Document ingestion · OpenSearch

From a folder of files to a searchable vector index.

A multi-format document ingestion pipeline for OpenSearch with automatic vectorization and RAG search. Upload, ingest and search from a simple web UI or a REST API.

MITPython 3.11+OpenSearch 3.xHelm
bash
$ git clone https://github.com/iwebbo/DocVector
$ cd DocVector && docker compose up -d
# or deploy with the official Helm chart
$ _
How it works

Four steps from file to answer.

01 · UPLOAD

Any common format

PDF, Word, Excel, Markdown, code (Python, Java, JS) and YAML.

02 · EMBED

Automatic vectorization

Embeddings generated with Sentence Transformers.

03 · INDEX

OpenSearch k-NN

A vector index on OpenSearch 3.x, using FAISS as the k-NN engine.

04 · SEARCH

Hybrid retrieval

Semantic and keyword search combined, ready for RAG on top of the vectorized content.

Everything is available through the web UI and a documented REST API, with curl examples for each endpoint.

Features

Ingestion without the glue code.

FORMATS

Multi-format ingestion

PDF, Word, Excel, Markdown, code and YAML in one pipeline.

VECTORS

Automatic vectorization

Sentence Transformers embeddings, no manual step.

SEARCH

Hybrid k-NN search

A vector index with semantic and keyword search together.

UI

Simple web UI

Upload, ingest and search through a GUI.

API

REST API

A programmatic interface with documented curl commands.

DEPLOY

Compose and Helm

Docker Compose to start fast, an official Helm chart for Kubernetes.

Every capabilityexpand

Core

  • Multi-format ingestion — PDF, Word, Excel, Markdown, code, YAML
  • Automatic vectorization — Sentence Transformers
  • OpenSearch k-NN — hybrid semantic and keyword
  • Web UI — upload, ingest, search
  • RAG search — retrieval on vectorized content

Deployment

  • Docker Compose — quick start
  • Kubernetes / Helm — production-ready chart
  • REST API — for developers
  • Curl commands — documented endpoints

Stack

  • Backend — Python 3.11+ with FastAPI
  • Database — OpenSearch 3.x
  • Orchestration — Docker and Kubernetes
  • k-NN engine — FAISS
Built withPythonFastAPIOpenSearchDockerKubernetesHelm
Star or fork on GitHub ↗