MCP Servers · dmayboroda

minima

Minima is an open-source RAG (Retrieval-Augmented Generation) platform that runs on-premises using containerized services. It supports four deployment modes: fully local with Ollama, custom OpenAI-compatible LLMs, ChatGPT integration, and Anthropic Claude integration. Users index local documents and query them through a web UI or AI assistant integrations.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/dmayboroda/minima

GitHub stars

103

Forks

Python

Primary language

MPL-2.0

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	dmayboroda/minima
Owner	dmayboroda
Primary language	Python
License	MPL-2.0 — OSI-approved
Stars	1k
Forks	103
Open issues	14
Latest release	Unknown
Last updated	2026-01-22
Source	https://github.com/dmayboroda/minima

What minima is

Python-based RAG system using Docker Compose for orchestration. Employs Sentence Transformers for embeddings, Qdrant for vector storage, and supports pluggable LLM backends (Ollama, vLLM, TGI, LocalAI). Includes optional reranking via HuggingFace CrossEncoder. Integrates with MCP protocol, ChatGPT custom GPTs, and Anthropic Claude desktop app.

Quickstart

Get the minima source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/dmayboroda/minima.git2cd minima3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Private document search for enterprises

Organizations needing to index and query sensitive documents without sending data to third-party LLM providers. Fully on-premises deployment ensures data remains under company control.

AI assistant enrichment for existing tools

Teams already using ChatGPT or Claude who want to augment these tools with access to local documents and knowledge bases via custom GPTs or MCP protocol.

Self-hosted LLM infrastructure

Teams running their own LLM servers (vLLM, TGI, etc.) who need a lightweight RAG layer to combine retrieval with generation without dependencies on external APIs.

Implementation considerations

Resource overhead: Embedding models, rerankers, and LLMs require significant VRAM/GPU depending on deployment mode. Fully local setup (Ollama) demands 8GB+ RAM minimum.
Vector database setup: Qdrant deployment and index initialization must complete before queries work. Initial indexing of large document collections can be time-consuming.
Environment variable management: Four separate Docker Compose files with different variable requirements. Misalignment (e.g., setting RERANKER_MODEL in custom LLM mode) silently fails or wastes resources.
LLM endpoint stability: Custom LLM mode depends on external OpenAI-compatible servers. Failures in vLLM, TGI, etc. cascade to the entire RAG pipeline.
Supported file formats are limited: .pdf, .xls, .docx, .txt, .md, .csv only. No HTML, JSON, or other formats.

When to avoid it — and what to weigh

Need production-grade support — Project has no release cadence (latestRelease: none), 14 open issues, and no commercial support structure. Not suitable for mission-critical deployments requiring SLAs.
Limited Docker/DevOps expertise — Setup requires manual .env configuration, Docker Compose orchestration, and understanding of LLM deployment patterns. Significant operational overhead for teams unfamiliar with containerization.
Search-heavy workloads without reranking — Custom LLM mode intentionally skips reranking for performance. Accuracy may suffer on large document collections without this refinement step.
Require cross-platform GUI without code changes — Electron app mentioned only for local Ollama setup. Other deployment modes rely on web UI at localhost:3000. Limited UI tooling for other configurations.

License & commercial use

Mozilla Public License 2.0 (MPL-2.0) is a weak copyleft license. Permits commercial use, modification, and distribution provided: (1) source code changes are disclosed, (2) original MPL-2.0 license is included, and (3) patent rights are covered. Allows proprietary derivative works if modified files are isolated. Commonly used in enterprise software.

MPL-2.0 explicitly permits commercial use and proprietary derivatives. No license restrictions on resale, SaaS deployment, or integration. Caution: internal modifications must be disclosed if distributed. Recommend legal review for SaaS product wrapping or embedded deployment to clarify derivative disclosure boundaries. No commercial support, maintenance, or warranty from maintainer.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	High
DEV.co fit	Good
Assessment confidence	Medium

Security considerations

On-premises architecture reduces third-party data exposure. Considerations: (1) No authentication layer documented for web UI (localhost:3000 accessible to any local user), (2) Custom LLM_API_KEY transmitted in .env file; no secrets management integration, (3) ChatGPT/Claude integration requires Firebase credentials in plaintext, (4) Qdrant vector database likely exposed on internal Docker network without explicit access controls, (5) No encryption-at-rest for indexed embeddings mentioned. Docker network isolation provides basic containment but not suitable for multi-tenant or hostile environments without additional hardening.

Alternatives to consider

LlamaIndex (previously GPT Index)

Framework-agnostic, handles more LLM backends, stronger documentation and community. Requires more code but offers finer control. Supports more file types and preprocessing pipelines.

Langchain with Chroma/Weaviate

Modular stack with mature integrations, better for custom pipelines. Chroma simpler than Qdrant for small deployments. Langchain ecosystem more extensive.

Production-ready RAG framework with cloud-native deployment options, stronger monitoring, and commercial support available. Steeper learning curve but more robust for enterprise.

Software development agency

Build on minima with DEV.co software developers

Evaluate Minima for your LLM infrastructure team. Best suited for organizations with Docker expertise seeking on-premises RAG without third-party data dependencies.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

mini-rag MiniRAG MiniMax-MCP

Related on DEV.co

Explore the category and the services that help you build with it.

MCP Servers AI application development API development Cloud deployment Talk to DEV.co

minima FAQ

Can I use Minima in a production SaaS product?

Technically yes under MPL-2.0, but not recommended. No versioning, no release cadence, 14 open issues, and no commercial support. Embedding Minima requires disclosing modifications if distributed. Consider managed RAG platforms or enterprise Haystack for SaaS.

How do I scale Minima beyond one machine?

Not clearly addressed. Qdrant can be distributed, but Minima's orchestration is single-Docker-Compose-instance. Multi-node deployment, load balancing, and horizontal scaling require architecture redesign outside documented scope.

What's the difference between Ollama and Custom LLM modes?

Ollama bundles LLM + reranker in containers (simpler, heavier). Custom LLM uses external OpenAI-compatible endpoint with no reranking (lighter, faster but potentially less accurate). Choose based on infrastructure and accuracy needs.

Does ChatGPT integration leak my documents to OpenAI?

No. Only the user query and relevant document snippets are sent to ChatGPT API. Documents remain indexed locally, though encryption and audit logging are not documented.

Custom software development services

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If minima is part of your mcp servers roadmap, our team can implement, customize, migrate, and maintain it.

Ready to build a private document search?

Evaluate Minima for your LLM infrastructure team. Best suited for organizations with Docker expertise seeking on-premises RAG without third-party data dependencies.

Talk to DEV.co Schedule a Consultation