mirror of
https://gh-proxy.org/https://github.com/klinsmaya/multica.git
synced 2026-11-09 01:46:46 +08:00
* feat(plugins): host plugin artifacts, bind installs to immutable versions A plugin used to be a URL. The manifest was frozen at install, but the surface script was fetched from the author's server every time a panel opened, with no integrity check — so an administrator consented to a manifest while the browser ran whatever that host served that day, inside the scopes already granted. The author's uptime was our uptime, every panel open leaked the reader's IP and "who read which issue" to the author, and publishing at all required a public HTTPS domain, which is why the repo's own deploy-sentinel example never ran. The author now uploads an artifact bundle and Multica stores it: - plugin_package / plugin_package_version / plugin_package_file. A version is insert-only, and the (package_id, version) unique index is what makes immutability a database rule rather than a convention. - An installation names one version. Publishing another changes nothing for a workspace until an administrator upgrades, which is a second consent. - The bundle is validated once, at publish: the manifest parses, every file it declares is present and loadable, and a surface entry with a top-level import is refused with the line named. That failure used to surface months later in a reader's browser. - The host inlines the surface script into the document it generates, so the CSP names no remote script origin at all. `connect-src 'none'` now means it: a surface with no net: scope cannot reach anywhere, including its author. - Install-by-URL is gone rather than kept alongside. Two paths would make "is this code frozen?" depend on how the plugin was installed. - MULTICA_PLUGIN_DIR stays the development channel, but publishes an ordinary immutable version instead of being a second render path; re-publishing an unchanged version lands as `+dev.N`. Hook endpoints and MCP servers remain on the author's own infrastructure. Migration 376 drops existing installations: they name a source URL that is no longer a concept, and the code they would run was never published here. plugins_v1 has never left its feature flag and holds no production data — the same call migration 344 made. Closes MUL-6469 Co-authored-by: multica-agent <[email protected]> * fix(plugins): serialize publish/install/delete, correct the isolation claim Review of #7321 found five problems. Four are fixed here; the fifth is real, needs its own design pass, and is now tracked and described honestly in the code instead of claimed away. Migration prefix collision. main took 376 for agent_task_durable_work_dir while this branch was open, so `migration prefix 376 is reused` failed backend-tests. Rebased and renumbered the six migrations to 377-382. TOCTOU between delete and install/publish. Relationships are application-owned by repository policy, so nothing made "the version this installation names still exists" true across statements: `delete counts zero installs` -> `install reads the version` -> `delete commits` -> `install commits` left an installation whose panel 404s forever, and a publish racing a delete left a version whose package row was gone. Publish, install and delete now hold a (workspace, plugin key) advisory lock for their whole transaction, and delete counts installations inside it. Both interleavings have regression tests that reproduce the bug with the lock removed. Failed publishes moved package state. upsertPackage ran before the version transaction, so a republish that lost the unique index had already renamed the package, and a first publish that failed afterwards left a plugin with zero versions. It now runs inside the same transaction. The top-level import check was wrong in BOTH directions. It read line prefixes, so it missed ` import x` and `/* c */ import x` — and it refused a valid file whose template literal contained the word at a line start, which is ordinary in a surface that renders code samples. A refusal blocks a publish with no way around it, so that false positive was the worse half. Replaced with a scanner that skips comments and string/template literals; where it is imprecise it is imprecise safely (`/` always reads as division), so a mistake can only cost a detection, never invent one. Sandbox self-navigation is NOT fixed. `sandbox="allow-scripts"` permits `_self` navigation by design, no shipped CSP directive covers it, and contentWindow is unchanged across it — so a hostile artifact can still reach its author once and hand the replacement document the bridge port. Closing it needs an embedder frame-src policy or a handshake a navigated document cannot complete; both are decisions of their own. What lands here is damage control, labelled as such: the generated document beacons on pagehide from a listener the plugin cannot detach, and the embedder drops the bridge and unmounts the frame. The "a surface cannot reach its author" wording is removed from the SDK README and the tests. Closes MUL-6469 Refs MUL-6485 Co-authored-by: multica-agent <[email protected]> * fix(plugins): harden surface validation and error reporting * fix(migrations): renumber plugin package migrations * fix(plugins): validate classic surfaces and reset failures --------- Co-authored-by: Lambda <[email protected]> Co-authored-by: multica-agent <[email protected]>
84 lines
4.1 KiB
SQL
84 lines
4.1 KiB
SQL
-- Platform-hosted plugin artifacts.
|
|
--
|
|
-- Until now a plugin was a URL: the manifest was fetched once and frozen as the
|
|
-- consented snapshot, but the surface script was loaded from the author's server
|
|
-- every time a panel opened. So the administrator consented to a manifest while
|
|
-- the browser ran whatever code that host served that day, the author's uptime
|
|
-- became our uptime, and every panel open told the author who was reading which
|
|
-- issue.
|
|
--
|
|
-- What replaces it: the author uploads an artifact bundle, Multica stores it,
|
|
-- and an installation is bound to one immutable version. Nothing about a
|
|
-- published version is ever updated in place — a new publish is a new row, and
|
|
-- an installed workspace keeps running the version it consented to until an
|
|
-- administrator upgrades it.
|
|
--
|
|
-- Hook endpoints and MCP servers stay on the author's own infrastructure. Only
|
|
-- the frontend artifact moved.
|
|
|
|
-- One publishable plugin identity per workspace. Publishing is workspace-private
|
|
-- for now; a public directory needs review, reporting and takedown, which is a
|
|
-- separate decision from where the bytes live.
|
|
CREATE TABLE plugin_package (
|
|
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
|
|
workspace_id UUID NOT NULL,
|
|
plugin_key TEXT NOT NULL CHECK (char_length(plugin_key) BETWEEN 3 AND 255),
|
|
name TEXT NOT NULL CHECK (char_length(name) BETWEEN 1 AND 160),
|
|
created_by UUID,
|
|
created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
|
|
updated_at TIMESTAMPTZ NOT NULL DEFAULT now()
|
|
);
|
|
|
|
-- One published version. Immutable by construction: no statement in the
|
|
-- application updates this row, and republishing an existing version is a
|
|
-- conflict rather than an overwrite. That is the whole point — "what the admin
|
|
-- approved" and "what the browser runs" have to name the same bytes.
|
|
--
|
|
-- workspace_id is denormalized so every read can be scoped without joining
|
|
-- through plugin_package; relationships stay application-owned by repository
|
|
-- policy, so there is no foreign key to join through cheaply anyway.
|
|
CREATE TABLE plugin_package_version (
|
|
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
|
|
package_id UUID NOT NULL,
|
|
workspace_id UUID NOT NULL,
|
|
version TEXT NOT NULL CHECK (char_length(version) BETWEEN 1 AND 64),
|
|
manifest JSONB NOT NULL CHECK (jsonb_typeof(manifest) = 'object'),
|
|
-- sha256 of the uploaded bundle, hex. Shown to the publisher so two people
|
|
-- can confirm they are looking at the same artifact.
|
|
digest TEXT NOT NULL CHECK (char_length(digest) = 64),
|
|
size_bytes BIGINT NOT NULL CHECK (size_bytes >= 0),
|
|
published_by UUID,
|
|
created_at TIMESTAMPTZ NOT NULL DEFAULT now()
|
|
);
|
|
|
|
-- The files a version ships: the surface scripts, the skill text, the icon.
|
|
--
|
|
-- Stored in Postgres rather than object storage. A bundle is capped at a few
|
|
-- hundred kilobytes by the service, and this way an installed plugin cannot be
|
|
-- half-available: the manifest snapshot and the code it names commit together
|
|
-- and are restored together. Object storage would reintroduce, against our own
|
|
-- bucket, exactly the availability split this change exists to remove.
|
|
CREATE TABLE plugin_package_file (
|
|
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
|
|
version_id UUID NOT NULL,
|
|
path TEXT NOT NULL CHECK (char_length(path) BETWEEN 1 AND 1024),
|
|
content BYTEA NOT NULL,
|
|
size_bytes BIGINT NOT NULL CHECK (size_bytes >= 0),
|
|
sha256 TEXT NOT NULL CHECK (char_length(sha256) = 64),
|
|
created_at TIMESTAMPTZ NOT NULL DEFAULT now()
|
|
);
|
|
|
|
-- Every existing installation names a source URL, which is no longer a concept.
|
|
-- There is nothing to migrate them to: the code they would run was never
|
|
-- published here, so binding them to a version would be an invention. plugins_v1
|
|
-- has never left its feature flag and holds no production data, so they are
|
|
-- removed — the same call migration 344 made for the same reason.
|
|
DELETE FROM skill WHERE plugin_installation_id IS NOT NULL;
|
|
DELETE FROM plugin_invocation;
|
|
DELETE FROM plugin_secret;
|
|
DELETE FROM plugin_storage;
|
|
DELETE FROM plugin_installation;
|
|
|
|
ALTER TABLE plugin_installation DROP COLUMN IF EXISTS source_url;
|
|
ALTER TABLE plugin_installation ADD COLUMN package_version_id UUID NOT NULL;
|