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>
This commit is contained in:
Jaroslav Beneš
@@ -17,10 +17,16 @@
|
||||
<div class="composer" {% if can.get("files.upload") %}data-dropzone{% endif %}>
|
||||
{% if can.get("files.upload") %}
|
||||
{# Outside the form: it is only ever read by JavaScript, and inside it would
|
||||
be submitted as an empty file part on every message. #}
|
||||
be submitted as an empty file part on every message.
|
||||
|
||||
Two inputs rather than one whose accept attribute is rewritten: changing
|
||||
accept and then calling click() in the same tick is unreliable in Safari,
|
||||
and two hidden inputs cost nothing. #}
|
||||
<input class="visually-hidden" type="file" id="file-input" multiple
|
||||
data-upload-url="/api/files{% if chat %}?chat_id={{ chat.id }}{% endif %}"
|
||||
accept="image/*,.pdf,.txt,.md,.csv,.json,.py,.js,.ts,.rs,.go,.sh,.sql,.yaml,.yml,.toml,.log"
|
||||
accept=".pdf,.txt,.md,.csv,.json,.py,.js,.ts,.rs,.go,.sh,.sql,.yaml,.yml,.toml,.log"
|
||||
onchange="window.lembas.uploadFiles(this.files); this.value = ''">
|
||||
<input class="visually-hidden" type="file" id="image-input" multiple accept="image/*"
|
||||
onchange="window.lembas.uploadFiles(this.files); this.value = ''">
|
||||
{% endif %}
|
||||
|
||||
@@ -47,11 +53,54 @@
|
||||
|
||||
<div class="composer__row">
|
||||
{% if can.get("files.upload") %}
|
||||
<button class="btn btn--icon composer__btn" type="button"
|
||||
aria-label="Attach a file" title="Attach a file"
|
||||
onclick="document.getElementById('file-input').click()">
|
||||
{{ icon("attach") }}
|
||||
</button>
|
||||
{# A menu rather than the file picker straight away: there are four ways
|
||||
to attach something now, and only one of them is a file on disk.
|
||||
Uses the same picker machinery as the model chooser -- see ui.js. #}
|
||||
<div class="picker picker--up" data-picker>
|
||||
<button class="btn btn--icon composer__btn" type="button" data-picker-toggle
|
||||
aria-haspopup="menu" aria-expanded="false"
|
||||
aria-label="Attach" title="Attach">
|
||||
{{ icon("attach") }}
|
||||
</button>
|
||||
|
||||
<div class="picker__menu picker__menu--compact" data-picker-menu role="menu"
|
||||
hidden aria-label="Attach">
|
||||
<button class="picker__option" type="button" role="menuitem"
|
||||
data-attach="file">
|
||||
{{ icon("attach", "icon--sm") }}
|
||||
<span class="picker__option-body">
|
||||
<span class="picker__option-name">File</span>
|
||||
<span class="picker__option-note">PDF, text, code</span>
|
||||
</span>
|
||||
</button>
|
||||
<button class="picker__option" type="button" role="menuitem"
|
||||
data-attach="image">
|
||||
{{ icon("image", "icon--sm") }}
|
||||
<span class="picker__option-body">
|
||||
<span class="picker__option-name">Image</span>
|
||||
<span class="picker__option-note">Sent only to vision models</span>
|
||||
</span>
|
||||
</button>
|
||||
<button class="picker__option" type="button" role="menuitem"
|
||||
data-attach="link">
|
||||
{{ icon("link", "icon--sm") }}
|
||||
<span class="picker__option-body">
|
||||
<span class="picker__option-name">Link</span>
|
||||
<span class="picker__option-note">Fetch a page and attach its text</span>
|
||||
</span>
|
||||
</button>
|
||||
{% if can.get("library.use") %}
|
||||
<button class="picker__option" type="button" role="menuitem"
|
||||
data-attach="knowledge">
|
||||
{{ icon("archive", "icon--sm") }}
|
||||
<span class="picker__option-body">
|
||||
<span class="picker__option-name">Knowledge</span>
|
||||
<span class="picker__option-note">From your library</span>
|
||||
</span>
|
||||
</button>
|
||||
{% endif %}
|
||||
</div>
|
||||
</div>
|
||||
{% endif %}
|
||||
|
||||
<textarea class="composer__input" name="content" rows="1"
|
||||
|
||||
@@ -0,0 +1,38 @@
|
||||
{% from "_macros.html" import icon %}
|
||||
{#
|
||||
The list inside the composer's Knowledge dialog.
|
||||
|
||||
Swapped in on its own so typing in the filter re-queries the server: the
|
||||
library is searched with FTS rather than filtered in the browser, which is
|
||||
what makes it work when there are five hundred documents rather than five.
|
||||
#}
|
||||
<div id="knowledge-results">
|
||||
{% if not documents %}
|
||||
<p class="muted text-sm" style="padding: var(--sp-3)">
|
||||
{% if q %}
|
||||
Nothing matches “{{ q }}”.
|
||||
{% else %}
|
||||
Your library is empty. <a href="/library/knowledge">Add something to it</a>.
|
||||
{% endif %}
|
||||
</p>
|
||||
{% else %}
|
||||
<ul class="picker__list">
|
||||
{% for document in documents %}
|
||||
<li>
|
||||
<button class="picker__option" type="button"
|
||||
data-attach-knowledge="{{ document.id }}">
|
||||
{{ icon("archive" if not document.is_image else "image", "icon--sm") }}
|
||||
<span class="picker__option-body">
|
||||
<span class="picker__option-name">{{ document.title }}</span>
|
||||
<span class="picker__option-note">
|
||||
{{ document.kind }}
|
||||
{%- if document.pages %} · {{ document.pages }}p{% endif %}
|
||||
{%- if document.owner_id != user.id %} · shared{% endif %}
|
||||
</span>
|
||||
</span>
|
||||
</button>
|
||||
</li>
|
||||
{% endfor %}
|
||||
</ul>
|
||||
{% endif %}
|
||||
</div>
|
||||
Reference in New Issue
Block a user