Compaction: a button, and automatically when the window fills

A long conversation eventually just stops working. Compaction summarises the
earlier turns and sends the summary in their place.

The messages are kept. They stay in the transcript behind a collapsed
divider and simply stop being part of the request, which is what makes the
button safe to press and automatic compaction safe to have at all: a summary
that came out badly is a bad turn, not a lost conversation.

Stored on the Chat, not as a synthetic Message. A synthetic row needs a
role -- `system` breaks the one-system-message rule the moment build_messages
emits it beside the harness, and user/assistant makes it a turn people can
edit, regenerate from and copy, indistinguishable from a real one in all
four places a bubble is rendered. Worse, "editing rewinds, it does not
branch" would silently delete it and leave no marker that compaction had
happened at all.

The summary goes out as a user turn and an assistant turn, not one. A
leading assistant breaks templates requiring the first non-system message to
be user; a lone leading user produces user, user whenever the kept history
starts on a user turn -- which it always does, because the cutoff lands on a
finished reply.

compacted_through_id is a plain id rather than a foreign key: migrations.py
compiles only the column type, so a REFERENCES clause would exist on a fresh
database and not on an upgraded one, and a constraint half the fleet has is
worse than none. cutoff_message validates it on every read instead, and a
rewind past the boundary clears it.

Compacting again summarises only the delta, with the previous summary
supplied to be subsumed. Re-summarising the whole chat each time grows
quadratically and eventually exceeds the window it is protecting.

Automatically at the top of _run, not in post_message: that route's contract
is to return immediately and leave the slow part to a resumable connection,
and it also means build_request is called once, after compaction, with no
second assembly path. The trigger is the last reply's recorded usage plus an
estimate of the new turn -- retrospective because true prompt_tokens are only
knowable after a response, plus the delta because otherwise fifty thousand
characters pasted into the composer overflow a window that read 90% last
turn. It never fires when the context length is unknown. It does fire on
estimated counts, which is safe here precisely because nothing is lost.

_maybe_compact never raises: a failure logs and sends the uncompacted
request. A `status` event says "Summarising earlier messages…" in the
meantime, because a silent multi-second pause before the first token is what
a hang looks like.

The wording is three fragments under Admin - Prompts. Clearing task.compact
turns compaction off entirely.

Also adds compaction.moment(): SQLite does not store the offset, so a row
loaded from disk is naive while one in the session's identity map keeps its
tzinfo, and comparing the two raises. Every comparison here is between
exactly those.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

src/lembas/web/templates/admin/general.html

@@ -41,6 +41,29 @@
     </div>
   </section>
 
+  <section class="card">
+    <h2 class="card__title">Compaction</h2>
+    <p class="card__lede">
+      A long conversation eventually fills the model's context. When it gets
+      close, the earlier turns are summarised and the summary is sent in their
+      place. The messages themselves are kept and stay readable in the
+      transcript — they simply stop being sent.
+    </p>
+    <div class="field">
+      <label class="field__label" for="compact-threshold">Compact at</label>
+      <input class="input" id="compact-threshold" name="compact_threshold" type="number"
+             min="0" max="99" value="{{ values.compact_threshold }}">
+      <p class="field__hint">
+        Percent of the model's context length. <code>0</code> turns automatic
+        compaction off; the button in each chat still works. Nothing happens for
+        a model whose context length is unset under
+        <a href="/admin/models">Models</a> — that is "unknown", not "small", and
+        this will not act on a number nobody supplied. The wording of the
+        summary is under <a href="/admin/prompts">Prompts</a>.
+      </p>
+    </div>
+  </section>
+
   <section class="card">
     <h2 class="card__title">
       Registration

src/lembas/web/templates/chat/_message.html

@@ -119,6 +119,10 @@
          reply is being written, which is where the hand already is. #}
       <div class="msg__waiting">
         <span class="dots"><i></i><i></i><i></i></span>
+        {# What the reply is doing when it is not producing tokens. A silent
+           multi-second pause before the first token is what a hang looks
+           like. #}
+        <span class="msg__status" sse-swap="status" hx-swap="innerHTML"></span>
       </div>
       {# Counts as the reply is written. Everything is an estimate until the
          usage chunk lands at the very end, and the chips say so. #}

src/lembas/web/templates/chat/_thread.html

@@ -1,12 +1,38 @@
+{% from "_macros.html" import icon %}
 {#
-  The whole thread. Returned after a rewind, which changes an arbitrary number
-  of messages at once -- replacing the lot is simpler and less error-prone than
-  working out which individual bubbles to remove. Also included by
-  chat/index.html, so the conversation is described in exactly one place.
+  The whole thread. Returned after a rewind or a compaction, either of which
+  changes an arbitrary number of messages at once -- replacing the lot is
+  simpler and less error-prone than working out which individual bubbles to
+  remove. Also included by chat/index.html, so the conversation is described in
+  exactly one place.
 
   Deliberately has no root element: it is swapped with innerHTML into #thread,
   and an outerHTML swap would take the container with it.
 #}
+{% if compacted %}
+{# Summarised turns are kept and still readable -- they have only stopped being
+   sent. A compaction that summarised badly is then a bad turn rather than a
+   lost conversation. #}
+<details class="compacted">
+  <summary class="compacted__summary">
+    {{ icon("archive", "icon--sm") }}
+    <span>{{ compacted | length }} earlier messages, summarised</span>
+    {{ icon("chevron-down", "icon--sm reasoning__chevron") }}
+  </summary>
+  <p class="compacted__note">
+    These are no longer sent to the model; a summary of them goes instead. They
+    are kept here so nothing is lost.
+  </p>
+  <div class="compacted__body">
+    {% for message in compacted %}
+      {% with body_html = bodies.get(message.id, "") %}
+        {% include "chat/_message.html" %}
+      {% endwith %}
+    {% endfor %}
+  </div>
+</details>
+{% endif %}
+
 {% for message in messages %}
   {% with body_html = bodies.get(message.id, "") %}
     {% include "chat/_message.html" %}

src/lembas/web/templates/chat/index.html

@@ -66,6 +66,18 @@
         </button>
         {% endif %}
 
+        {% if chat and messages %}
+        <button class="btn btn--icon" type="button" aria-label="Compact this chat"
+                title="Summarise the earlier messages so they stop taking up context"
+                hx-post="/api/chats/{{ chat.id }}/compact"
+                hx-target="#thread" hx-swap="innerHTML"
+                hx-confirm="Summarise everything before the last reply? The messages stay in the transcript; they just stop being sent to the model."
+                data-confirm-title="Compact this chat"
+                data-confirm-label="Compact">
+          {{ icon("archive") }}
+        </button>
+        {% endif %}
+
         {% if chat and user.is_admin %}
         <button class="btn btn--icon" type="button" aria-label="Inspect this chat"
                 title="Inspect this chat" aria-expanded="false" data-toggle="#inspector">