Agent Tools
The iterable.tools module exposes stable, JSON-serializable wrappers for common
IterableData operations. All tools return {"ok": true, "data": ...} or
{"ok": false, "error": "...", "code": "..."}.
Tools
| Tool | Description |
|---|---|
detect_format(path) | Detect format and compression |
describe_capabilities(format_id) | Format metadata + capabilities |
read_sample(path, n=10, redact=False) | Bounded row sample |
infer_schema(path) | Schema inference |
analyze_dataset(path, autodoc=False) | Structure analysis |
compute_stats(path) | Column statistics |
convert_file(input, output, confirm=False, dry_run=False) | Format conversion |
generate_documentation(path, **kwargs) | AI documentation |
validate_data(path, rules, mode="stats") | Row validation |
from iterable.tools import detect_format, read_sample, infer_schema
result = detect_format("data.csv")
if result["ok"]:
print(result["data"]["format"])
sample = read_sample("data.csv", n=5, redact=True)
schema = infer_schema("data.csv")
Schema export
from iterable.tools import schemas
openai_tools = schemas.to_openai_functions()
anthropic_tools = schemas.to_anthropic_tools()
schemas.call_tool("detect_format", {"path": "data.csv"})
LangChain
pip install iterabledata[langchain]
from iterable.tools.langchain import get_tools
tools = get_tools() # list of StructuredTool instances
MCP server
pip install iterabledata[mcp]
iterable-mcp