API & Integration

1. Load Adapter on Demand

POST /load_adapter
Content-Type: application/json

{
  "adapter_name": "my_finetuned_model",
  "source": "huggingface"
}

• Loads a specific LoRA adapter from Hugging Face dynamically.

2. Merge Multiple Adapters

POST /merge_adapters
Content-Type: application/json

{
  "adapters": ["adapter_1", "adapter_2"],
  "merge_type": "ensemble"
}

Merges multiple adapters for improved model performance.

3. Run Inference

POST /generate
Content-Type: application/json

{
  "prompt": "Explain quantum computing in simple terms.",
  "adapter": "scientific_finetuned"
}

Runs inference with a selected fine-tuned adapter.

4. Unload Adapter

POST /unload_adapter
Content-Type: application/json

{
  "adapter_name": "scientific_finetuned"
}

• Unloads the adapter to free GPU memory.