instructcanvas 0.2.1

InstructCanvas - DAG-based instruction format for agentic AI tools

instructcanvas

Python bindings for InstructCanvas (.icvs) — a DAG-based instruction format for agentic AI tools.

Install

pip install instructcanvas

Usage

import instructcanvas as icvs

source = """[node: rule1]
  type = rule
  content = "Use 2-space indentation"
  severity = must

[edge: rule1 -> rule2]

[target: claude]
  resolve = [rule1]
"""

# Parse
doc = icvs.parse(source)
print(f"Nodes: {doc.node_count()}, Edges: {doc.edge_count()}")

# Validate
result = icvs.validate(source)
print("Valid" if result.is_valid else "Invalid")

# Export Markdown
md = icvs.export_markdown(source, "claude")
print(md)

# Export DOT graph
dot = icvs.export_dot(source)

# .icvs ↔ Markdown
icvs_back = icvs.md_to_icvs("# Hello")
md_back = icvs.icvs_to_md(source)

# Convert to AI agent format
agent_json = icvs.convert_agent(source, "claude")

# Apply template variables
result = icvs.apply_template(source, {"FRAMEWORK": "React"})

API

Function	Description
parse(input)	Parse .icvs → Document
validate(input)	Validate → ValidationReport
export_markdown(input, target)	Export per-target Markdown
export_dot(input)	Export DOT graph
export_merge(input)	Merge all nodes → Markdown
md_to_icvs(markdown)	Convert Markdown → .icvs
icvs_to_md(input)	Convert .icvs → Markdown
convert_agent(input, format)	Export as Claude/OpenAI/JSON
apply_template(input, vars_dict)	Apply {{ VAR }} template

CLI (Rust)

cargo install icvs
icvs validate rules.icvs
icvs convert --target claude rules.icvs

License

MIT