Account
SANGAM CONSOLE

Namaste, friend.

Lend a laptop, run any model, train a LoRA, explore the marketplace.

Sangam credits —
My nodes — 0 online
Inferences — 0 credits spent
LoRAs — 0 on marketplace

Quick start

  1. Step 1 — Register a node to lend your laptop and receive a worker token.
  2. Step 2 — Download the Sangam node-client, paste the token, and let it run in the background.
  3. Step 3 — Open the Chat & Inference tab and try Llama 3 70B, Claude Opus, or Sarvam-1.
  4. Step 4 — Fine-tune your own LoRA, or buy one someone else trained from the Marketplace.

Your swarm nodes

Run the node-client

The Sangam node-client is a small Python program that runs on macOS, Linux, and Windows. After registering above, paste your worker token and run it. 

git clone https://github.com/Glixy-Labs/glixy_labs_1
cd glixy_labs_1/scripts/sangam-node
# macOS / Linux:
SANGAM_API=http://localhost:3001 \
SANGAM_NODE_ID=<your-node-id> \
SANGAM_TOKEN=<your-worker-token> \
  python3 worker.py
# Windows (PowerShell):
$env:SANGAM_API="http://localhost:3001"
$env:SANGAM_NODE_ID="<node-id>"
$env:SANGAM_TOKEN="<token>"
python worker.py
Browse hub →
Ask anything — community models route through the Sangam swarm; frontier models (Claude, GPT-4o, Gemini) route direct to the vendor. English, Hindi, any Indian language.

MVP note: community models use a vendor-fallback router until the full Petals fork ships. Frontier models (Claude Opus, GPT-4o, Gemini 2.5 Pro) route directly to the upstream vendor.

Start a new LoRA job

Cost preview: —

Active & recent training jobs

My LoRAs

List one for sale and earn credits whenever someone uses it.

Credit ledger

Every credit movement is recorded here — node uptime, inference, training, marketplace.

WhenReasonΔRef