feat: implement Inference + GenerateEmbedding endpoints (#42) #140
Reference in New Issue
Block a user
Delete Branch "feature/issue-42-inference-embedding-endpoints"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary\n- Implement unary
Inferenceendpoint: validates request, routes model, calls Ollama generate(), returns text + finish_reason + tokens_used\n- Implement unaryGenerateEmbeddingendpoint: validates request, resolves embedding model, calls Ollama embed(), returns embedding vector + dimensions\n- Both endpoints use model routing, audit logging, and consistent error mapping\n- Added 7 validation tests, removed 2 stale unimplemented stub tests\n\n## Test plan\n- [x] All 74 model-gateway tests pass\n- [x] Clippy clean (no warnings)\n- [x] Request validation covers missing params, missing context, empty prompt/text, empty session_id\n- [ ] Integration tests with mocked Ollama (deferred to issue #43)