review changes

Author: meffmadd

scripts/README.md

@@ -22,7 +22,6 @@ scripts/
   cleanup_recordings.py        # Remove orphaned test recordings
   diagnose_recordings.py       # Debug test recording issues
   normalize_recordings.py      # Normalize test recordings for consistency
-  gen_inference_store_disabled_recordings.py # Regenerate recordings for the inference-store-disabled integration test
   docker.sh                    # Docker build helper
   install.sh                   # Installation helper
   integration-tests.sh         # Run integration test suite
@@ -67,9 +66,6 @@ uv run python scripts/cleanup_recordings.py
 uv run python scripts/diagnose_recordings.py
 # Normalize recordings
 uv run python scripts/normalize_recordings.py
-# Regenerate recordings for the inference-store-disabled integration test
-# (records against a local mock OpenAI server; isolates the shared recordings dir)
-uv run python scripts/gen_inference_store_disabled_recordings.py
 ```
 
 ### Remote test recording (via GitHub Actions)

scripts/gen_inference_store_disabled_recordings.py

@@ -60,11 +60,6 @@ async def get_auto_router_impl(
     api_to_dep_impl = {}
     # TODO: move pass configs to routers instead
     if api == Api.inference:
-        # An absent inference store reference disables chat completion
-        # persistence: no store is constructed, no table is created, and no
-        # background write workers are started. The inference router handles a
-        # missing store (it guards every write and raises NotImplementedError
-        # on the history endpoints), mirroring the optional Responses store.
         inference_ref = run_config.storage.stores.inference
         if inference_ref is not None:
             inference_store = InferenceStore(

src/ogx/core/routers/__init__.py

@@ -55,28 +55,16 @@ Storage is configured in `StackConfig.storage` via `StorageConfig`. The `stores`
 
 See `datatypes.py` for all config types and `StorageBackendType` for the enum of supported backends.
 
-### Optional Stores (null to disable)
-
-Some store references are optional: setting the reference to `null` (not omitting
-it) means OGX does not construct the store at all, and the API that depends on
-it degrades gracefully. This is an explicit operator choice made in a run config;
-the default for every optional reference remains an enabled reference (except
-`responses`, which defaults to `None`), so existing deployments are unaffected
-unless they opt in.
-
-- **`inference`** -- when `null`, no `InferenceStore` is constructed: no
-  `inference_store` table is created and no background write workers run. Chat
-  completions still work (streaming and non-streaming); the chat completion
-  history endpoints (`list`, `retrieve`, `messages`) report that persistence is
-  not configured (HTTP 501) rather than returning an empty list or a 404.
-- **`responses`** -- the `responses` reference is nullable in the config schema
-  (`default=None`) and `null` passes validation, but unlike `inference` it is not
-  a runtime toggle: the built-in responses provider always constructs and
-  initializes its store from its own `persistence.responses` (a required,
-  non-nullable reference), and the shared `storage.stores.responses` reference
-  is only validated, never consumed at startup. Setting it to `null` is accepted
-  but does not currently disable Responses persistence.
-
-Other stores (`datasets`, `eval`, `files`, `prompts`, `vector_io`) are not
-affected by disabling the inference store, so persistence can be turned off for
-one API independently of the rest of the storage layer.
+### Inference Store (null to disable)
+
+Setting the `inference` store reference to `null` explicitly disables Chat
+Completions persistence. Omitting the reference keeps the store enabled for
+backward compatibility.
+
+When disabled, no `InferenceStore` is constructed, no `inference_store` table is
+created, and no background write workers run. Streaming and non-streaming Chat
+Completions continue to work. The history endpoints (`list`, `retrieve`, and
+`messages`) report that persistence is not configured (HTTP 501).
+
+Other stores (`responses`, `datasets`, `eval`, `files`, `prompts`, `vector_io`)
+are unaffected by disabling the inference store.

src/ogx/core/storage/README.md

@@ -73,6 +73,5 @@ completions still work for both streaming and non-streaming requests; the
 history endpoints report that persistence is not configured (HTTP 501) rather
 than returning an empty list or a 404. Other stores (`responses`, `datasets`,
 `eval`, `files`, `prompts`, `vector_io`) stay enabled, so disabling inference
-persistence is independent of the rest of the storage layer. This follows the
-same optional-store pattern the Responses store already uses; see the storage
+persistence is independent of the rest of the storage layer. See the storage
 module README for details.