AI Document Assistant — AI skill for Claude Code
Secure, multi-tenant enterprise document intelligence & RAG platform: sandboxed ingestion, hybrid pgvector/Qdrant search with authorization compiled into every query, verified citations, PostgreSQL ro.
How to install AI Document Assistant
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open mojtaba-py-code/ai-document-assistant and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What AI Document Assistant does
Secure, multi-tenant enterprise document intelligence & RAG platform: sandboxed ingestion, hybrid pgvector/Qdrant search with authorization compiled into every query, verified citations, PostgreSQL row-level security and a governed Claude LLM gateway. 1746 tests, 90% branch coverage.
Alternatives in AI
- Manifest — Real-time cost observability for OpenClaw agents — track tokens, costs, messages, and model usage 4.1k ★
- CliRelay — Self-hosted AI gateway for coding CLIs — one OpenAI/Claude/Gemini/Codex-compatible endpoint, with a multi-tena 1k ★
- Hazmat — Containment for AI agents - user isolation, sandboxed execution, network controls, backup/rollback 167 ★
README
AI Document Assistant
**Secure, multi-tenant enterprise document intelligence and RAG platform.** Upload contracts, invoices, policies and technical documents; search them by meaning and by keyword; ask questions and get answers that cite the exact page they came from — while every user only ever sees the documents they are entitled to.
[](https://github.com/mojtaba-py-code/ai-document-assistant/actions/workflows/ci.yml) [](https://github.com/mojtaba-py-code/ai-document-assistant/actions/workflows/codeql.yml)  
Why this is not a toy RAG demo
| Typical RAG demo | This platform |
|---|---|
| One user, one vector index | Many organisations, departments, roles, clearances and per-document grants |
| Retrieve top-k, then filter what the user may see | Authorization is compiled into the SQL of every search, and PostgreSQL row-level security enforces the tenant boundary underneath it |
| The prompt says "ignore instructions in documents" | The answering model has no tools; sources are nonce-delimited and escaped; injected passages are scored and excluded; every citation quote is verified in code |
| Everything goes to the LLM API | Data governance: RESTRICTED text never leaves for an external provider (LLM or embeddings); PII is pseudonymised before external calls |
| "Which contracts expire in 90 days?" answered by the LLM | Answered deterministically from extracted fields — exact dates, page citations, no model call |
| Parsers run in the web process | Untrusted files are parsed in a secret-free, resource-limited subprocess; macros, PDF JavaScript and zip bombs are quarantined |
Features
**Document management*
Related Skills
Achainsaw
A compiler for the code AI agents write: a flat SSA language with JSON diagnostics, compiled in under a millis
Retriva
Self-hosted, production-ready RAG chatbot agent — upload documents, get accurate AI answers with citations, ag
Orange Hourglass
Orange Hourglass 橙色沙漏 — 本地优先、AI 驱动的个人注意力管理工具。Local-first, AI-driven attention management with 4-Agent architec
Khipu
Persistent, cross-session memory for AI coding agents. Mac app and CLI, PostgreSQL 19 with pgvector, a knowled
Krusch Git
Git DAG indexing, AST symbol graphs, and hybrid semantic code search MCP server for AI coding agents (Claude C
Secure AI Customer Support Platform
Secure LLM-powered customer support platform: an AI agent (Claude) with grounded RAG answers, authorization-en
Related Agents
Appsec AI Security Reviewer
Use for AI/LLM application security review: direct/indirect prompt injection, tool and agent authorization, RA
Engineering Identity Access Engineer
Expert identity engineer for OAuth 2.0/OIDC flows, enterprise SSO (SAML/OIDC) and SCIM provisioning, passkeys/
AI Specialist
Expert in integrating LLMs, building RAG systems (pgvector), and creating MCP servers.