docs(identity): drop Phase 1/2 framing from module docstrings

.gitignore

@@ -3,6 +3,7 @@
 .venv_policy_test
 .env
 .claude
+.claude-team-notes/
 .newenv
 newenv/*
 litellm/proxy/myenv/*

litellm/identity/__init__.py

@@ -1,13 +1,13 @@
 """Caller-identity module.
 
-Phase 1: domain types + extractors + bidirectional adapter to
+Domain types, the identity cache/store, and a bidirectional adapter to
 ``UserAPIKeyAuth``. The proxy still drives identity through
 ``litellm/proxy/auth/`` today; this module is the new home those flows
-will migrate to.
+migrate to.
 
 The public surface is small on purpose; downstream code should depend on
 ``IdentityContext`` and the ``Principal`` union, not on individual
-extractor internals.
+internals.
 """
 
 from litellm.identity.cache import IdentityCache

litellm/identity/adapter.py

@@ -1,8 +1,8 @@
 """Bidirectional bridge between ``IdentityContext`` and ``UserAPIKeyAuth``.
 
-Phase 1 keeps the legacy Pydantic model as the universal carrier. These
-two pure functions let new code work in terms of ``IdentityContext``
-without forcing call sites to migrate today.
+The legacy Pydantic model stays the universal carrier. These two pure
+functions let new code work in terms of ``IdentityContext`` without
+forcing call sites to migrate today.
 
 Invariants:
 - ``identity_context_to_user_api_key_auth(uak.to_identity_context())``

litellm/identity/context.py

@@ -1,9 +1,9 @@
 """The per-request identity bundle.
 
 ``IdentityContext`` is what downstream consumers (auth, spend, guardrails,
-logging, audit) should read identity from once Phase 2 migration is done.
-In Phase 1 it travels alongside the legacy ``UserAPIKeyAuth`` via the
-adapter functions in ``litellm.identity.adapter``.
+logging, audit) should read identity from. Today it travels alongside the
+legacy ``UserAPIKeyAuth`` via the adapter functions in
+``litellm.identity.adapter``.
 
 The bundle is mutable on purpose: identity fields like ``end_user_id`` are
 sometimes resolved or overridden after initial extraction, and the

litellm/identity/invalidation.py

@@ -12,9 +12,9 @@ Two flavors:
   a mismatch as a miss.
 
 The legacy ``_delete_cache_key_object`` and the per-table cache deletes
-in ``auth_checks.py`` are left in place by Phase 2. These hooks run
-side-by-side so we don't strand a partially-deployed fleet that's still
-reading from the legacy cache keys.
+in ``auth_checks.py`` stay in place. These hooks run side-by-side so we
+don't strand a partially-deployed fleet that's still reading from the
+legacy cache keys.
 """
 
 from __future__ import annotations

litellm/identity/principal.py

@@ -3,7 +3,7 @@
 A ``Principal`` answers "who is making this request" using only the fields
 that uniquely identify the caller. Per-row enrichment (budgets, team rows,
 object permissions) is intentionally not modeled here; that data continues
-to ride on ``UserAPIKeyAuth`` in Phase 1.
+to ride on ``UserAPIKeyAuth``.
 
 Each subtype is a frozen dataclass with a ``kind`` discriminator suitable
 for ``match``-style dispatch.