contact messages: easypow proof-of-work for sends, pending-upload cleanup

New dependency: easypow (PBKDF2-SHA512 challenges, bundled JS solver).

Sends are PoW-gated: registration/resync and every accepted message
come with a fresh single-use challenge (~5 s solve), held in per-socket
RAM state only — nothing in the database. A message burns one
solution; a missing, stale or spent solution is dropped and answered
with a fresh 'required' challenge. The bundled browser solver is
served at /_contact/easypow.js straight from the installed package, so
client and validator versions always match.

Uploads are deliberately not PoW-gated (the transfer rate is the real
cost). Instead, uploads are tracked as pending per session and a quota
caps pending files at 10 — answered 429, not 403, which could trip
auth flows. A sent message strikes the files it references off the
pending set; when the session's last socket closes, the remaining
pending files are deleted from the store.

AGENTS.md

@@ -15,7 +15,7 @@ Pagerite is a CMS. See `docs` for the full design and implementation details. Ke
   - `files.py` — `FileStore` (content-addressed, RAM-cached), image derivative helpers (`store_image`; `store_attachment` keeps non-media uploads as inert `{hash}.file` blobs), file routes (`/_api/files`, `/_f/` — media by hash, attachments only via `/_f/{hash}/{filename}`, all forced-download + nosniff, `/_themes/`, `/_fonts/`, favicon settings).
   - `api.py` — editor REST + WS: `/_api/pages`, `/_api/structure`, `/_api/settings`, `/_api/toggle-task`, `/_api/translations`, `/_api/ws/editor`, `/_translate/{key}`.
   - `tracking.py` — visit analytics: GeoIP, client enrichment, favicon fetch, `/_ws`, `/_api/ws/analytics`, the `/_a` page (docs/analytics.md). Both sockets also carry the contact-message protocol (docs/messages.md): `/_ws` routes token-bearing visitor frames, the analytics socket the admin inbox.
-  - `messages.py` — visitor contact threads (docs/messages.md): strict message rendering (`html=False` markdown-it — threads never carry user markup), thread store ops, the visitor/admin socket protocols, and the public `/_contact/files` upload route.
+  - `messages.py` — visitor contact threads (docs/messages.md): strict message rendering (`html=False` markdown-it — threads never carry user markup), thread store ops, the visitor/admin socket protocols, easypow PoW gating (single-use challenges in per-socket RAM, browser solver served at `/_contact/easypow.js`), and the public `/_contact/files` upload route (pending-upload quota, unreferenced files deleted when the session's last socket closes).
   - `pages.py` — public content pages: `/`, `/sitemap.xml`, `/robots.txt`, the `/{path:path}` catch-all. Articles serve their Markdown (title injected as `# title`, per-language hybrid) at a `.md`-suffixed URL, or at the plain URL when Accept explicitly names text/markdown; pages link it via `<link rel="alternate" type="text/markdown">`, and llms.txt links point at the `.md` URLs.
   - `feeds.py` — machine-readable exports: `/llms.txt` (compact index linking each article's `.md` Markdown URL), `/feed.json` (JSON Feed 1.1), `/feed.xml` (RSS 2.0 + atom:link); the feeds carry all published articles with full content, linked from every page `<head>` and the sitemap, recorded in analytics.
   - `data.py` — msgspec Structs for the kanta database.
@@ -49,7 +49,7 @@ Server run by CLI entry point `uv run pagerite` (no auto reloads, build needed).
 
 ## Toolchain
 
-- Python >= 3.14, managed with **uv**. Dependencies: `fastapi[standard]`, `fastapi-vue`, `html5tagger`, `kanta`, `markdown-it-py`, `mdit-py-plugins`, `platformdirs`, `pygments`, `tracerite`; dev group has `httpx`. Run anything via `uv run ...` (the venv is `.venv`).
+- Python >= 3.14, managed with **uv**. Dependencies: `fastapi[standard]`, `fastapi-vue`, `easypow`, `html5tagger`, `kanta`, `markdown-it-py`, `mdit-py-plugins`, `platformdirs`, `pygments`, `tracerite`; dev group has `httpx`. Run anything via `uv run ...` (the venv is `.venv`).
 - Key libraries:
   - **html5tagger** — all HTML generation (`E`, `Document`, `Template`, `HTML` for trusted/raw HTML).
     - To create stand alone pages, begin with `doc = Document(...)` that gives a HTML5 page header

docs/messages.md

@@ -49,12 +49,28 @@ no extra connection). Frames **to** the server:
   `pagerite.js` re-sends this on every socket (re)connect, which is the
   resync mechanism (the socket idle-closes after 5 min and reconnects on
   activity).
-- `{token, msg, fr}` — append a message; `msg` is the Markdown-subset
-  source, `fr` the page path it was sent from (stored for context).
+- `{token, msg, pow, fr}` — append a message; `msg` is the Markdown-subset
+  source, `pow` the solved proof-of-work (below), `fr` the page path it
+  was sent from (stored for context).
 
 Frames **from** the server: `{type: "thread", id, created, unread,
 messages: [{id, t, admin, html, page}]}` — a full thread snapshot
-(replace/upsert by `id`), or `{type: "contact-error", detail}`.
+(replace/upsert by `id`); `{type: "pow", challenge, required}` — a fresh
+single-use PoW challenge (`required` means the last message was rejected
+for a missing/stale solution and should be re-sent once solved); or
+`{type: "contact-error", detail}`.
+
+### Proof of work (anti-spam)
+
+Sends are gated by easypow (PBKDF2 challenges, ~5 s solving cost). The
+server hands a fresh single-use challenge with every registration/resync
+and after every accepted message; the panel solves it in the background
+with the bundled solver module (served from the installed package at
+`/_contact/easypow.js`, so client and validator versions always match),
+typically finishing while the user is still writing. Each message burns
+one solution — a send with a missing, stale or already-spent solution is
+dropped and answered with a fresh `required` challenge, which the client
+solves and retries once. Challenges live in per-socket RAM state only.
 
 ## Admin side
 
@@ -110,11 +126,16 @@ filename needed — the URL's last segment names the download) and
 rendered or sniffed into something executable in the site's origin;
 subresource fetches (`<img>`, `<video>`, favicons) ignore the
 disposition, so inline media is unaffected. The chip anchors
-additionally carry a `download` attribute. Gates today: a
-session-shaped `x-contact-token` header and a 20 MB size cap. Admin
-uploads from the same composer go to the admin-gated `/_api/files`
-instead. Files uploaded but never referenced by a sent message
-currently linger (no orphan GC).
+additionally carry a `download` attribute. Uploads are deliberately
+**not** PoW-gated (the transfer rate is the real cost there); the gates
+are a session-shaped `x-contact-token` header, a 20 MB size cap, and a
+pending-upload quota — 10 un-sent files per session, answered 429 (not
+403, which could trip auth flows). Every upload stays *pending* per
+session: a sent message strikes the files it references off the pending
+set, and when the session's last socket closes the remaining pending
+files are deleted from the store. Admin uploads from the same composer
+go to the admin-gated `/_api/files` instead. Files left behind by a
+server restart (the pending set is RAM-only) still linger — no GC.
 
 ## Storage
 
@@ -129,11 +150,9 @@ transactions (`contact` / `contact:reply` / `contact:read`).
 
 ## Limits and deliberate gaps (MVP)
 
-- Message source ≤ 8000 chars; uploads ≤ 20 MB; payloads cap at 100
-  threads × last 200 messages.
-- **No proof-of-work yet**: the send and upload paths are the two seams
-  where the PoW stamp plugs in (challenge on token registration, verify on
-  `msg`/upload; one stamp may cover several messages).
-- No thread moderation (delete/block), no orphan-upload GC, panel strings
-  are English-only, and the admin inbox lists threads by token-hash prefix
-  rather than enriched visitor identity.
+- Message source ≤ 8000 chars; uploads ≤ 20 MB and ≤ 10 pending per
+  session; payloads cap at 100 threads × last 200 messages.
+- No thread moderation (delete/block), no orphan-upload GC for files
+  orphaned by a server restart, panel strings are English-only, and the
+  admin inbox lists threads by token-hash prefix rather than enriched
+  visitor identity.

pagerite/messages.py

@@ -10,6 +10,15 @@ Admins watch and reply over the analytics socket
 frames go to ``admin_frame``. Both sides get pushed full thread payloads
 (threads are small); there is no REST polling anywhere.
 
+Sends are proof-of-work gated (easypow): registration/resync and every
+accepted message come with a fresh single-use challenge, solved in the
+background by the browser (the bundled JS solver is served at
+/_contact/easypow.js); a message burns one solution. Uploads are not
+PoW-gated — the transfer rate is the real cost there — but uploaded
+files stay pending per session and are deleted when its last socket
+closes without a sent message referencing them; a quota caps pending
+files.
+
 Message bodies are a Markdown subset (emphasis, links, lists, code plus
 /_f/ image and file references) serialized client-side from a constrained
 contenteditable composer. They are rendered here with a strict MarkdownIt
@@ -29,12 +38,14 @@ from pathlib import Path
 from urllib.parse import quote
 
 import blake3
+import easypow
 from fastapi import APIRouter, HTTPException, Request, WebSocket
+from fastapi.responses import Response
 from markdown_it import MarkdownIt
 from markdown_it.common.utils import escapeHtml
 
 from pagerite.data import ContactMessage, ContactThread
-from pagerite.files import _ext, store_attachment, store_image
+from pagerite.files import _ext, file_store, store_attachment, store_image
 from pagerite.state import data, kanta
 
 logger = logging.getLogger(__name__)
@@ -51,6 +62,16 @@ MAX_MESSAGE = 8000
 _visitor_ws: dict[bytes, set[WebSocket]] = {}
 _admin_ws: set[WebSocket] = set()
 
+#: PoW state (RAM only): the outstanding single-use challenge per visitor
+#: socket, and uploaded-but-unreferenced file digests per session — the
+#: latter are deleted when the session's last socket closes.
+_pow: dict[WebSocket, str] = {}
+_pending_uploads: dict[bytes, set[str]] = {}
+
+#: easypow solving cost per message and the pending-upload quota.
+POW_SECONDS = 5.0
+MAX_PENDING_UPLOADS = 10
+
 
 # --- Strict rendering -----------------------------------------------------
 # A separate instance from markdown.py's page renderer: no raw HTML (the
@@ -182,9 +203,18 @@ def _append(key: bytes, msg: ContactMessage, action: str, user: str | None) -> N
 # --- Visitor protocol (/_ws frames carrying "token") ------------------------
 
 
+async def _issue_pow(ws: WebSocket, *, required: bool = False) -> None:
+    """Send a fresh single-use PoW challenge (~POW_SECONDS of solving),
+    replacing any outstanding one for the socket."""
+    challenge = easypow.generate(POW_SECONDS)
+    _pow[ws] = challenge
+    await ws.send_json({"type": "pow", "challenge": challenge, "required": required})
+
+
 async def visitor_frame(ws: WebSocket, raw: dict) -> None:
     """Handle one visitor frame: ``{token}`` registers/resyncs (the thread
-    state is pushed back), ``{token, msg, fr?}`` appends a message and
+    state is pushed back, with a fresh PoW challenge), ``{token, msg, pow,
+    fr?}`` burns the single-use PoW solution, appends the message and
     broadcasts the updated thread to both sides."""
     token = str(raw.get("token") or "")
     if not TOKEN_RE.fullmatch(token):
@@ -196,10 +226,19 @@ async def visitor_frame(ws: WebSocket, raw: dict) -> None:
     if not text:
         thread = data.contact.get(key)
         await ws.send_json(_payload(key, thread) if thread else _empty_payload(key))
+        await _issue_pow(ws)
         return
     if len(text) > MAX_MESSAGE:
         await ws.send_json({"type": "contact-error", "detail": "message too long"})
         return
+    challenge = _pow.pop(ws, None)  # every send attempt consumes it
+    try:
+        if challenge is None:
+            raise ValueError("no outstanding challenge")
+        easypow.validate(challenge, str(raw.get("pow") or ""))
+    except ValueError:
+        await _issue_pow(ws, required=True)
+        return
     now = datetime.now(UTC)
     _append(
         key,
@@ -212,16 +251,26 @@ async def visitor_frame(ws: WebSocket, raw: dict) -> None:
         "contact",
         None,
     )
+    # Files the message references are no longer pending cleanup.
+    if pending := _pending_uploads.get(key):
+        for digest in re.findall(r"/_f/([0-9a-f]{12})", text):
+            pending.discard(digest)
     await _broadcast(key, data.contact[key])
+    await _issue_pow(ws)  # the next message needs a fresh solution
 
 
 def visitor_disconnect(ws: WebSocket) -> None:
-    """Drop a closed /_ws socket from every thread registry."""
+    """Drop a closed /_ws socket from every registry. When a session's
+    last socket closes, its uploaded files that no sent message ended up
+    referencing are deleted from the store."""
+    _pow.pop(ws, None)
     empty = [key for key, sockets in _visitor_ws.items() if not sockets - {ws}]
     for sockets in _visitor_ws.values():
         sockets.discard(ws)
     for key in empty:
         del _visitor_ws[key]
+        for digest in _pending_uploads.pop(key, ()):
+            file_store.delete(digest)
 
 
 # --- Admin protocol (/_api/ws/analytics frames) -----------------------------
@@ -318,16 +367,23 @@ async def contact_upload(name: str, request: Request) -> dict:
     """Store a contact-panel upload in the content-addressed store.
 
     Public like the messaging socket (only /_api is gated): the caller
-    must present a session-shaped token, size is capped, and abuse
-    limiting is meant to ride on the proof-of-work stamp once plugged in.
+    must present a session-shaped token and size is capped. Uploads are
+    deliberately not PoW-gated (the transfer rate is the real cost);
+    instead the file stays pending per session and is deleted when the
+    session's last socket closes without a sent message referencing it,
+    and a quota (429, not 403 — no auth flows here) caps pending files.
     Images (raster and SVG) get the usual derivative treatment and are
     referenced extension-less; any other file is stored as an inert
     {hash}.file blob and referenced as /_f/{hash}/{original-name}, which
     is always served as a forced download. Returns the /_f/ path and
     whether the file is an inline-able image.
     """
-    if not TOKEN_RE.fullmatch(request.headers.get("x-contact-token", "")):
+    token = request.headers.get("x-contact-token", "")
+    if not TOKEN_RE.fullmatch(token):
         raise HTTPException(403, "no session")
+    pending = _pending_uploads.setdefault(thread_key(token), set())
+    if len(pending) >= MAX_PENDING_UPLOADS:
+        raise HTTPException(429, "too many un-sent uploads — send your message first")
     try:
         length = int(request.headers.get("content-length") or 0)
     except ValueError:
@@ -346,10 +402,24 @@ async def contact_upload(name: str, request: Request) -> dict:
         # Decodable images come back extension-less (GIFs and as-is SVGs
         # keep their extension); an undecodable body attaches like any file.
         if "." not in stored or ext in (".gif", ".svg"):
+            pending.add(stored.partition(".")[0])
             return {"path": f"/_f/{stored}", "image": True, "name": orig}
     stored = await asyncio.to_thread(store_attachment, body)
+    pending.add(stored.removesuffix(".file"))
     return {
         "path": f"/_f/{stored.removesuffix('.file')}/{quote(orig, safe='')}",
         "image": False,
         "name": orig,
     }
+
+
+@router.get("/_contact/easypow.js")
+async def pow_solver() -> Response:
+    """The easypow browser solver as an ES module, served from the
+    installed package so the client solver and the server validator are
+    always the same version."""
+    return Response(
+        easypow.js,
+        media_type="text/javascript",
+        headers={"cache-control": "public, max-age=86400"},
+    )

pyproject.toml

@@ -17,6 +17,7 @@ readme = "README.md"
 requires-python = ">=3.14"
 dependencies = [
     "blake3>=1.0.9",
+    "easypow>=0.1.0",
     "fastapi-vue~=1.7.2",
     "fastapi[standard]>=0.141.1",
     "html5tagger>=2.0.0",