feed · watching · stats · help with your tokens
mlimarenko/IronRAG
IronRAG is a self-hosted document search and grounded answer system for teams and AI agents.
Solves the need for self-hosted document search and grounded answers by providing a knowledge system for teams and AI agents that integrates with various LLM providers and MCP clients. Runs on Docker Compose or Kubernetes with PostgreSQL, pgvector, and Redis, using a hybrid retrieval pipeline across vector, lexical, graph, and technical-fact lanes.
Similar
- Arjun-Ingole/hop
hop is a Windows application that installs Linux alongside Windows by creating a bootable partition on the internal disk.
- JustVugg/colibri
colibri is a C inference engine that runs very large open mixture-of-experts models on the machine you already have.
- deepseek-ai/dsh-superagent-prototype
dsh-superagent is a research prototype for a proactive coding agent that works on a single GitHub repository using DeepSeek Harness.
- anthropics/FlashProve
FlashProve is a research prototype for zero-knowledge proofs of AI inference that allows providers to prove model identity without revealing weights.
- Muvon/octomind
Octomind is an open-source coding agent harness that gives an LLM real access to a repository through MCP tools.
- pgEdge/ai-dba-workbench
A PostgreSQL monitoring system that combines traditional and AI-based monitoring for instances v14+ including Supabase and Amazon RDS.
- moesaif/dna-studio
DNA Studio is a self-hosted, open-source AI marketing platform that generates on-brand social media content from a website URL.
- NVIDIA/auto-ontology
An open context layer that turns database metadata into an ontology, so AI agents answer questions about enterprise data with fewer wrong answers.
sources
Claims below include README evidence. A matching quote or worker agreement does not establish that a claim is correct. Category majority among 2 workers. Version 1.
IronRAG is a self-hosted document search and grounded answer system for teams and AI agents.
Self-hosted document search and grounded answers for teams and AI agents.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Solves the need for self-hosted document search and grounded answers by providing a knowledge system for teams and AI agents that integrates with various LLM providers and MCP clients.
Self-hosted document search and grounded answers for teams and AI agents.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Runs on Docker Compose or Kubernetes with PostgreSQL, pgvector, and Redis, using a hybrid retrieval pipeline across vector, lexical, graph, and technical-fact lanes.
Docker Compose (PostgreSQL + pgvector, Redis, backend, worker, frontend) or Helm on Kubernetes; data stays on your servers
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Builds a typed knowledge graph with entities, relations, and chunk-level citations for hybrid retrieval.
Typed knowledge graph — entities, typed relations, and chunk-level citations; hybrid retrieval across vector, lexical, graph, and technical-fact lanes
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Includes a native MCP server with 23 tools for documents, graph, web ingest, and grounded answers, compatible with Claude, Cursor, Codex, VS Code, Zed, and others.
Native MCP server — 23 tools for documents, graph, web ingest, and grounded grounded_answer; works with Claude Desktop / Code, Cursor, Codex, VS Code (Continue / Cline / Roo), Zed, OpenClaw, Hermes, and custom HTTP MCP clients
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Supports provider-agnostic AI bindings for OpenAI, DeepSeek, Qwen, GPTunnel, OpenRouter, RouterAI, MiniMax, Ollama, and LiteLLM.
Provider-agnostic AI — OpenAI, DeepSeek, Qwen / DashScope, GPTunnel, OpenRouter, RouterAI, MiniMax, Ollama, LiteLLM / OpenAI-compatible gateways
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Performs code-aware document ingestion using 15-language tree-sitter AST, typed JSON/YAML/TOML records, XLSX tables, and LLM Vision for scanned pages.
Code-aware document ingest — 15-language tree-sitter AST; typed JSON / YAML / TOML records and named XLSX tables; native Xberg parsing for document text and structure, with LLM Vision for scanned pages and images
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Features multi-tenant IAM with principals, scoped tokens, permission groups, and an audit log.
Multi-tenant IAM — principals, scoped tokens (system / workspace / library), permission groups, audit log
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Provides durable operations including restart-safe document jobs, recoverable assistant turns, and streaming tar.zst backup/restore.
Durable ops — restart-safe document jobs with page-range checkpoints; durable assistant turns that recover after stream drops; streaming tar.zst backup / restore
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Offers pluggable connectors for BookStack, Confluence, YouTrack, and Git repos with poll and webhook intake.
Pluggable connectors — BookStack, Confluence, YouTrack, Git repos, or build your own; see connectors/. They live in this repository and ship on this version (poll + webhook intake)
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Provides a self-hosted runtime via Docker Compose or Helm on Kubernetes, ensuring data remains on user servers.
Self-hosted runtime — Docker Compose (PostgreSQL + pgvector, Redis, backend, worker, frontend) or Helm on Kubernetes; data stays on your servers
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Maintains a typed knowledge graph with entities, relations, and chunk-level citations for precise information retrieval.
Typed knowledge graph — entities, typed relations, and chunk-level citations; hybrid retrieval across vector, lexical, graph, and technical-fact lanes
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Supports provider-agnostic AI integration with models from OpenAI, DeepSeek, Ollama, and others via configurable bindings.
Provider-agnostic AI — OpenAI, DeepSeek, Qwen / DashScope, GPTunnel, OpenRouter, RouterAI, MiniMax, Ollama, LiteLLM / OpenAI-compatible gateways
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Features code-aware document ingestion using tree-sitter AST for 15 languages and native parsing for structured data formats.
Code-aware document ingest — 15-language tree-sitter AST; typed JSON / YAML / TOML records and named XLSX tables; native Xberg parsing for document text and structure
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Offers multi-tenant identity and access management with scoped tokens, permission groups, and comprehensive audit logging.
Multi-tenant IAM — principals, scoped tokens (system / workspace / library), permission groups, audit log
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Ensures durable operations with restart-safe jobs, checkpoint recovery, and streaming backup/restore capabilities.
Durable ops — restart-safe document jobs with page-range checkpoints; durable assistant turns that recover after stream drops; streaming tar.zst backup / restore
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Supports pluggable connectors for platforms like BookStack and Confluence, allowing users to build custom integrations.
Pluggable connectors — BookStack, Confluence, YouTrack, Git repos, or build your own; see connectors/. They live in this repository and ship on this version
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Requires a credential master key and provider API keys in the .env file to start; the API and workers refuse to start without them.
First start only: the API and workers refuse to start without a credential master key, and .env.example ships it empty. Add IRONRAG_AI_PROVIDER_API_KEYS_JSON_B64 to the same file.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Must not overwrite an existing .env or regenerate the master key while the PostgreSQL volume persists, or stored credentials become undecryptable.
Never overwrite an existing .env, and never regenerate its key while the deployment's PostgreSQL volume persists: the stored credentials would stay undecryptable.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Users must never overwrite the existing .env file or regenerate the master key while the PostgreSQL volume persists, as stored credentials would become undecryptable.
Never overwrite an existing .env, and never regenerate its key while the deployment's PostgreSQL volume persists: the stored credentials would stay undecryptable.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Changing the vector dimension requires re-embedding data into a new shard, which can be a disruptive process.
Changing dimension requires re-embedding into a new shard.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: A wedged ingest queue can mimic ordinary load, where poisoned jobs hold leased rows and count against parallel job limits, potentially causing ingestion delays.
A wedged ingest queue looks like ordinary load. A poisoned job (a terminal ops_async_operation behind a claimable ingest_job) holds leased rows, and those count against ingestion_max_parallel_jobs_per_library.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: The default database configuration uses a local PostgreSQL instance with a specific URL, which may not be suitable for all production environments without modification.
The default in apps/api/src/app/config/loading.rs is postgres://postgres:postgres@127.0.0.1:5432/ironrag, and TempPostgresDatabase starts with drop database if exists.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: Changes to the answer pipeline must be verified through both the UI and an external MCP client, adding complexity to the testing process.
A change to the answer pipeline is verified through both channels. The built-in assistant is an MCP-client-shaped agent over the same tool surface an external MCP client gets, and grounded_answer stays the one high-level grounded-answer tool
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
Limitation: The vector dimension is not a fixed constant but depends on the active binding, requiring careful configuration and potential re-embedding when changed.
Vector dimension is a property of a library, not a constant of the product. It comes from the active EmbedChunk/EmbedQuestion binding.
Quote matched the README or GitHub description at extraction. Same claim and quote from 1 worker(s). Interpretation not reviewed by a maintainer.
ai-ml · homelab