remove deprecation warning from byo-llm

Signed-off-by: Jordan Dubrick <jdubrick@redhat.com>

Author: Jdubrick

docs/design/llama-stack-config-merge/llama-stack-config-merge.md

@@ -36,8 +36,7 @@ Key shape:
 - `llama_stack.config.baseline: default | byo-llm | empty` — pick
   LCORE's built-in baseline (includes a conditional OpenAI provider),
   the same baseline without that OpenAI row, or an empty dict (used by
-  the migration tool for exact round-trip). `default` emits a deprecation
-  WARN naming `byo-llm`.
+  the migration tool for exact round-trip).
 - Legacy two-file mode (`llama_stack.library_client_config_path` +
   external `run.yaml`) is preserved during a deprecation window;
   mutually exclusive with the unified *synthesis inputs* (a non-empty
@@ -459,8 +458,7 @@ September 2026.
    Else if `unified` and `unified.baseline == "empty"` → `{}`. Else →
    `default_baseline` arg or `load_default_baseline()`. If the selector
    is `byo-llm`, strip the built-in conditional OpenAI inference row.
-   If the selector is `default` or omitted, emit a deprecation WARN
-   naming `byo-llm`. `empty` and `profile:` are unchanged.
+   `empty` and `profile:` are unchanged.
 3. Run `dedupe_providers_vector_io` on the baseline.
 4. Apply existing enrichment: `enrich_byok_rag`, `enrich_solr` (Azure
    Entra ID intentionally stays separate because it's a `.env`
@@ -558,7 +556,7 @@ reference.
 |---|---|---|
 | 2026-04-23 | Initial version | Spike completion |
 | 2026-08-20 | Default baseline openai provider is conditional on `OPENAI_API_KEY` | LCORE-3607: `baseline: default` must load when the key is unset |
-| 2026-08-21 | Add `baseline: byo-llm` (default_run.yaml minus the OpenAI row); WARN on `default`/omitted | LCORE-3654: opt-in openai-free baseline |
+| 2026-08-21 | Add `baseline: byo-llm` (default_run.yaml minus the OpenAI row) | LCORE-3654: opt-in openai-free baseline |
 
 ## Appendix A — Worked example: legacy → unified migration
 

src/llama_stack_configuration.py

@@ -1252,19 +1252,9 @@ def synthesize_configuration(  # pylint: disable=too-many-locals
     ls_config: dict[str, Any] = copy.deepcopy(baseline)
 
     # Profile and empty are unchanged. The shipped file either keeps OpenAI
-    # (default/omitted, with a deprecation WARN) or drops it (byo-llm).
-    if loaded_shipped_baseline:
-        if unified and unified.get("baseline") == "byo-llm":
-            _strip_default_openai_inference(ls_config)
-        else:
-            logger.warning(
-                "DEPRECATED: llama_stack.config.baseline 'default' includes a "
-                "built-in conditional OpenAI inference provider (%s). Set "
-                "baseline to 'byo-llm' to start without that provider and "
-                "declare LLMs under inference.providers. 'byo-llm' will "
-                "become the default in a future release.",
-                CONDITIONAL_OPENAI_PROVIDER_ID,
-            )
+    # (default/omitted) or drops it (byo-llm).
+    if loaded_shipped_baseline and unified and unified.get("baseline") == "byo-llm":
+        _strip_default_openai_inference(ls_config)
 
     # 3. Normalize duplicated vector_io providers in the baseline.
     dedupe_providers_vector_io(ls_config)

tests/unit/test_llama_stack_synthesize.py

@@ -8,7 +8,6 @@
 
 # pylint: disable=too-many-lines
 
-import logging
 import os
 import stat
 from pathlib import Path
@@ -741,15 +740,6 @@ def test_synthesize_vllm_appends_and_keeps_conditional_openai(
     assert any(entry["provider_id"] == "vllm" for entry in _inference_entries(resolved))
 
 
-def _byo_llm_deprecation_warnings(caplog: pytest.LogCaptureFixture) -> list[str]:
-    """Return WARN records that name the byo-llm baseline replacement."""
-    return [
-        record.getMessage()
-        for record in caplog.records
-        if record.levelno == logging.WARNING and "byo-llm" in record.getMessage()
-    ]
-
-
 @pytest.mark.parametrize(
     "lcs",
     [
@@ -758,43 +748,29 @@ def _byo_llm_deprecation_warnings(caplog: pytest.LogCaptureFixture) -> list[str]
         {},
     ],
 )
-def test_synthesize_default_path_warns_and_keeps_conditional_openai(
-    caplog: pytest.LogCaptureFixture, lcs: dict[str, Any]
+def test_synthesize_default_path_keeps_conditional_openai(
+    lcs: dict[str, Any],
 ) -> None:
-    """default or omitted baseline keeps the OpenAI row and warns naming byo-llm."""
-    with caplog.at_level(
-        "WARNING", logger="lightspeed_stack.llama_stack_configuration"
-    ):
-        result = synthesize_configuration(lcs)
+    """default or omitted baseline keeps the OpenAI row."""
+    result = synthesize_configuration(lcs)
     openai_entries = _openai_inference_entries(result)
     assert len(openai_entries) == 1
     assert openai_entries[0]["provider_id"] == OPENAI_CONDITIONAL_PROVIDER_ID
     ids = [entry["provider_id"] for entry in _inference_entries(result)]
     assert "sentence-transformers" in ids
-    warnings = _byo_llm_deprecation_warnings(caplog)
-    assert len(warnings) == 1
-    assert "byo-llm" in warnings[0]
 
 
-def test_synthesize_byo_llm_strips_openai_without_warn(
-    caplog: pytest.LogCaptureFixture,
-) -> None:
-    """byo-llm drops the built-in OpenAI row, keeps the embedder, and does not warn."""
+def test_synthesize_byo_llm_strips_openai() -> None:
+    """byo-llm drops the built-in OpenAI row and keeps the embedder."""
     lcs = {"llama_stack": {"config": {"baseline": "byo-llm"}}}
-    with caplog.at_level(
-        "WARNING", logger="lightspeed_stack.llama_stack_configuration"
-    ):
-        result = synthesize_configuration(lcs)
+    result = synthesize_configuration(lcs)
     assert _openai_inference_entries(result) == []
     ids = [entry["provider_id"] for entry in _inference_entries(result)]
     assert ids == ["sentence-transformers"]
     assert "model-context-protocol" in _tool_runtime_ids(result)
-    assert _byo_llm_deprecation_warnings(caplog) == []
 
 
-def test_synthesize_byo_llm_with_vllm_appends_without_openai(
-    caplog: pytest.LogCaptureFixture,
-) -> None:
+def test_synthesize_byo_llm_with_vllm_appends_without_openai() -> None:
     """byo-llm + high-level vLLM appends vLLM and does not restore OpenAI."""
     lcs = {
         "llama_stack": {"config": {"baseline": "byo-llm"}},
@@ -808,10 +784,7 @@ def test_synthesize_byo_llm_with_vllm_appends_without_openai(
             ]
         },
     }
-    with caplog.at_level(
-        "WARNING", logger="lightspeed_stack.llama_stack_configuration"
-    ):
-        result = synthesize_configuration(lcs)
+    result = synthesize_configuration(lcs)
     assert _openai_inference_entries(result) == []
     vllm = next(
         entry for entry in _inference_entries(result) if entry["provider_id"] == "vllm"
@@ -820,34 +793,25 @@ def test_synthesize_byo_llm_with_vllm_appends_without_openai(
     assert "sentence-transformers" in [
         entry["provider_id"] for entry in _inference_entries(result)
     ]
-    assert _byo_llm_deprecation_warnings(caplog) == []
 
 
-def test_synthesize_byo_llm_with_openai_appends_one_row(
-    caplog: pytest.LogCaptureFixture,
-) -> None:
-    """byo-llm + high-level openai appends a single openai row and does not warn."""
+def test_synthesize_byo_llm_with_openai_appends_one_row() -> None:
+    """byo-llm + high-level openai appends a single openai row."""
     lcs = {
         "llama_stack": {"config": {"baseline": "byo-llm"}},
         "inference": {
             "providers": [{"type": "openai", "api_key_env": "OPENAI_API_KEY"}]
         },
     }
-    with caplog.at_level(
-        "WARNING", logger="lightspeed_stack.llama_stack_configuration"
-    ):
-        result = synthesize_configuration(lcs)
+    result = synthesize_configuration(lcs)
     openai_entries = _openai_inference_entries(result)
     assert len(openai_entries) == 1
     assert openai_entries[0]["provider_id"] == "openai"
     assert openai_entries[0]["config"]["api_key"] == "${env.OPENAI_API_KEY}"
-    assert _byo_llm_deprecation_warnings(caplog) == []
 
 
-def test_synthesize_empty_baseline_does_not_warn_byo_llm(
-    caplog: pytest.LogCaptureFixture,
-) -> None:
-    """baseline: empty is unchanged: no strip, no byo-llm WARN."""
+def test_synthesize_empty_baseline_does_not_strip_openai() -> None:
+    """baseline: empty is unchanged: no OpenAI strip."""
     lcs = {
         "llama_stack": {
             "config": {
@@ -856,18 +820,12 @@ def test_synthesize_empty_baseline_does_not_warn_byo_llm(
             }
         }
     }
-    with caplog.at_level(
-        "WARNING", logger="lightspeed_stack.llama_stack_configuration"
-    ):
-        result = synthesize_configuration(lcs)
+    result = synthesize_configuration(lcs)
     assert result == {"version": 2, "apis": ["inference"]}
-    assert _byo_llm_deprecation_warnings(caplog) == []
 
 
-def test_synthesize_profile_ignores_byo_llm_and_does_not_warn(
-    tmp_path: Path, caplog: pytest.LogCaptureFixture
-) -> None:
-    """profile: wins over baseline: byo-llm; no strip and no deprecation WARN."""
+def test_synthesize_profile_ignores_byo_llm(tmp_path: Path) -> None:
+    """profile: wins over baseline: byo-llm; no OpenAI strip."""
     profile = {
         "version": 2,
         "apis": ["inference"],
@@ -891,15 +849,11 @@ def test_synthesize_profile_ignores_byo_llm_and_does_not_warn(
             }
         }
     }
-    with caplog.at_level(
-        "WARNING", logger="lightspeed_stack.llama_stack_configuration"
-    ):
-        result = synthesize_configuration(lcs, config_file_dir=str(tmp_path))
+    result = synthesize_configuration(lcs, config_file_dir=str(tmp_path))
     assert result["marker"] == "from-profile"
     openai_entries = _openai_inference_entries(result)
     assert len(openai_entries) == 1
     assert openai_entries[0]["provider_id"] == "openai"
-    assert _byo_llm_deprecation_warnings(caplog) == []
 
 
 def test_synthesize_loads_profile_relative_to_config_dir(tmp_path: Path) -> None: