|

Overview

This project helps an internal user search SharePoint documents through an assistant workflow. The request can include a document query, a folder hint, a file type, and an instruction to deliver the best match back to Slack.

The workflow searches Microsoft Graph, ranks possible folders, falls back to a bounded recursive crawl when needed, and can deliver a selected file after safety checks.

The private workflow keeps the real SharePoint structure, drive IDs, Slack channel IDs, credentials, and document names inside n8n. The public repo documents only the architecture and reusable helper patterns.

The challenge

Internal document lookup can be slow when files are stored across SharePoint folders, nested structures, and mixed file types. A user might know the topic, a partial folder name, or only the type of file they need.

The challenge was to make search useful from a simple assistant request without exposing the SharePoint structure and without letting fallback logic crawl too much data.

My role

I built the n8n workflow structure for SharePoint search, folder ranking, recursive crawl fallback, file-type filtering, and Slack delivery.

I also prepared the sanitized public documentation and helper snippets so the architecture can be reviewed without publishing workflow exports or internal identifiers.

Workflow structure

Core modules

Technical decisions

Tech stack

Proof / Evidence

This case study is based on a private SharePoint search assistant. The public proof focuses on architecture, design decisions, and sanitized snippets. Original workflow exports, drive IDs, folder paths, Slack IDs, credential IDs, document names, and tenant details are not published.

Live system

Click to enlarge

Public code & documentation

Code excerpt: file type filtering

The assistant accepts human file-type words and maps them to file extensions before formatting the SharePoint results.

const TYPE_MAP = {
  image: ["jpg", "jpeg", "png", "gif", "webp"],
  pdf: ["pdf"],
  email: ["msg", "eml"],
  excel: ["xlsx", "xls", "csv"],
  word: ["docx", "doc"],
};

function filterFilesByType(items, fileType) {
  const filesOnly = items.filter((item) => item.file);
  const allowed = TYPE_MAP[String(fileType || "").toLowerCase()];

  if (!allowed) return filesOnly;

  return filesOnly.filter((item) => {
    const ext = String(item.name || "").split(".").pop().toLowerCase();
    return allowed.includes(ext);
  });
}

Full sanitized example in snippets/file-result-formatter.js. The repository also documents query cleanup, folder ranking, recursive crawl queue handling, and Slack delivery guardrails.