Skip to main content

​
What is Moorcheh On-Prem?

Moorcheh On-Prem runs on your machine: semantic search over text documents and precomputed vectors. Text embeddings use Ollama (local), OpenAI, or Cohere — configured once with moorcheh configure or on first moorcheh up. You install moorcheh-client from PyPI, start the stack with one command, and call the API from your app or the CLI.
PyPI packagemoorcheh-client
CLImoorcheh
Python importfrom moorcheh import MoorchehClient
Server imagemoorcheh/server:latest (Docker Hub)
Default APIhttp://localhost:8080
Data on disk~/.moorcheh/data
On-prem has no Moorcheh cloud API keys. If you use OpenAI or Cohere for embeddings, your provider API key is stored locally in ~/.moorcheh/config.json. Ollama needs no API key.

​
What you can do

  • Create text namespaces (documents embedded via your configured provider) or vector namespaces (you supply vectors)
  • Upload text documents by id, or upload files (PDF, DOCX, etc.) from ~/.moorcheh/uploads
  • Upload, get, and delete items by id; list and delete indexed files separately
  • Semantic search across one or more namespaces (top hit can include batch summary_text)
  • AI Answer — RAG or direct LLM via POST /answer (Ollama, OpenAI, or Cohere)
  • Check global storage quota via /health

​
Prerequisites

  • Python 3.10+
  • Docker (Desktop or Engine)
  • Embedding provider for text workflows — Ollama, OpenAI, or Cohere (Ollama is optional if you use cloud embeddings or vector-only)

​
Install and run

pip install moorcheh-client
moorcheh configure   # optional; first moorcheh up can prompt instead
moorcheh up
moorcheh status
See Quickstart for a full upload-and-search walkthrough.
Config vs runtime: moorcheh configure writes ~/.moorcheh/config.json only. To apply new embedding or LLM settings to a server that is already running, use moorcheh down then moorcheh up. If you already have data under ~/.moorcheh/data, changing the embedding model can break text search and RAG until you re-upload — see Embedding providers.

​
Limits

  • Unlimited namespaces - no cloud-tier namespace count limit; create as many text or vector namespaces as you need
  • 100,000 items total across all namespaces (text + vectors)
  • Item ids are unique per namespace (the same id string may exist in different namespaces)
  • Get/delete by id: at most 100 ids per request

​
Documentation

SectionUse when
PrerequisitesInstall Python, Docker; optional Ollama
Embedding providersModels, dimensions, configure / up behavior
QuickstartFirst run end-to-end
API referenceIntegrating via HTTP / curl
Python clientBuilding apps with MoorchehClient
CLILocal ops and testing with moorcheh

​
Data persistence

Vectors and documents are stored under: 
~/.moorcheh/data/
 moorcheh_data_store.json
 namespace_registry.json
 file_registry.json
~/.moorcheh/uploads/     # host files mounted as /uploads in the server (read-only)
moorcheh down stops containers but does not delete this folder. Back up ~/.moorcheh to save everything. File upload indexes content from uploads/; deleting an index does not remove files on disk.