Practical overview
When it fits
Use Chunky before indexing documents in a RAG pipeline, especially when PDF layout, Markdown quality, chunk boundaries, or metadata need review.
Who it is for
- RAG developers preparing source documents before indexing them in a vector store.
- Teams or individuals who need to inspect PDF conversions and compare chunking strategies on actual documents.
- Users who want to enrich document or chunk metadata before ingestion.
Documented uses
- Convert PDFs into Markdown using a converter suited to the document type.
- Clean conversion artifacts and compare the output with the original PDF.
- Inspect chunks and compare supported splitting strategies and settings.
- Generate context-aware titles, summaries, keywords, and retrieval questions for chunks.
- Process multiple documents through conversion, enrichment, and chunking.
Requirements & limits
The project is described as local and open-source under the MIT license. The README documents local startup on macOS, Linux, WSL, and Windows PowerShell, plus a Docker Compose route.
- The available converters suit different inputs: Docling is listed for complex layouts, while the vision-model option targets scanned PDFs, handwriting, and diagrams.
- The vision-model converter rasterizes pages at 300 DPI by default; its default target is local Ollama, and the UI can configure a model name, base URL, and API key.
- A cloud converter sends the PDF to a configurable external endpoint.
- Chunk size and overlap controls apply only where the selected splitting strategy supports them.
- Saved chunks are tied to a Markdown revision and splitter configuration; changed Markdown must be re-chunked before replacing the current saved version.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Start Chunky locally or with Docker
Launch the documented local application using one of its two repository setup routes.
Before you start
- A clone of the Chunky repository.
- For local startup, macOS, Linux, WSL, or Windows PowerShell; alternatively, a Docker setup capable of running Docker Compose.
Steps
Clone the repository and change into its directory.
On macOS, Linux, or WSL, run ./start_all.sh; on Windows PowerShell, run .\start_all.ps1. Alternatively, run docker compose up --build.
Expected outcome
The README lists the frontend at http://localhost:5173, backend at http://localhost:8000, and Swagger interface at http://localhost:8000/docs.
Guide limitations
- The instructions describe startup but do not establish that a particular machine or environment has been tested.
- The README notes that Docling and local Ollama conversion may be significantly slower on CPU-only machines.
Compare options for Retrieval-augmented applications
These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.
| Project | Documented purpose | Platform | Access and requirements |
|---|---|---|---|
| Chunky | Convert PDFs into Markdown using a converter suited to the document type. | Self-hosted, Web | The project is described as local and open-source under the MIT license. The README documents local startup on macOS, Linux, WSL, and Windows PowerShell, plus a Docker Compose route. |
| Rag Powered AI Assistant Transforming Data Retrieval And Analysis Across The Web And Pdfs | Retrieve relevant passages from web and PDF content to answer a question. | Web | |
| LangChain in Supabase Edge Functions | Initialize and start a Supabase project locally. | Cloud, Server |
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- GiovanniPasq/chunky · Official homepage title · retained in its source language