fix(pat): align replay scopes, keep 404/405, retire expired names, document the PAT_ENABLED switch

The ASGI message events route now accepts the same scopes as its Flask
sibling (conversations:read or chat:run) through a shared constant. A token
request that fails routing gets Flask's 404/405 instead of a 403. Creating a
token retires an expired token that still held the name. allowed_ids uses
is_pat instead of a bare literal. PAT_ENABLED is documented as the master
switch it is: turning it off stops every existing token from authenticating.
The docs explain that sources are matched by name (oldest wins) and point CI
flows at sources upload --replace.

docsgpt/api/asgi_auth.py

@@ -9,7 +9,7 @@ from __future__ import annotations
 
 import uuid
 from contextvars import Token
-from typing import Optional, Tuple
+from typing import Optional, Sequence, Tuple, Union
 
 import anyio
 from starlette.requests import Request
@@ -23,13 +23,13 @@ from docsgpt.core.settings import settings
 
 
 async def authenticate(
-    request: Request, *, pat_scope: Optional[str] = None
+    request: Request, *, pat_scope: Union[str, Sequence[str], None] = None
 ) -> Tuple[Optional[dict], Optional[JSONResponse]]:
     """Decode the caller's JWT the way Flask's ``authenticate_request`` does.
 
     Args:
         request: The incoming Starlette request.
-        pat_scope: Scope a personal access token needs for this route. Left
+        pat_scope: Scope (or any-of scopes) a personal access token needs for this route. Left
             unset, the route rejects PATs outright (deny by default, matching
             the Flask rule table in ``docsgpt/api/pat/rules.py``). A token with
             a resource filter is always rejected.
@@ -48,7 +48,8 @@ async def authenticate(
     if is_pat(decoded):
         # A PAT lookup already excludes revoked tokens and deactivated users,
         # so the session denylist below does not apply to it.
-        if pat_scope is None or pat_scope not in (decoded.get("scopes") or []):
+        accepted = (pat_scope,) if isinstance(pat_scope, str) else tuple(pat_scope or ())
+        if not set(accepted).intersection(decoded.get("scopes") or []):
             return None, JSONResponse(
                 {"success": False, "message": "Token lacks the required scope", "error": "insufficient_scope"},
                 status_code=403,

docsgpt/api/async_sse.py

@@ -25,6 +25,7 @@ from starlette.responses import Response
 from starlette.routing import Route
 
 from docsgpt.api.asgi_auth import authenticate, bind_log_context, json_error
+from docsgpt.api.pat.rules import MESSAGE_REPLAY_SCOPES
 from docsgpt.api.asgi_stream import sse_response
 from docsgpt.core.settings import settings
 from docsgpt.storage.db.session import db_readonly
@@ -94,7 +95,8 @@ async def stream_message_events(request: Request) -> Response:
     """
     # Same JWT decoder and OIDC revocation check as the Flask routes. With
     # AUTH_TYPE unset the caller resolves to ``{"sub": "local"}``.
-    decoded, error = await authenticate(request, pat_scope="chat:run")
+    # Same scopes as its Flask sibling GET /api/messages/<id>/tail.
+    decoded, error = await authenticate(request, pat_scope=MESSAGE_REPLAY_SCOPES)
     if error is not None:
         return error
     user_id = decoded.get("sub") if isinstance(decoded, dict) else None

docsgpt/api/pat/routes.py

@@ -139,6 +139,7 @@ class PersonalAccessTokens(Resource):
                     return _error(
                         f"Token limit reached ({settings.PAT_MAX_PER_USER}); revoke one first", 409
                     )
+                repo.retire_expired_name(user_id, name)
                 if repo.name_in_use(user_id, name):
                     return _error("A token with this name already exists", 409)
                 row = repo.create(

docsgpt/api/pat/rules.py

@@ -25,6 +25,8 @@ import uuid
 from dataclasses import dataclass
 from typing import Any, Callable, Iterable, Optional
 
+from docsgpt.api.pat.tokens import is_pat
+
 VIEW, QUERY, JSON, FORM, BODY = "view", "query", "json", "form", "body"
 
 Locator = tuple[str, str]
@@ -53,6 +55,11 @@ def _rule(scope: Optional[str] = None, *ids: Locator, any_of: tuple[str, ...] =
     return Rule(scopes=scopes, family=family, ids=tuple(ids), **kwargs)
 
 
+#: Scopes that admit a token to message replay. Shared by the Flask tail route
+#: below and its ASGI sibling GET /api/messages/<id>/events (docsgpt/api/async_sse.py),
+#: which sits outside this table.
+MESSAGE_REPLAY_SCOPES = ("conversations:read", "chat:run")
+
 _ALL_FAMILIES = ("agents", "sources", "prompts", "tools", "workflows")
 
 # Ids of other families that agent create/update accept in their JSON-or-form body.
@@ -220,7 +227,7 @@ RULES: dict[tuple[str, str], Rule] = {
     ("/api/get_single_conversation", "GET"): _rule("conversations:read", blocked_by=("agents",)),
     # A message cannot be tied to an allowlist from here, so any restricted token is kept out.
     ("/api/messages/<string:message_id>/tail", "GET"): _rule(
-        any_of=("conversations:read", "chat:run"), family=None, blocked_by=_ALL_FAMILIES
+        any_of=MESSAGE_REPLAY_SCOPES, family=None, blocked_by=_ALL_FAMILIES
     ),
     ("/api/delete_conversation", "POST"): _rule("conversations:write", blocked_by=("agents",)),
     ("/api/delete_all_conversations", "GET"): _rule("conversations:write", blocked_by=("agents",)),
@@ -360,7 +367,11 @@ def _all_allowed(ids: Iterable[str], allowed: Iterable[str]) -> bool:
 def authorize(request, decoded_token: dict) -> Optional[tuple[dict, int]]:
     """Check a PAT request against the table. ``None`` allows; otherwise ``(body, status)``."""
     url_rule = getattr(request, "url_rule", None)
-    rule = RULES.get((url_rule.rule, request.method)) if url_rule is not None else None
+    if url_rule is None:
+        # Routing failed (unknown path or wrong method): no view will run, so
+        # let Flask answer 404/405 instead of masking it with a 403.
+        return None
+    rule = RULES.get((url_rule.rule, request.method))
     if rule is None:
         return (
             {
@@ -420,8 +431,8 @@ def allowed_ids(request, family: str) -> Optional[set[str]]:
 
     Used by listing routes (``listing=True``) and by ``in_route`` handlers.
     """
-    decoded = getattr(request, "decoded_token", None) or {}
-    if decoded.get("auth_method") != "pat":
+    decoded = getattr(request, "decoded_token", None)
+    if not is_pat(decoded):
         return None
     ids = (decoded.get("resource_filter") or {}).get(family)
     if ids is None: