LLARS Documentation¶

Welcome to the documentation of the LLM Assisted Research System (LLARS).

Overview¶

LLARS is a platform for AI-supported analysis and evaluation of email counseling conversations. Core features:

Mail Rating & Ranking: Structured assessment and comparison of threads
LLM Integration: OpenAI and LiteLLM/Mistral
Authentication: Authentik (OIDC) with roles and permissions
Collaboration: Yjs for real-time synchronization
RAG Pipeline: Knowledge-grounded answers via ChromaDB

Architecture¶

Service Overview (internal ports)

flowchart LR
    client((Browser)) -->|HTTP| nginx[nginx :80]
    nginx --> frontend[Vue :5173]
    nginx -->|api| backend[Flask :8081]
    nginx -->|auth| backend
    nginx -->|authentik| authentik[Authentik :9000]
    nginx -->|collab| yjs[Yjs :8082]
    nginx -->|analytics| matomo[Matomo :80]
    nginx -->|docs| mkdocs[MkDocs :8000]

    backend <--> mariadb[(MariaDB :3306)]
    authentik <--> pg[(PostgreSQL)]

    backend -->|OIDC| authentik
    yjs -->|JWT| authentik

Standard Ports (Development) - 55080 -> nginx (Frontend + API + Matomo + Docs Proxy) - 55095 -> Authentik (optional direct; also via nginx /authentik/) - 55306 -> MariaDB (optional direct; debug only) - 55800 -> Docs (MkDocs, optional direct; also via nginx /mkdocs/ in dev, /mkdocs/ in prod)

In production, only 80/443 are exposed externally.

Quick Start¶

Clone the repository
Copy .env.template.development to .env and adjust as needed
Run the start script: ./start_llars.sh
Open: http://localhost:55080

Further Information¶

See Getting Started for installation and configuration details.