🚀 ITP/TTP Organizational LLM Service

🤖 What you can do

• Point any OpenAI-compatible chat app or internal tool at the gateway and work with institute data behind your own login.
• Hand your coding harness a base URL and a key, so it can call the model from your editor or terminal.
• Summarise, extract, classify or clean data in batch jobs — a language-model step for an existing pipeline.
• Everything runs on hardware we run ourselves, so institute data stays inside the institute.

🔐 Quick Start

1. Sign in with your institute account in the box above.
2. Create a key in the key panel that appears after sign-in, then copy it once and store it like a password.
3. Point your tool at the gateway — base URL https://llm.particle.kit.edu/llm/, key sent as a bearer token:

curl https://llm.particle.kit.edu/llm/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"your-model-id","messages":[{"role":"user","content":"Hello"}]}'

🔑 Key management

• Sign-in issues one primary key; it is shown in the panel above and stays the same until you sign in again.
• Named keys are for one purpose each (editor, batch job, CI) and can be revoked individually without touching the primary key.
• Click any key value to select it, then copy it. Keys are shown in full only when they are created — keep them like a password.

🛡 Security & privacy

• Keys are personal. Never commit one, never paste one into a shared channel.
• Revoke a key from the panel above any time — it stops working immediately.
• If a key leaks, sign in again and rotate the primary key; the old value stops working.
• We do not store or process prompts or inference data in any way — nothing you send to the model is kept after the request is answered.

📖 Support

Questions, broken keys or feature requests:

• 👤 Bakar Chargeishvili