skills.rest
vellum-ai
vellum-aiOFFICIAL
Explore all skills by this author →
Website𝕏@vellum_aiGitHub

chat-complex-documents

Connects the Unstructured Transform MCP server to parse and search documents across 60+ formats.

1.2k|166|Updated Feb 7, 2026
Search Tags:#mcp#OCR#PDF#table extraction#document parsing#unstructured#rAG
One-click install
npx skills add https://github.com/vellum-ai/vellum-assistant --skill chat-complex-documents
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: chat-complex-documents
Source: https://github.com/vellum-ai/vellum-assistant/tree/main/skills/chat-complex-documents
Command: npx skills add https://github.com/vellum-ai/vellum-assistant --skill chat-complex-documents

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes assets (resource) components.

What problem does it solve?

Getting usable content out of messy documents — scanned PDFs, slide decks, spreadsheets, emails, images — usually requires building and maintaining custom parsing and OCR pipelines. This Skill connects the hosted Unstructured Transform MCP server so the assistant can partition, enrich, and structure documents into clean, AI-ready data for search, Q&A, and summarization.

Core Features & Use Cases

  • Document parsing across 60+ formats: Partitions PDFs, Word/Excel/PowerPoint, images, scanned files, and emails into structured elements like titles, paragraphs, and tables.
  • Enrichment: Adds metadata, table and image descriptions, and entity recognition to extracted content.
  • Grounded Q&A and search: Makes document sets searchable and answerable for RAG, knowledge bases, and summarization workflows.
  • Use Case: Hand the assistant a folder of contracts and scanned reports, then ask it to extract the tables from each PDF and answer questions grounded in the source content.

Quick Start

Ask the assistant to connect Unstructured Transform, sign in during the OAuth step, then say "Use Unstructured Transform to parse these files and extract the tables from this PDF."

Frequently Asked Questions about chat-complex-documents

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract tables and text from PDFs with Unstructured Transform?▼

Add the Unstructured Transform MCP server, authenticate via OAuth, then ask the assistant in plain language to parse your files. Transform partitions each document into structured elements like titles, paragraphs, and tables and returns structured JSON output.

What file formats does Unstructured Transform support?▼

Unstructured Transform supports over 60 formats, including PDFs, Word, Excel, PowerPoint, images, scanned files, and emails. Each file must be a supported type and 50 MB or smaller, with up to 10 files per request.

Does this work on both the Vellum desktop app and web app?▼

Yes, it works on both. On the desktop app, commands run via host_bash and OAuth opens your local browser; on the web app, commands run via bash and the platform handles the browser redirect.

Why does the Unstructured Transform server show 'Needs authentication'?▼

This means the OAuth step has not completed or the cached tokens are invalid. Re-run the auth command for the unstructured-transform server, sign in with your Unstructured account, then verify it shows Connected in the MCP server list.

What are the limits of Unstructured Transform requests?▼

Each file must be 50 MB or less, each request may include at most 10 files, and at most 5 requests may run concurrently. Plan batch processing around these limits when working with large document sets.

When should I not use Unstructured Transform for document parsing?▼

If you only need the plain text of one small local file, use the built-in file reading instead. Transform is intended for complex, varied, or scanned documents where structured extraction, enrichment, and OCR justify a hosted pipeline.

Related Skills

Explore More Data & Analytics Skills→
onebrain-aionebrain-ai
23

import

Import local files into structured Markdown notes under the OneBrain resources folder.

Official
Advanced
sylin-orgsylin-org
4

koan-ai-integration

Integrate chat endpoints, embeddings, and RAG workflows with vector search.

Official
Advanced
leonardoteodorooleonardoteodoroo

rag-implementation

Builds minimal RAG pipelines combining vector stores and document retrieval for grounded answers.

Community
Advanced
mrwyx12369mrwyx12369

file_reader

Read and summarize local text files, extracting key insights from JSON, YAML, CSV, and logs.

Community
Intermediate