A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
/* Copyright 2026 Marimo. All rights reserved. */
import type { LucideIcon } from "lucide-react";
import { BookPlusIcon, FileSymlink } from "lucide-react";
import type { StorageEntry, StorageNamespace } from "@/core/storage/types";
type BackendType = StorageNamespace["backendType"];
export interface StorageSnippetContext {
variableName: string;
protocol: string;
entry: StorageEntry;
backendType: BackendType;
}
export interface StorageSnippet {
id: string;
label: string;
icon: LucideIcon;
/** Return the code string, or null to hide the snippet for this context. */
getCode: (ctx: StorageSnippetContext) => string | null;
}
const NOT_SIGNABLE_PROTOCOLS = new Set(["http", "file", "in-memory"]);
function escapeForPythonString(value: string): string {
return JSON.stringify(value).slice(1, -1);
}
interface ParsedHfRepoPath {
repoType: "model" | "dataset" | "space";
repoId: string;
filename: string;
}
// Namespaced prefixes are reserved: they can never be the first segment of a
// model repo id, so a path like "datasets/org" that is missing a filename
// must not fall through and be misparsed as the model repo "datasets/org".
const RESERVED_PREFIXES = new Set(["datasets", "spaces", "buckets"]);
function parseHfRepoPath(path: string): ParsedHfRepoPath | null {
const parts = path.split("/").filter(Boolean);
if (parts[0] === "datasets" && parts.length >= 4) {
return {
repoType: "dataset",
repoId: `${parts[1]}/${parts[2]}`,
filename: parts.slice(3).join("/"),
};
}
if (parts[0] === "spaces" && parts.length >= 4) {
return {
repoType: "space",
repoId: `${parts[1]}/${parts[2]}`,
filename: parts.slice(3).join("/"),
};
}
if (parts.length >= 3 && !RESERVED_PREFIXES.has(parts[0])) {
return {
repoType: "model",
repoId: `${parts[0]}/${parts[1]}`,
filename: parts.slice(2).join("/"),
};
}
return null;
}
function hfHubDownloadSnippet(parsed: ParsedHfRepoPath): string {
const repoId = escapeForPythonString(parsed.repoId);
const filename = escapeForPythonString(parsed.filename);
const repoType = escapeForPythonString(parsed.repoType);
return `from huggingface_hub import hf_hub_download\n\nlocal_path = hf_hub_download(\n repo_id="${repoId}",\n filename="${filename}",\n repo_type="${repoType}",\n)`;
}
export const STORAGE_SNIPPETS: StorageSnippet[] = [
{
id: "read-file",
label: "Insert read snippet",
icon: BookPlusIcon,
getCode: (ctx) => {
if (ctx.entry.kind === "directory") {
return null;
}
const path = escapeForPythonString(ctx.entry.path);
if (ctx.backendType === "huggingface") {
const parsed = parseHfRepoPath(ctx.entry.path);
if (!parsed) {
return null;
}
return `${hfHubDownloadSnippet(parsed)}\n\nwith open(local_path, "rb") as f:\n _data = f.read()\n_data`;
}
if (ctx.backendType === "obstore") {
return `_data = ${ctx.variableName}.get("${path}").bytes()\n_data`;
}
return `_data = ${ctx.variableName}.cat_file("${path}")\n_data`;
},
},
{
id: "download-file",
label: "Insert download snippet",
icon: FileSymlink,
getCode: (ctx) => {
if (ctx.entry.kind === "directory") {
return null;
}
const path = escapeForPythonString(ctx.entry.path);
if (ctx.backendType === "huggingface") {
const parsed = parseHfRepoPath(ctx.entry.path);
if (!parsed) {
return null;
}
return `${hfHubDownloadSnippet(parsed)}\nlocal_path`;
}
if (ctx.backendType === "obstore") {
if (NOT_SIGNABLE_PROTOCOLS.has(ctx.protocol)) {
return null;
}
return `from datetime import timedelta\nfrom obstore import sign\n\nsigned_url = sign(\n ${ctx.variableName}, "GET", "${path}",\n expires_in=timedelta(hours=1),\n)\nsigned_url`;
}
const filename = escapeForPythonString(
ctx.entry.path.split("/").pop() || "download",
);
return `${ctx.variableName}.get("${path}", "${filename}")`;
},
},
];