More actor refresh improvements

app/activitypub.py

@@ -6,7 +6,6 @@ from typing import Any
 
 import httpx
 from loguru import logger
-from markdown import markdown
 
 from app import config
 from app.config import ALSO_KNOWN_AS
@@ -14,6 +13,7 @@ from app.config import AP_CONTENT_TYPE  # noqa: F401
 from app.config import MOVED_TO
 from app.httpsig import auth
 from app.key import get_pubkey_as_pem
+from app.source import dedup_tags
 from app.source import hashtagify
 from app.utils.url import check_url
 
@@ -101,6 +101,19 @@ class VisibilityEnum(str, enum.Enum):
 
 
 _LOCAL_ACTOR_SUMMARY, _LOCAL_ACTOR_TAGS = hashtagify(config.CONFIG.summary)
+_LOCAL_ACTOR_METADATA = []
+if config.CONFIG.metadata:
+    for kv in config.CONFIG.metadata:
+        kv_value, kv_tags = hashtagify(kv.value)
+        _LOCAL_ACTOR_METADATA.append(
+            {
+                "name": kv.key,
+                "type": "PropertyValue",
+                "value": kv_value,
+            }
+        )
+        _LOCAL_ACTOR_TAGS.extend(kv_tags)
+
 
 ME = {
     "@context": AS_EXTENDED_CTX,
@@ -113,7 +126,7 @@ ME = {
     "outbox": config.BASE_URL + "/outbox",
     "preferredUsername": config.USERNAME,
     "name": config.CONFIG.name,
-    "summary": markdown(_LOCAL_ACTOR_SUMMARY, extensions=["mdx_linkify"]),
+    "summary": _LOCAL_ACTOR_SUMMARY,
     "endpoints": {
         # For compat with servers expecting a sharedInbox...
         "sharedInbox": config.BASE_URL
@@ -121,16 +134,7 @@ ME = {
     },
     "url": config.ID + "/",  # XXX: the path is important for Mastodon compat
     "manuallyApprovesFollowers": config.CONFIG.manually_approves_followers,
-    "attachment": [
-        {
-            "name": kv.key,
-            "type": "PropertyValue",
-            "value": markdown(kv.value, extensions=["mdx_linkify", "fenced_code"]),
-        }
-        for kv in config.CONFIG.metadata
-    ]
-    if config.CONFIG.metadata
-    else [],
+    "attachment": _LOCAL_ACTOR_METADATA,
     "icon": {
         "mediaType": mimetypes.guess_type(config.CONFIG.icon_url)[0],
         "type": "Image",
@@ -141,7 +145,7 @@ ME = {
         "owner": config.ID,
         "publicKeyPem": get_pubkey_as_pem(config.KEY_PATH),
     },
-    "tag": _LOCAL_ACTOR_TAGS,
+    "tag": dedup_tags(_LOCAL_ACTOR_TAGS),
 }
 
 if ALSO_KNOWN_AS:

app/actor.py

@@ -1,6 +1,7 @@
 import hashlib
 import typing
 from dataclasses import dataclass
+from datetime import timedelta
 from functools import cached_property
 from typing import Union
 from urllib.parse import urlparse
@@ -12,6 +13,8 @@ from sqlalchemy.orm import joinedload
 from app import activitypub as ap
 from app import media
 from app.database import AsyncSession
+from app.utils.datetime import as_utc
+from app.utils.datetime import now
 
 if typing.TYPE_CHECKING:
     from app.models import Actor as ActorModel
@@ -189,26 +192,64 @@ async def fetch_actor(
     if existing_actor:
         if existing_actor.is_deleted:
             raise ap.ObjectNotFoundError(f"{actor_id} was deleted")
-        return existing_actor
-    else:
-        if save_if_not_found:
-            ap_actor = await ap.fetch(actor_id)
-            # Some softwares uses URL when we expect ID
-            if actor_id == ap_actor.get("url"):
-                # Which mean we may already have it in DB
-                existing_actor_by_url = (
-                    await db_session.scalars(
-                        select(models.Actor).where(
-                            models.Actor.ap_id == ap.get_id(ap_actor),
-                        )
-                    )
-                ).one_or_none()
-                if existing_actor_by_url:
-                    return existing_actor_by_url
 
-            return await save_actor(db_session, ap_actor)
+        if now() - as_utc(existing_actor.updated_at) > timedelta(hours=24):
+            logger.info(
+                f"Refreshing {actor_id=} last updated {existing_actor.updated_at}"
+            )
+            try:
+                ap_actor = await ap.fetch(actor_id)
+                await update_actor_if_needed(
+                    db_session,
+                    existing_actor,
+                    RemoteActor(ap_actor),
+                )
+                return existing_actor
+            except Exception:
+                logger.exception(f"Failed to refresh {actor_id}")
+                # If we fail to refresh the actor, return the cached one
+                return existing_actor
         else:
-            raise ap.ObjectNotFoundError(actor_id)
+            return existing_actor
+
+    if save_if_not_found:
+        ap_actor = await ap.fetch(actor_id)
+        # Some softwares uses URL when we expect ID
+        if actor_id == ap_actor.get("url"):
+            # Which mean we may already have it in DB
+            existing_actor_by_url = (
+                await db_session.scalars(
+                    select(models.Actor).where(
+                        models.Actor.ap_id == ap.get_id(ap_actor),
+                    )
+                )
+            ).one_or_none()
+            if existing_actor_by_url:
+                # Update the actor as we had to fetch it anyway
+                await update_actor_if_needed(
+                    db_session,
+                    existing_actor_by_url,
+                    RemoteActor(ap_actor),
+                )
+                return existing_actor_by_url
+
+        return await save_actor(db_session, ap_actor)
+    else:
+        raise ap.ObjectNotFoundError(actor_id)
+
+
+async def update_actor_if_needed(
+    db_session: AsyncSession,
+    actor_in_db: "ActorModel",
+    ra: RemoteActor,
+) -> None:
+    # Check if we actually need to udpte the actor in DB
+    if _actor_hash(ra) != _actor_hash(actor_in_db):
+        actor_in_db.ap_actor = ra.ap_actor
+        actor_in_db.handle = ra.handle
+        actor_in_db.ap_type = ra.ap_type
+        actor_in_db.updated_at = now()
+        await db_session.flush()
 
 
 @dataclass

app/boxes.py

@@ -24,6 +24,7 @@ from app.actor import Actor
 from app.actor import RemoteActor
 from app.actor import fetch_actor
 from app.actor import save_actor
+from app.actor import update_actor_if_needed
 from app.ap_object import RemoteObject
 from app.config import BASE_URL
 from app.config import BLOCKED_SERVERS
@@ -32,6 +33,7 @@ from app.config import MANUALLY_APPROVES_FOLLOWERS
 from app.config import set_moved_to
 from app.database import AsyncSession
 from app.outgoing_activities import new_outgoing_activity
+from app.source import dedup_tags
 from app.source import markdownify
 from app.uploads import upload_to_attachment
 from app.utils import opengraph
@@ -347,6 +349,7 @@ async def fetch_conversation_root(
     db_session: AsyncSession,
     obj: AnyboxObject | RemoteObject,
     is_root: bool = False,
+    depth: int = 0,
 ) -> str:
     """Some softwares do not set the context/conversation field (like Misskey).
     This means we have to track conversation ourselves. To do so, we fetch
@@ -354,12 +357,13 @@ async def fetch_conversation_root(
      - use the context field if set
      - or build a custom conversation ID
     """
-    if not obj.in_reply_to or is_root:
-        if obj.ap_context:
-            return obj.ap_context
-        else:
-            # Use the root AP ID if there'no context
-            return f"microblogpub:root:{obj.ap_id}"
+    logger.info(f"Fetching convo root for ap_id={obj.ap_id}/{depth=}")
+    if obj.ap_context:
+        return obj.ap_context
+
+    if not obj.in_reply_to or is_root or depth > 10:
+        # Use the root AP ID if there'no context
+        return f"microblogpub:root:{obj.ap_id}"
     else:
         in_reply_to_object: AnyboxObject | RemoteObject | None = (
             await get_anybox_object_by_ap_id(db_session, obj.in_reply_to)
@@ -375,15 +379,21 @@ async def fetch_conversation_root(
                 ap.FetchError,
                 ap.NotAnObjectError,
             ):
-                return await fetch_conversation_root(db_session, obj, is_root=True)
+                return await fetch_conversation_root(
+                    db_session, obj, is_root=True, depth=depth + 1
+                )
             except httpx.HTTPStatusError as http_status_error:
                 if 400 <= http_status_error.response.status_code < 500:
                     # We may not have access, in this case consider if root
-                    return await fetch_conversation_root(db_session, obj, is_root=True)
+                    return await fetch_conversation_root(
+                        db_session, obj, is_root=True, depth=depth + 1
+                    )
                 else:
                     raise
 
-        return await fetch_conversation_root(db_session, in_reply_to_object)
+        return await fetch_conversation_root(
+            db_session, in_reply_to_object, depth=depth + 1
+        )
 
 
 async def send_move(
@@ -542,7 +552,7 @@ async def send_create(
         "context": context,
         "conversation": context,
         "url": outbox_object_id(note_id),
-        "tag": tags,
+        "tag": dedup_tags(tags),
         "summary": content_warning,
         "inReplyTo": in_reply_to,
         "sensitive": is_sensitive,
@@ -562,7 +572,7 @@ async def send_create(
     for tag in tags:
         if tag["type"] == "Hashtag":
             tagged_object = models.TaggedOutboxObject(
-                tag=tag["name"][1:],
+                tag=tag["name"][1:].lower(),
                 outbox_object_id=outbox_object.id,
             )
             db_session.add(tagged_object)
@@ -1490,7 +1500,7 @@ async def _handle_update_activity(
             )
 
         # Update the actor
-        from_actor.ap_actor = updated_actor.ap_actor
+        await update_actor_if_needed(db_session, from_actor, updated_actor)
     elif (ap_type := wrapped_object["type"]) in [
         "Question",
         "Note",
@@ -1513,6 +1523,7 @@ async def _handle_update_activity(
             # Everything looks correct, update the object in the inbox
             logger.info(f"Updating {existing_object.ap_id}")
             existing_object.ap_object = wrapped_object
+            existing_object.updated_at = now()
     else:
         # TODO(ts): support updating objects
         logger.info(f'Cannot update {wrapped_object["type"]}')

app/httpsig.py

@@ -88,8 +88,12 @@ def _body_digest(body: bytes) -> str:
     return "SHA-256=" + base64.b64encode(h.digest()).decode("utf-8")
 
 
-async def _get_public_key(db_session: AsyncSession, key_id: str) -> Key:
-    if cached_key := _KEY_CACHE.get(key_id):
+async def _get_public_key(
+    db_session: AsyncSession,
+    key_id: str,
+    should_skip_cache: bool = False,
+) -> Key:
+    if not should_skip_cache and (cached_key := _KEY_CACHE.get(key_id)):
         logger.info(f"Key {key_id} found in cache")
         return cached_key
 
@@ -101,15 +105,18 @@ async def _get_public_key(db_session: AsyncSession, key_id: str) -> Key:
             select(models.Actor).where(models.Actor.ap_id == key_id.split("#")[0])
         )
     ).one_or_none()
-    if existing_actor and existing_actor.public_key_id == key_id:
-        k = Key(existing_actor.ap_id, key_id)
-        k.load_pub(existing_actor.public_key_as_pem)
-        logger.info(f"Found {key_id} on an existing actor")
-        _KEY_CACHE[key_id] = k
-        return k
+    if not should_skip_cache:
+        if existing_actor and existing_actor.public_key_id == key_id:
+            k = Key(existing_actor.ap_id, key_id)
+            k.load_pub(existing_actor.public_key_as_pem)
+            logger.info(f"Found {key_id} on an existing actor")
+            _KEY_CACHE[key_id] = k
+            return k
 
     # Fetch it
     from app import activitypub as ap
+    from app.actor import RemoteActor
+    from app.actor import update_actor_if_needed
 
     # Without signing the request as if it's the first contact, the 2 servers
     # might race to fetch each other key
@@ -133,6 +140,12 @@ async def _get_public_key(db_session: AsyncSession, key_id: str) -> Key:
             f"failed to fetch requested key {key_id}: got {actor['publicKey']}"
         )
 
+    if should_skip_cache and actor["type"] != "Key" and existing_actor:
+        # We had to skip the cache, which means the actor key probably changed
+        # and we want to update our cached version
+        await update_actor_if_needed(db_session, existing_actor, RemoteActor(actor))
+        await db_session.commit()
+
     _KEY_CACHE[key_id] = k
     return k
 
@@ -216,7 +229,17 @@ async def httpsig_checker(
     has_valid_signature = _verify_h(
         signed_string, base64.b64decode(hsig["signature"]), k.pubkey
     )
-    # FIXME: fetch/update the user if the signature is wrong
+
+    # If the signature is not valid, we may have to update the cached actor
+    if not has_valid_signature:
+        logger.info("Invalid signature, trying to refresh actor")
+        try:
+            k = await _get_public_key(db_session, hsig["keyId"], should_skip_cache=True)
+            has_valid_signature = _verify_h(
+                signed_string, base64.b64decode(hsig["signature"]), k.pubkey
+            )
+        except Exception:
+            logger.exception("Failed to refresh actor")
 
     httpsig_info = HTTPSigInfo(
         has_valid_signature=has_valid_signature,

app/main.py

@@ -763,7 +763,7 @@ async def tag_by_name(
     _: httpsig.HTTPSigInfo = Depends(httpsig.httpsig_checker),
 ) -> ActivityPubResponse | templates.TemplateResponse:
     where = [
-        models.TaggedOutboxObject.tag == tag,
+        models.TaggedOutboxObject.tag == tag.lower(),
         models.OutboxObject.visibility == ap.VisibilityEnum.PUBLIC,
         models.OutboxObject.is_deleted.is_(False),
     ]
@@ -789,7 +789,7 @@ async def tag_by_name(
         return ActivityPubResponse(
             {
                 "@context": ap.AS_CTX,
-                "id": BASE_URL + f"/t/{tag}",
+                "id": BASE_URL + f"/t/{tag.lower()}",
                 "type": "OrderedCollection",
                 "totalItems": tagged_count,
                 "orderedItems": [

app/source.py

@@ -21,15 +21,16 @@ if typing.TYPE_CHECKING:
 
 _FORMATTER = HtmlFormatter(style=CODE_HIGHLIGHTING_THEME)
 _HASHTAG_REGEX = re.compile(r"(#[\d\w]+)")
-_MENTION_REGEX = re.compile(r"@[\d\w_.+-]+@[\d\w-]+\.[\d\w\-.]+")
+_MENTION_REGEX = re.compile(r"(@[\d\w_.+-]+@[\d\w-]+\.[\d\w\-.]+)")
+_URL_REGEX = re.compile(
+    "(https?:\\/\\/(?:www\\.)?[-a-zA-Z0-9@:%._\\+~#=]{1,256}\\.[a-zA-Z0-9()]{1,6}\\b(?:[-a-zA-Z0-9()@:%_\\+.~#?&\\/=]*))"  # noqa: E501
+)
 
 
 class AutoLink(SpanToken):
     parse_inner = False
     precedence = 10
-    pattern = re.compile(
-        "(https?:\\/\\/(?:www\\.)?[-a-zA-Z0-9@:%._\\+~#=]{1,256}\\.[a-zA-Z0-9()]{1,6}\\b(?:[-a-zA-Z0-9()@:%_\\+.~#?&\\/=]*))"  # noqa: E501
-    )
+    pattern = _URL_REGEX
 
     def __init__(self, match_obj: re.Match) -> None:
         self.target = match_obj.group()
@@ -38,7 +39,7 @@ class AutoLink(SpanToken):
 class Mention(SpanToken):
     parse_inner = False
     precedence = 10
-    pattern = re.compile(r"(@[\d\w_.+-]+@[\d\w-]+\.[\d\w\-.]+)")
+    pattern = _MENTION_REGEX
 
     def __init__(self, match_obj: re.Match) -> None:
         self.target = match_obj.group()
@@ -47,7 +48,7 @@ class Mention(SpanToken):
 class Hashtag(SpanToken):
     parse_inner = False
     precedence = 10
-    pattern = re.compile(r"(#[\d\w]+)")
+    pattern = _HASHTAG_REGEX
 
     def __init__(self, match_obj: re.Match) -> None:
         self.target = match_obj.group()
@@ -88,9 +89,13 @@ class CustomRenderer(HTMLRenderer):
 
     def render_hashtag(self, token: Hashtag) -> str:
         tag = token.target[1:]
-        link = f'<a href="{BASE_URL}/t/{tag}" class="mention hashtag" rel="tag">#<span>{tag}</span></a>'  # noqa: E501
+        link = f'<a href="{BASE_URL}/t/{tag.lower()}" class="mention hashtag" rel="tag">#<span>{tag}</span></a>'  # noqa: E501
         self.tags.append(
-            dict(href=f"{BASE_URL}/t/{tag}", name=token.target, type="Hashtag")
+            dict(
+                href=f"{BASE_URL}/t/{tag.lower()}",
+                name=token.target.lower(),
+                type="Hashtag",
+            )
         )
         return link
 
@@ -134,17 +139,22 @@ async def _prefetch_mentioned_actors(
     return actors
 
 
-def hashtagify(content: str) -> tuple[str, list[dict[str, str]]]:
-    # TODO: fix this, switch to mistletoe?
+def hashtagify(
+    content: str,
+) -> tuple[str, list[dict[str, str]]]:
     tags = []
-    hashtags = re.findall(_HASHTAG_REGEX, content)
-    hashtags = sorted(set(hashtags), reverse=True)  # unique tags, longest first
-    for hashtag in hashtags:
-        tag = hashtag[1:]
-        link = f'<a href="{BASE_URL}/t/{tag}" class="mention hashtag" rel="tag">#<span>{tag}</span></a>'  # noqa: E501
-        tags.append(dict(href=f"{BASE_URL}/t/{tag}", name=hashtag, type="Hashtag"))
-        content = content.replace(hashtag, link)
-    return content, tags
+    with CustomRenderer(
+        mentioned_actors={},
+        enable_mentionify=False,
+        enable_hashtagify=True,
+    ) as renderer:
+        rendered_content = renderer.render(Document(content))
+        tags.extend(renderer.tags)
+
+    # Handle custom emoji
+    tags.extend(emoji.tags(content))
+
+    return rendered_content, tags
 
 
 async def markdownify(
@@ -174,3 +184,17 @@ async def markdownify(
     tags.extend(emoji.tags(content))
 
     return rendered_content, tags, list(mentioned_actors.values())
+
+
+def dedup_tags(tags: list[dict[str, str]]) -> list[dict[str, str]]:
+    idx = set()
+    deduped_tags = []
+    for tag in tags:
+        tag_idx = (tag["type"], tag["name"])
+        if tag_idx in idx:
+            continue
+
+        idx.add(tag_idx)
+        deduped_tags.append(tag)
+
+    return deduped_tags