Compile any document into AI-ready knowledge bases with built-in agent memory. One command. Zero friction.
60+ file formats. PDFs, DOCX, HTML, Markdown, code in 20+ languages, spreadsheets, emails, presentations.
3-tier memory OS with instant writes. Pad, episodic, and fact tiers give your agents persistent context.
Query any document by keyword or ID. Native wrappers for LangChain and LlamaIndex. Zero-config retrieval.
Automatically redact emails, phone numbers, and SSNs before compilation. Privacy by default.
No pickle. No arbitrary code execution. Uses safetensors + msgpack. Safe to share.
macOS, Windows, and Linux. Pure Python with zero OS-specific dependencies.
The same .aura archive works for RAG retrieval and agent memory.
from aura.rag import AuraRAGLoader
loader = AuraRAGLoader("knowledge.aura")
# Instant text retrieval
text = loader.get_text_by_id("doc_001")
# Framework wrappers
docs = loader.to_langchain_documents()
docs = loader.to_llama_index_documents()from aura.memory import AuraMemoryOS
memory = AuraMemoryOS()
# Write to memory tiers
memory.write("fact", "User prefers dark mode")
memory.write("episodic", "Discussed API design")
memory.write("pad", "TODO: check auth module")
# Search across all memory
results = memory.query("user preferences")Native integrations for the AI agent platforms you already use.
Persistent RAG and memory for always-on agents.
View on GitHub →Custom /aura commands for context-aware coding.
View on GitHub →Knowledge-backed Codex agents with memory.
View on GitHub →Slash commands and MCP support for Gemini CLI.
View on GitHub →$ pip install auralith-aura$ aura compile ./my_data/ --output knowledge.aura --pii-maskfrom aura.memory import AuraMemoryOS
memory = AuraMemoryOS()
memory.write("fact", "User prefers dark mode")
results = memory.query("preferences")Aura compiles entirely on your local hardware — no cloud uploads, no external APIs, no telemetry. Your documents stay yours.
Compiles on any modern laptop or desktop. Your machine, your setup — no cloud dependency.
Zero internet required after install. Compile sensitive documents without any data leaving your network.
macOS, Windows, and Linux. Python 3.8+ — runs wherever your team works.
Aura handles local compilation. For enterprise-scale training pipelines, model fine-tuning, and production-grade agent infrastructure — there's OMNI.
Open source. Cross-platform. Zero friction.