pipeshub-ai/pipeshub-ai

PipesHub – Open‑Source Workplace AI Platform

What it is – PipesHub is a self‑hostable platform that lets you connect large‑language‑model (LLM) applications to the data stored in your company’s everyday tools (Slack, Google Drive, GitHub, Microsoft 365, Notion, …). It builds a searchable knowledge graph from those sources, enforces the original permissions on every query, and returns explainable answers with precise citations.

Key capabilities

• Explainable, citation‑backed answers – each response points to the exact block of the source document that supports it.
• Permission‑aware search – the system checks the user’s access rights in the source system at query time, so no data is exposed to the wrong person.
• Bring‑Your‑Own‑Model – any LLM provider can be plugged in via LangChain; the model runs inside your VPC, so your data never leaves your infrastructure.
• 50+ enterprise connectors – ready‑made integrations for Slack, Google Workspace, GitHub, Microsoft 365, Notion and many others, with real‑time or scheduled indexing.
• Graph‑backed retrieval – Neo4j or ArangoDB stores relationships between entities, while Qdrant/OpenSearch/Redis provide vector similarity search.
• Multimodal support – PDFs (including scanned), Word/Excel/PowerPoint, Markdown, HTML, images, and upcoming audio/video handling.
• No‑code agent builder – visually design AI agents that can act on connected tools without writing code.
• Developer‑first extensibility – REST‑style MCP server, Python/TypeScript/Go SDKs, custom connectors, and independent micro‑services that can be scaled separately.

Typical use cases

• Unified enterprise search with source‑level citations.
• Knowledge‑augmented Q&A or research assistants that respect corporate data policies.
• Building internal AI agents (e.g., ticket triage, report generation) that act with the same permissions a human would have.
• Automating workflows that need to read/write to multiple SaaS tools in a single, governed context.

Tech stack (as described in the repo)

Layer	Technologies
Frontend	Next.js (App Router), TypeScript, Radix UI Themes, Zod, React Hook Form
Backend	FastAPI, Celery, Kafka/Redis Streams, Redis cache & KV store, etcd, MongoDB document store, Neo4j/ArangoDB graph DB, Qdrant/OpenSearch/Redis vector DB, S3/Azure Blob storage
LLM integration	LangChain (multi‑provider), sentence‑transformers / fastembed for embeddings
Document processing	pdfplumber, selectolax, markdown‑it, openpyxl, csv, optional Docling, LibreOffice, CairoSVG
Data analysis	pandas

Getting started

1. Install Docker Compose (v2).
2. Run the one‑liner installer:

   curl -fsSL https://get.pipeshub.com/install | bash
   

   – The script downloads the compose files, runs an interactive wizard (slim vs. full deployment, graph DB choice, etc.), creates a .env with random secrets, pulls the images and starts the stack.
3. Open http://localhost:3000 (HTTPS required on cloud hosts).
4. For developers, clone the repo and run ./install.sh (add --build to compile images locally).

Extending / integrating

• Use the MCP server (repo pipeshub-ai/mcp-server) to expose the context to any MCP‑compatible client.
• SDKs are provided for Python, TypeScript, and Go (separate repos). They wrap the REST APIs and MCP endpoints, handling authentication tokens that respect per‑user permissions.
• Custom connectors can be added by implementing the connector interface and registering it in the backend configuration.

Roadmap highlights

• Personalised search based on team/role/history (in progress).
• PageRank‑augmented relevance over the knowledge graph.
• Fully managed PipesHub Cloud (wait‑list open).

Community & support

• Discord community, GitHub issues, security policy, contribution guide, and extensive docs are linked from the README.

---

PipesHub is a genuine, open‑source AI‑focused project aimed at enterprise knowledge integration, searchable AI assistants, and permission‑safe agent workflows.