ShapeCalyx provides a Windows project workspace for AI-assisted counting of tree shapes. Ask the connected model to explain the recurrence and integer bounds, then verify the proposed implementation.
def tree_shapes(internal_nodes):
if type(internal_nodes) is not int or not 0 <= internal_nodes <= 100:
raise ValueError("shapes.nodes")
counts = [1] + [0]*internal_nodes
for size in range(1, internal_nodes+1):
counts[size] = sum(counts[left]*counts[size-1-left] for left in range(size))
return counts[internal_nodes]
Base URL, credentials and a reachable server.
Test the single leaf and small known counts, rejecting Boolean or out-of-range inputs without constructing all trees.
Compare actual behavior with the expected result. If a check fails, investigate it before applying more changes.
def tree_shapes(internal_nodes):
if type(internal_nodes) is not int or not 0 <= internal_nodes <= 100:
raise ValueError("shapes.nodes")
counts = [1] + [0]*internal_nodes
for size in range(1, internal_nodes+1):
counts[size] = sum(counts[left]*counts[size-1-left] for left in range(size))
return counts[internal_nodes]
ShapeCalyx gives you a Windows coding workspace for examining structure-related logic with an AI model connected to your project. Start by explaining what one structure represents and which distinctions count as different results. Ask the assistant to inspect the source and its callers before proposing a change, so an apparently simple numeric output retains the meaning the application requires.
Counting ordered full binary-tree shapes is one example task. Bring small known counts and describe the single-leaf case before discussing the recurrence. Review whether the proposed code counts structures or constructs them, and whether it preserves exact integer results. Those choices matter to the caller even when both implementations look reasonable on the smallest inputs.
After inspecting the patch, run your project's reference checks and rejected-input cases, including the declared upper boundary. Compare any changed output handling with the original requirement instead of judging only the calculation's final line. ShapeCalyx connects to supported OpenAI-compatible or Anthropic endpoints with credentials when necessary. An unfiltered model is an option if your chosen service offers one; its tools and context limits depend on the server. The work remains a reviewable software change tied to explicit structural definitions.
It determines which structures count as different. Left and right placement can distinguish shapes under an ordered definition. Include small examples that expose this distinction so a proposed replacement does not silently count a different family of objects.
The illustrated contract counts the single leaf as one shape. Keep that base case explicit in tests and in any explanation, because changing it affects later counts and the meaning of an empty or minimal structure.
The existing task asks for a count using exact integers. Building all structures changes the work and output substantially. If you need actual trees, make that a separate requirement rather than accepting it as an incidental rewrite.
Review type checks, supported bounds and how the caller handles the returned integer. Test Boolean and out-of-range inputs as well as known counts, so a numerically correct rewrite does not weaken the surrounding software contract.
ShapeCalyx · Updated 2026-10-25