Knowledge, notes, memory and skills, and a harness to make them used

Four places a model can reach for, differing in who writes a record and how it
gets in front of the model.

**Knowledge** is uploaded by a person and searched by the model. It goes through
`services/files.py:prepare` — the same pipeline as a chat attachment — so the
same PDF produces the same text whichever way it arrived, and `Document` carries
the same content columns as `Attachment` for the same reason.

**Notes** are written by the model and edited by you. Too long to inject, so
they are searched.

**Memory** is short facts, and every one of them goes into every request. That
single decision is where the rest of its design comes from: records are capped
short, the block has a budget, there is no search tool because the model is
already looking at them, and they are not shareable — a record about a person is
not content to hand round.

**Skills** are saved procedures. Only the name and description are injected; the
body is fetched when the model decides one applies, which is what makes a
hundred skills affordable. A model may write and revise its own — the safety
story is not a gate but a record: every revision is kept, attributed and
revertible. A model that has just read a hostile page can save a skill that
outlives the conversation, and the honest mitigation is that it is visible and
undoable rather than that it was prevented.

**The harness** is why any of it gets used. A model handed a tools array
ignores it and answers from recall, because nothing in the request suggests
otherwise. `services/harness.py` assembles a preamble from what this chat
actually has: when to reach for each tool, the memories, the skill index.

This is an exception to "system prompts are precedence, not concatenation", and
a deliberate one. That rule governs the three *authored* layers and is
untouched — exactly one still wins. The harness is a different axis: it
describes the machinery rather than the behaviour, nobody authored it, and there
is nothing for it to disagree with. It is prepended to whichever authored prompt
won, in one system message, since several endpoints reject a second.

Supporting changes:

- **Sharing**, in one helper. `visible_to()` is the only definition of who can
  see a library item and every listing and tool goes through it. Sharing grants
  *reading*; two people editing one note with no history and no merge is worse
  than copying it. **Administrators do not bypass this** — they bypass
  permissions elsewhere because an admin can grant themselves those anyway, but
  reading somebody's private notes is a different act.
- **FTS5**, created by `db/migrations.py:ensure_fts` with the triggers an
  external-content index needs. Idempotent, like the column sync beside it.
  Terms are ANDed and then ORed: the caller is usually a model writing a whole
  question, and requiring every word loses the match on one absent term.
- **The attach button is a menu** — file, image, a web page, or a document from
  the library. Attaching a document copies it, because history must not change
  when a document is edited later.
- **A URL fetcher with an SSRF guard.** This server can reach the router, the
  other services on the box and LLeMbas itself, and the address can come from a
  model. Private ranges are refused *after resolution* and redirects are followed
  by hand so every hop is checked. An admin can open it deliberately.
- **Model capabilities split** into protocol support and a toggle per built-in
  tool. Rows predating the split have no `tool_*` keys, and absent counts as on
  when `tools` is on — otherwise an upgrade silently takes web search away from
  every model already configured for it.

Also fixes the test fixture, which built the schema with `create_all` and so ran
against a database without the FTS tables production has; it now runs
`sync_schema`, the same path startup takes.

430 tests, ruff clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

src/lembas/services/library/memories.py

@@ -0,0 +1,119 @@
+"""Memory: short facts, in front of the model on every turn.
+
+The whole design follows from being injected rather than searched.
+
+* Each record is **capped short**, because every one of them costs tokens on
+  every request forever. A tool that writes an essay gets it trimmed and is
+  told so, rather than the write failing -- the model can then decide to put
+  the long version in a note.
+* There is a **budget** for the block as a whole. Past it the oldest are left
+  out rather than the request growing without limit; the user can see the whole
+  list in their settings and prune it.
+* There is **no search tool**. Searching something the model is already looking
+  at is a round trip for nothing.
+* They are **not shareable**. A record about a person is not content to hand
+  round, and nobody asked to share their memories with a group.
+"""
+
+from __future__ import annotations
+
+import logging
+
+from sqlalchemy import func, select
+from sqlalchemy.orm import Session as DBSession
+
+from lembas.db.models import AUTHOR_MODEL, AUTHOR_USER, Memory, User
+
+log = logging.getLogger(__name__)
+
+# One fact, not a paragraph. Long enough for "prefers metric units and a 24-hour
+# clock", short enough that fifty of them are still affordable.
+MAX_MEMORY_CHARS = 400
+
+# Ceiling on the injected block. Reached, the oldest records drop out of the
+# prompt -- they are still listed in settings, so nothing disappears silently.
+MAX_TOTAL_CHARS = 4000
+
+# A hard stop on how many can exist, so an enthusiastic model cannot fill a
+# database with variations on one fact.
+MAX_RECORDS = 200
+
+
+def all_for(db: DBSession, user: User | None) -> list[Memory]:
+    if user is None:
+        return []
+    return list(
+        db.scalars(
+            select(Memory).where(Memory.owner_id == user.id).order_by(Memory.created_at)
+        )
+    )
+
+
+def get(db: DBSession, memory_id: str, user: User | None) -> Memory | None:
+    memory = db.get(Memory, memory_id)
+    if memory is None or user is None or memory.owner_id != user.id:
+        return None
+    return memory
+
+
+def add(db: DBSession, *, owner: User, content: str, author: str = AUTHOR_MODEL) -> Memory:
+    """Record a fact. Raises ValueError when there is no room or nothing to say."""
+    content = " ".join((content or "").split())
+    if not content:
+        raise ValueError("A memory cannot be empty.")
+
+    count = db.scalar(
+        select(func.count()).select_from(Memory).where(Memory.owner_id == owner.id)
+    )
+    if (count or 0) >= MAX_RECORDS:
+        raise ValueError(
+            f"There are already {MAX_RECORDS} memories. Remove one first, or put "
+            f"this in a note instead."
+        )
+
+    memory = Memory(
+        owner_id=owner.id,
+        content=content[:MAX_MEMORY_CHARS],
+        author=author if author in (AUTHOR_USER, AUTHOR_MODEL) else AUTHOR_MODEL,
+    )
+    db.add(memory)
+    db.commit()
+    return memory
+
+
+def update(db: DBSession, memory: Memory, content: str) -> Memory:
+    content = " ".join((content or "").split())
+    if not content:
+        raise ValueError("A memory cannot be empty.")
+    memory.content = content[:MAX_MEMORY_CHARS]
+    db.commit()
+    return memory
+
+
+def delete(db: DBSession, memory: Memory) -> None:
+    db.delete(memory)
+    db.commit()
+
+
+def block(db: DBSession, user: User | None) -> str:
+    """The memories as they appear in the prompt, within the budget.
+
+    Oldest first, and truncation drops the *newest* -- a fact that has survived
+    a long time is more likely to be a standing preference than something said
+    once this morning.
+    """
+    records = all_for(db, user)
+    if not records:
+        return ""
+
+    lines: list[str] = []
+    total = 0
+    for memory in records:
+        line = f"- {memory.content}"
+        if total + len(line) > MAX_TOTAL_CHARS:
+            lines.append(f"- (…{len(records) - len(lines)} more, see your settings)")
+            break
+        lines.append(line)
+        total += len(line)
+
+    return "\n".join(lines)