POST /v1/embeddings returns 404 for loaded nomic-embed-text-v1.5

[horacebramwell]

Summary

POST /v1/embeddings returns HTTP 404 for a loaded nomic-embed-text-v1.5 embedding model on BaseRT 0.2.2.

Chat completions work with a concurrently loaded Gemma model.

Environment

• macOS on Apple Silicon
• BaseRT 0.2.2
• basecompute/nomic-embed-text-v1.5 Q4
• basecompute/gemma-4-E4B-it Q4
• Local OpenAI-compatible server with bearer authentication

Reproduction

basert pull basecompute/nomic-embed-text-v1.5
basert serve basecompute/gemma-4-E4B-it \
  --model /path/to/nomic-embed-text-v1.5-Q4.base \
  --api-key test-key \
  --port 8081

Then:

curl --fail --show-error \
  -H 'Authorization: Bearer test-key' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "nomic-embed-text-v1.5-Q4.base",
    "input": "BaseRT embedding check"
  }' \
  http://127.0.0.1:8081/v1/embeddings

Actual result

HTTP 404.

The BaseRT startup log reports both models loaded, including Nomic.

Expected result

A successful OpenAI-compatible embeddings response containing a 768-dimension, L2-normalized vector, as documented for this model.

Additional notes

• Retried after upgrading from 0.1.6 to 0.2.2.
• Also tried the hub-style model identifier basecompute/nomic-embed-text-v1.5; same 404.
• POST /v1/chat/completions succeeds with gemma-4-E4B-it.