Installing LIT Platform with pip#
Current version: 2.5.3 · Release notes · All releases
The fastest way to get started with LIT Platform for chat and AI development. No Docker required.
Personal & Non-Commercial Use Only
This installation is licensed for personal and non-commercial use. If you're using LIT Platform within a business, as part of a product or service, or in any revenue-generating context, a commercial license is required. Contact us to discuss terms.
Requirements#
- Python 3.10 - 3.13 (Windows: 3.10 or 3.11 recommended — see note below)
- One of the following AI providers:
- Claude — Claude Pro/Max account + Node.js (to install the Claude CLI)
- Gemini — free API key from Google AI Studio (no other software needed)
- ChatGPT — OpenAI account + Node.js (to install the OpenAI Codex CLI)
Installation#
Download and run the installer — it auto-detects your OS, architecture, and Python version:
Windows: native or WSL?
- Native Windows — simplest path; fully supported for chat and agents. GPU training requires Python 3.10/3.11.
- WSL2 + Ubuntu — recommended for ML work: full GPU support on any Python version, and the platform behaves exactly like the macOS / Linux docs everywhere else.
curl -fsSL https://github.com/Positronic-AI/lit-releases/releases/latest/download/install.py | python3
Then start the platform:
Or download and run in two steps:
irm https://github.com/Positronic-AI/lit-releases/releases/latest/download/install.py -OutFile install.py
python install.py
Then start the platform:
Recommended for ML and GPU training work. Install Ubuntu under WSL from an administrator PowerShell:
Why Ubuntu-24.04 and not Ubuntu?
The bare Ubuntu name installs the newest release, which may ship a Python version we don't have wheels for yet (e.g. Python 3.14). Ubuntu 24.04 LTS ships Python 3.12, which is fully supported.
Reboot if prompted, then launch Ubuntu and create your Unix user account when asked. Fresh Ubuntu images don't include pip, and Ubuntu 24.04+ blocks installs into the system Python — so set up a virtual environment first, then run the standard installer inside it:
sudo apt update && sudo apt install -y python3-venv
python3 -m venv ~/lit-env
source ~/lit-env/bin/activate
curl -fsSL https://github.com/Positronic-AI/lit-releases/releases/latest/download/install.py | python3
lit serve
In future shells, activate the venv again first (source ~/lit-env/bin/activate) — or skip activation and run ~/lit-env/bin/lit serve directly.
WSL-specific notes:
- Open the app from your Windows browser at http://localhost:8080 — WSL2 forwards localhost automatically.
- Install Node.js (needed for the Claude or Codex CLI) inside WSL via
aptornvm, not on Windows. - For GPU training, a current NVIDIA driver on the Windows side is all you need — do not install a driver inside WSL.
- If the install fails with a
CreateVmtimeout, see troubleshooting.
If you prefer to install manually, pick the wheel for your Python version and platform from the releases page:
# Example: Python 3.12 on Linux (lightweight — chat & agents only)
pip install https://github.com/Positronic-AI/lit-releases/releases/download/v2.5.3/positronic_lit-2.5.3-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
To include the deep learning studio and ML pipeline, add the [full] extra:
# Full install with ML dependencies (TensorFlow, scikit-learn, etc.)
pip install "positronic-lit[full] @ https://github.com/Positronic-AI/lit-releases/releases/download/v2.5.3/positronic_lit-2.5.3-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl"
See Choosing Your Python Version and Install Extras below for details.
Open your browser to http://localhost:8080 and follow the setup wizard to connect your AI provider.
Windows: Use Python 3.10 or 3.11 for GPU training
Python 3.12/3.13 on Windows lacks GPU support for ML training due to TensorFlow's deprecation of native Windows GPU builds. For GPU-accelerated training on Windows, use Python 3.10/3.11 — or install under WSL2 with Ubuntu (see the Windows (WSL) tab above), which has full GPU support on any Python version.
Choosing Your Python Version#
The installer handles this automatically. If installing manually, check your version and use the matching wheel:
| Python Version | Wheel Suffix | Windows |
|---|---|---|
| 3.10 | cp310-cp310 |
✅ Recommended |
| 3.11 | cp311-cp311 |
✅ Recommended |
| 3.12 | cp312-cp312 |
⚠️ CPU-only (no GPU training) |
| 3.13 | cp313-cp313 |
⚠️ CPU-only (no GPU training) |
Install Extras#
The base install gives you the full agentic platform — chat, autonomous agents, persistent channels, and all AI provider integrations. ML and data science features are optional extras that you can add when needed:
| Extra | What it adds | Who needs it |
|---|---|---|
[ml] |
TensorFlow, OpenCV, scikit-learn, scipy, pandas, LangChain | Deep learning studio and ML pipeline users |
[trading] |
pandas-ta technical analysis indicators | Trading feature users |
[stats] |
tdigest streaming percentiles | Stats/analytics users |
[full] |
All of the above | Full platform deployments |
The lightweight install is faster, uses less disk, and avoids C compiler requirements on Windows. If you're just using LIT for chat and AI agents, the base install is all you need.
First-Time Setup#
When you first open LIT Platform, the setup wizard will guide you through connecting an AI provider:
Option A: Claude (Anthropic)#
Requires a Claude Pro or Max subscription and Node.js.
First, install Node.js from nodejs.org (LTS version), then install the Claude CLI:
Then in the LIT setup wizard:
- Click the Claude card
- Click the button — your browser will open to sign in at claude.ai
- After signing in, copy the authentication code shown in the browser
- Paste the code into the field and click Save
Option B: Gemini (Free)#
No additional software required — just a Google account.
- Click the Gemini card
- Click the button to open Google AI Studio
- Create a free API key
- Paste the key and click Save
Option C: ChatGPT (OpenAI)#
Requires an OpenAI account and Node.js.
First, install Node.js from nodejs.org (LTS version), then install the Codex CLI:
Then in the LIT setup wizard:
- Click the ChatGPT card
- Click Authenticate with ChatGPT — the wizard will display a URL and a short device code
- Open the URL in your browser and enter the code
- Authentication completes automatically — no need to paste anything back
Verify Installation#
Once setup is complete, you should see the chat interface.
Try sending a message:
Hello! Can you tell me what platform you're running on?
The AI should respond and you're ready to go!
Updating#
Run the same installer again — it detects your existing installation and upgrades to the latest version:
To install a specific version:
Your settings and agents are stored separately in ~/.config/lit/ and are preserved across updates.
Troubleshooting#
Port already in use#
If port 8080 is already in use:
Windows: "lit" is not recognized#
Use python -m lit serve instead. The pip Scripts directory isn't in your PATH.
Import errors on startup#
Make sure you're using the wheel that matches your Python version. Run python --version to check.
Windows: No GPU acceleration#
TensorFlow dropped native Windows GPU support after version 2.11. On Windows with Python 3.12 or 3.13, ML training will run on CPU only. Options:
- Downgrade to Python 3.10 or 3.11 — best compatibility for GPU training on Windows
- Use WSL2 + Ubuntu — full GPU support, recommended for serious ML work (see the Windows (WSL) install tab above)
Windows: WSL2 install times out (error 0x800705b4)#
wsl --install -d Ubuntu fails with:
This operation returned because the timeout period expired.
Error code: Wsl/InstallDistro/Service/RegisterDistro/CreateVm/0x800705b4
WSL2 runs its own lightweight VM, and this error means that VM can't launch. Task Manager showing "Virtualization: Enabled" is not sufficient — that only means the CPU flag is visible, not that the hypervisor actually booted. Work through these in order:
-
Bare metal: if Task Manager → Performance → CPU shows
Virtualization: Disabled, enable VT-x (Intel) or AMD-V/SVM in your BIOS/UEFI settings. -
Hypervisor not launching: in an administrator PowerShell:
Reboot, then verify with
systeminfo | findstr /i hyper— you want to see "A hypervisor has been detected." If you instead see a requirements checklist, the hypervisor still isn't running. -
Windows is itself a virtual machine (VMware, VirtualBox, KVM/QEMU, Hyper-V): WSL2 needs nested virtualization enabled on the host hypervisor. If you can't enable it — or WSL2 still times out — fall back to WSL1, which needs no VM at all:
WSL1 runs the platform itself fine (and is often faster inside a VM), but with caveats: no GPU training, no systemd, and the Claude backend won't work — Claude Code requires WSL2 or newer and refuses to install on WSL1. Under WSL1, use the Gemini backend (API key only, no Node.js needed). If you need Claude and can't run WSL2, install LIT natively on Windows instead — see the Windows (PowerShell) tab.
License#
The pip install is a fully functional platform — not a limited trial or demo. You get the complete single-user experience: chat, autonomous agents, persistent channels, and (with the [full] extra) the deep learning studio. Personal and non-commercial use is free, permanently.
What the pip install doesn't include is business infrastructure:
- Multi-user access and team administration
- SSO / Keycloak authentication
- Shared team channels across users
- On-premises support and SLAs
- Enterprise support
Think of it as try-before-you-buy: evaluate the complete platform at no cost, and when you're ready to deploy to a team or use it in a commercial context, contact us for commercial licensing.
Commercial use — deploying LIT Platform within a business, as part of a product or service, or in any revenue-generating context — requires a commercial license. View licensing options →
Next Steps#
- Explore the chat interface and try different AI models
- Check out Creating a Model from Scratch for a data science walkthrough
- Read about Human-AI Collaboration best practices
- Windows users: Run LIT Platform automatically at startup