diff --git a/architecture/client.md b/architecture/client.md
index 73c446e..74001a2 100644
--- a/architecture/client.md
+++ b/architecture/client.md
@@ -4,7 +4,7 @@
 
 ## The internal terminal
 
-The bottom of the middleware chain (the "terminal") is internal. It calls `self._httpx2_client.send(request)`, maps `httpx2` errors to `httpware` errors, and raises a `StatusError` subclass on 4xx/5xx. The error-mapping table (what `httpx2` exception maps to which `httpware` exception) lives at the terminal in `src/httpware/client.py`; status-keyed exceptions are looked up via the `STATUS_TO_EXCEPTION` table in `src/httpware/errors.py`. The same terminal lifecycle holds in both worlds — `Client.send` calls `httpx2.Client.send`, `AsyncClient.send` calls `httpx2.AsyncClient.send`.
+The bottom of the middleware chain (the "terminal") is internal. It calls `self._httpx2_client.send(request)`, maps `httpx2` errors to `httpware` errors, and raises a `StatusError` subclass on 4xx/5xx. The error-mapping table (what `httpx2` exception maps to which `httpware` exception) lives at the terminal in `src/httpware/client.py`; status-keyed exceptions are looked up via the `STATUS_TO_EXCEPTION` table in `src/httpware/errors.py`. The same terminal lifecycle holds in both worlds: `Client.send` / `AsyncClient.send` enter the middleware chain first, and it is the internal terminal — `Client._terminal` / `AsyncClient._terminal` — that calls `httpx2.Client.send` / `httpx2.AsyncClient.send`.
 
 ## Sync/async parity
 
@@ -14,7 +14,7 @@ The async middleware surface uses the `Async*`/`async_*` prefix, aligning with h
 
 ## Streaming
 
-`AsyncClient.stream()` provides a context-manager API for chunked response bodies. It bypasses the middleware chain by design.
+Both `Client.stream()` (sync) and `AsyncClient.stream()` (async) provide a context-manager API for chunked response bodies. Both bypass the middleware chain by design.
 
 ## Proxy environment (`trust_env`)
 
diff --git a/architecture/extras.md b/architecture/extras.md
index 2cb34ba..1780a06 100644
--- a/architecture/extras.md
+++ b/architecture/extras.md
@@ -15,7 +15,7 @@ A protocol seam is a documented internal boundary. AI agents and contributors mu
 
 ```toml
 [project.optional-dependencies]
-pydantic = ["pydantic>=2"]
+pydantic = ["pydantic>=2.0,<3.0"]
 msgspec = ["msgspec>=0.18"]
 ```
 
diff --git a/docs/middleware.md b/docs/middleware.md
index 0549de0..0ce2a3b 100644
--- a/docs/middleware.md
+++ b/docs/middleware.md
@@ -38,7 +38,7 @@ Whatever you do, return an `httpx2.Response`. Raising an exception propagates up
 For the common cases where you don't need state-keeping on `self` and don't need to wrap the full `await next(...)` call, `httpware.middleware` exports three decorators that turn a single async function into an `AsyncMiddleware`:
 
 ```python
-from httpware.middleware import async_before_request, async_after_response, async_on_error
+from httpware import async_before_request, async_after_response, async_on_error
 ```
 
 | Decorator | Function signature | When to use |
@@ -64,7 +64,7 @@ import uuid
 import httpx2
 
 from httpware import AsyncClient, AsyncRetry
-from httpware.middleware import AsyncNext
+from httpware import AsyncNext
 
 
 _LOGGER = logging.getLogger("myapp.request_id")
@@ -152,7 +152,7 @@ import logging
 import httpx2
 
 from httpware import Client
-from httpware.middleware import Next
+from httpware import Next
 
 
 _LOGGER = logging.getLogger("myapp.logging_middleware")
diff --git a/docs/recipes/phase-decorator-patterns.md b/docs/recipes/phase-decorator-patterns.md
index ff6b432..81a0549 100644
--- a/docs/recipes/phase-decorator-patterns.md
+++ b/docs/recipes/phase-decorator-patterns.md
@@ -14,7 +14,7 @@ The smallest useful case — add a static `Authorization` header to every outgoi
 import httpx2
 
 from httpware import AsyncClient
-from httpware.middleware import async_before_request
+from httpware import async_before_request
 
 
 @async_before_request
@@ -43,7 +43,7 @@ import contextvars
 import httpx2
 
 from httpware import AsyncClient, AsyncRetry
-from httpware.middleware import async_before_request
+from httpware import async_before_request
 
 
 _CORRELATION_ID: contextvars.ContextVar[str | None] = contextvars.ContextVar(
@@ -83,7 +83,7 @@ from collections.abc import Callable
 import httpx2
 
 from httpware import AsyncClient
-from httpware.middleware import AsyncMiddleware, async_after_response
+from httpware import AsyncMiddleware, async_after_response
 
 
 MetricSink = Callable[[str, int], None]
@@ -126,8 +126,8 @@ When the upstream is unreachable, return a synthesized 503 with a sentinel heade
 import httpx2
 
 from httpware import AsyncClient
-from httpware.errors import NetworkError
-from httpware.middleware import async_on_error
+from httpware import NetworkError
+from httpware import async_on_error
 
 
 @async_on_error
diff --git a/planning/changes/active/2026-06-14.06-audit-doc-fixes/change.md b/planning/changes/active/2026-06-14.06-audit-doc-fixes/change.md
new file mode 100644
index 0000000..68c0da1
--- /dev/null
+++ b/planning/changes/active/2026-06-14.06-audit-doc-fixes/change.md
@@ -0,0 +1,31 @@
+---
+status: draft
+date: 2026-06-14
+slug: audit-doc-fixes
+supersedes: null
+superseded_by: null
+pr: null
+outcome: null
+---
+
+# Change: Deep-audit documentation accuracy fixes
+
+**Lane:** docs/docstring-only; spec is the [2026-06-14 deep audit](../../../audits/2026-06-14-deep-audit.md).
+
+## Goal
+
+Close the remaining confirmed documentation-accuracy findings.
+
+## Findings
+
+- **M4** — `architecture/client.md` streaming section omitted `Client.stream()` → now documents both `Client.stream()` and `AsyncClient.stream()`.
+- **L13** — `architecture/client.md` attributed the `httpx2.*.send` call to `Client.send`/`AsyncClient.send` → corrected to the internal `_terminal` (and `.send()` enters the chain first).
+- **Nit4** — `src/httpware/errors.py` module docstring attributed auto-raise to `AsyncClient`'s terminal only → now lists all four raise sites (both terminals + both `stream()` methods).
+- **Nit15** — `architecture/extras.md` showed `pydantic>=2` without the `<3.0` ceiling → synced to `pydantic>=2.0,<3.0` (matches `pyproject.toml`).
+- **Nit14** — `docs/middleware.md` and `docs/recipes/phase-decorator-patterns.md` imported root-`__all__` symbols via submodule paths → standardized to `from httpware import X`.
+
+(Nit2 — `_is_streaming_body_async` reliance note — moot: PR #64 symmetrized the code instead.)
+
+## Verification
+
+- [ ] `just lint-ci` clean (CI-exact, no autofix); `just test` 100% (docstring change adds no lines).
diff --git a/src/httpware/errors.py b/src/httpware/errors.py
index 05d44e6..d14f255 100644
--- a/src/httpware/errors.py
+++ b/src/httpware/errors.py
@@ -1,6 +1,8 @@
 """Status-keyed exception hierarchy.
 
-Auto-raise rule lives at AsyncClient's internal terminal (see client.py).
+Auto-raise fires at four sites (all in client.py): both clients' internal
+terminals (Client._terminal / AsyncClient._terminal) and both stream() methods
+(Client.stream / AsyncClient.stream).
 Unknown 4xx falls back to ClientStatusError; unknown 5xx to ServerStatusError.
 The fallback assumes 400 <= status < 600.