This PR adds the embedding implementation for golem-llm, including:

✅ Basic embedding generation (tested)

✅ Support for multiple inputs

✅ Support for custom dimensions (e.g., 256)

✅ Proper error handling (invalid model tested)

✅ Support for multiple output formats (float & base64)

All tests run via ./run_embed_tests.sh have passed successfully. Logs confirm that the implementation meets the bounty requirements as specified.

Example logs:

yaml

✅ Basic embedding generation successful ✅ Multiple inputs test successful ✅ Custom dimensions: 256 ✅ Error handling: returned expected 404 ✅ Base64 format: successfully decoded /claim #16