makes the maintainer handoff explicit before anything is re-published under a company scope.
Publish
Publish the version your company actually owns.
rawctx keeps review, snapshot download, Python access, and upstream boundaries explicit at the point of use.
Python SDK
Load typed models or compress prompt context without leaving the hub
load() turns a package into typed semantic objects for Python, and to_prompt() turns the same package into compact LLM-ready context.
Python SDK
Typed semantic access with load()
Use the package ref directly in Python and inspect datasets, measures, dimensions, and relationships without parsing YAML yourself.
import rawctx
model = rawctx.load("@scope/name")
print(model.datasets) # ["users", "trips", "expense_reports"]
print(model.measures) # [Measure(name="total_amount", ...)]
print(model.dimensions) # [Dimension(name="currency_code", ...)]
print(model.relationships) # [Relationship(name='...', ...)]Python SDK
LLM-ready context with to_prompt()
Scope the prompt to a few datasets, keep it token-bounded, and pass the result straight into your agent or notebook workflow.
import rawctx
prompt = rawctx.to_prompt(
"@scope/name",
datasets=["expense_reports", "users"],
max_tokens=2000,
)
print(prompt)