OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Longparser

Longparser is a Python document-processing system that extracts PDFs, DOCX, PPTX, XLSX and CSV files, validates and chunks their contents, and supports human review before embedding and indexing for RAG. It also provides a FastAPI server with document jobs and memory-based chat.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Longparser — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Longparser when a RAG workflow needs document parsing and chunking, particularly when a human review stage should precede embedding and indexing.

Who it is for

  • Developers preparing office documents for retrieval-augmented generation.
  • Teams that want human approval or editing before extracted content is embedded.
  • Builders integrating document ingestion and document-grounded chat through Python or FastAPI.

Documented uses

  • Extract structured content from PDFs and common office formats.
  • Create token-aware, heading-aware and table-aware retrieval chunks.
  • Review, edit or reject extracted blocks and chunks before embedding.
  • Embed and index reviewed content in a supported vector store, then chat about indexed documents.
Before you choose

Requirements & limits

The project is identified as open source and its repository is public. Python 3.11 or newer is required.

  • The core SDK and server are distinct installation choices; the server extra adds FastAPI, MongoDB, Redis and LangChain chat.
  • The default PDF OCR path requires Tesseract and the appropriate language data.
  • Vector-store options named in the documentation include Chroma, FAISS and Qdrant; Qdrant is a separate extra.
  • The project describes local extraction and embeddings, but hosted LLM or embedding providers receive content sent to them.
  • Memory chat is described as short-term turns, a rolling summary and long-term facts.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Process and chunk a document with the Python SDK

Use the SDK to extract a supported file and produce chunks.

Before you start

  • Python 3.11 or newer
  • Install the core package with pip install longparser
  • For the default PDF OCR path, Tesseract and the appropriate language data

Steps

  1. Create a DocumentPipeline with ProcessingConfig().

  2. Pass a document path to process_file().

  3. Pass the result and a ChunkingConfig() to pipeline.chunk().

Expected outcome

The example exposes document page metadata and the resulting chunk count and text.

Guide limitations

  • The documented example does not cover embedding, indexing or chat.
  • The default PDF OCR path requires Tesseract and language data.
Documentation 1 ↗
Source-based practical guide

Run the documented API workflow from upload to document chat

Configure the service, upload a document, finalize and embed it, then create a chat session.

Before you start

  • Python 3.11 or newer
  • Docker Compose
  • MongoDB and Redis, started with the documented Compose setup
  • An API key and a configured .env file

Steps

  1. Copy .env.example to .env and edit the configuration.

  2. Start the API, worker, MongoDB and Redis with the command below.

    Documented command
    docker compose up -d --build

  3. Upload a supported document to the /jobs endpoint and check its status until it is ready_for_review.

  4. Finalize the job and enqueue embedding using the documented API requests.

  5. After the job is indexed, create a chat session and submit a question about the document.

Expected outcome

The documented sequence takes a document job through review readiness and indexing to a document-grounded chat request.

Guide limitations

  • The example finalizes with approve_all_pending; it does not show individual block or chunk edits.
  • The example uses a Hugging Face embedding model and Chroma; Qdrant and optional extraction backends are separate extras.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
ENDEVSOLS/LongParser · Official homepage title · retained in its source language
How we review sourcesSuggest a correction