ParisNeo/lollms
ParisNeo/lollms has 106 GitHub stars and 37 forks. GitHot measured 0 stars in 24 hours and 0 in 7 days. Last checked 2026-11-17 02:47 UTC. Last push 2026-11-15.
What you can do
- Multi-User Support: Secure token-based authentication. Each user has their own isolated data.
- Simplified Installation: Get started quickly with simple run.sh or run_windows.bat scripts.
- Environment-Based Configuration: Easy setup using a .env file, automatically generated from .env.example.
- Persistent Discussions: Chat histories are saved per user and can be revisited, renamed, starred, and deleted.
Measured star growth
- in 24h
- 0
- in 7 days
- 0
- in 15 days
- Collecting history
- in 1 month
- Collecting history
Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.
About
An all in one AI solution compatible with any known AI service on the planet
AIREADME.md
OriginalLoLLMs - v2.3.0 "HeartBeat"
One tool to rule them all!
A multi-user FastAPI backend and responsive Vue/Tailwind CSS frontend application designed to provide a chat interface powered by the lollms_client library. It features integrated Retrieval-Augmented Generation (RAG) using safe_store, a versatile personality system, multimodal chat, user management, a friend system with direct messaging, and enhanced sharing capabilities.
Live Project: https://github.com/ParisNeo/lollms
✨ Features
- Multi-User Support: Secure token-based authentication. Each user has their own isolated data.
- Simplified Installation: Get started quickly with simple
run.shorrun_windows.batscripts. - Environment-Based Configuration: Easy setup using a
.envfile, automatically generated from.env.example. - Persistent Discussions: Chat histories are saved per user and can be revisited, renamed, starred, and deleted.
- LLM Integration: Uses
lollms-clientto interact with various LLM backends. - Streaming Responses: AI responses are streamed for a real-time experience.
- Multimodal Chat: Upload images with text prompts for vision-capable models.
- Advanced Personality System: Create, edit, and delete custom personalities with unique system prompts, scripts, and data sources.
- Retrieval-Augmented Generation (RAG):
- Organize documents into multiple DataStores per user.
- Upload various file types (
.txt,.pdf,.docx, etc.). - Toggle RAG usage per discussion and select specific DataStores.
- Share DataStores with friends with configurable permissions.
- Friend System & Direct Messaging (DM):
- Send, accept, and reject friend requests.
- Engage in real-time direct messaging with friends.
- Admin Panel:
- Manage users (Add, List, Delete, Reset Password).
- Configure global settings and manage LLM/TTI/TTS bindings.
- Data Export/Import: Users can export/import their discussions and settings.
- Responsive UI: Built with Vue.js and Tailwind CSS, featuring Markdown rendering, code highlighting, and math rendering with KaTeX.
🚀 Getting Started
Prerequisites
- Python 3.10+
- Git
Installation
The easiest way to get started is by using the provided run scripts, which handle setup and execution.
Clone the Repository:
git clone https://github.com/ParisNeo/lollms.git cd lollmsRun the Installer:
- On Windows: Double-click
run_windows.bat. - On macOS or Linux:
chmod +x run.sh ./run.sh
The first time you run the script, it will create a Python virtual environment, install all required dependencies, and create a default
.envfile from.env.example. Subsequent runs will just start the application.- On Windows: Double-click
Usage
- Access the UI: Once the server is running, open your web browser and go to
http://localhost:9642(or the host and port you configured). - Create an Admin Account: On the first launch, you will be prompted to create an administrator account.
- Login: Use your newly created credentials to log in.
- Explore:
- Start a new chat.
- Go to Settings to configure your profile and select an LLM model.
- Visit the Admin Panel to configure LLM bindings and other global settings.
⚙️ Configuration (.env file)
Configuration is managed through the .env file in the project's root directory. When you first run the application, this file is created for you from .env.example.
SERVER_HOST&SERVER_PORT: The host and port the application will run on.DATABASE_URL: The location of the main SQLite database file.SECRET_KEY: Change this to a long, random string for production.ALLOW_NEW_REGISTRATIONS: Set tofalseto disable public sign-ups.INITIAL_ADMIN_USERNAME&INITIAL_ADMIN_PASSWORD: Used on first startup if no admin exists.
For detailed information on all available settings, please refer to the comments within the .env.example file.
📚 API Documentation
- Swagger UI:
http://localhost:9642/docs - ReDoc:
http://localhost:9642/redoc
🧳 Personalities (Handbags)
Personalities are filesystem handbags, not database rows. Behavior (system
prompt, tools, skills, RAG, docs) is defined by each handbag folder's contents
(SOUL.md, handbag.yaml, tools/, skills/, rag/, docs/) and loaded at
runtime via LollmsPersonality.from_handbag().
Storage layout (under the app data folder):
data/personalities/
├── global/ # admin / zoo / uploaded handbags, visible to all users
└── users/<user_id>/ # per-user handbags from the Personality Studio or zip upload
Installing a personality is a folder copy: zip uploads are extracted to a staging directory and atomically renamed into the target root; zoo installs are plain copies. The DB never stores prompt text, tools, skills, or RAG metadata.
Scopes: the global header groups personas by global / personal / shared.
Users can share a personal handbag with friends or groups — sharing is a
metadata-only ACL (PersonalityShareLink); the handbag physically stays in the
owner's folder and consumers get read-only discovery. The only other DB table
is UserPersonalityPreference, a pointer to the user's selected (scope, handbag_name, owner_id).
Discussions resolve the active preference to a handbag path (with ACL enforcement for shared scope) and call:
from lollms_client.lollms_personality import LollmsPersonality
personality = LollmsPersonality.from_handbag(resolved_handbag_path)
Missing or malformed folders degrade gracefully: the path is logged and the chat loop falls back to a safe default (NullPersonality contract).
🎙️ Voices Studio
Voice-profile selection is persisted under [voices_studio] in config.toml
(mirrored on the user record as active_voice_id).
voices_studio.active_voice_id— required. The id of the currently selected voice profile. It must match anidreturned by the voice registry (GET /voices). This key is the single source of truth for "which voice is active." Do not reuselollms_client_defaults.ai_namefor this — that field is a prompt label, not a voice selector, and conflating the two is a common cause of persisted-selection overwrites.
Troubleshooting: if the wrong persona appears after selecting a different
voice, verify in config.toml that voices_studio.active_voice_id matches the
desired voice's id, and that the id is present in GET /voices. A mismatch
between these two makes the UI render the previous persona even though the new
one was clicked.
🤝 Contributing
Contributions are welcome! Please open an issue or submit a pull request.
📜 License
Apache License 2.0. See LICENSE.