Named AI provider connections, plus nginx and updater fixes by jhd3197 · Pull Request #141 · jhd3197/ServerKit

jhd3197 · 2026-09-08T05:23:14Z

The AI assistant now supports named provider connections with encrypted credentials and a panel default. New chats choose a connection and model; existing conversations retain their original selection. This promotion releases ServerKit 1.10.0.

Changes

Discover installed Prompture chat providers and configure native providers or gateways such as OmniRoute, Prompture Hub, and LM Studio. Preserve exact model IDs and routing aliases.
Discover models and test connections through bounded real requests. Use Prompture 1.11 native gateway streaming and tool calls.
Keep credentials isolated from ambient process configuration. Clear retained credentials when endpoints change, and reject destination changes for connections already used by chats.
Migrate existing provider settings into encrypted saved connections and attach compatible historical conversations.
Allow authenticated JWT and API-key clients to list connection IDs, names, providers, and models without configuration or secrets. Connection mutations remain admin-only; existing JWT-only routes retain their policy.
Keep connection persistence in the service layer and render typed errors through the global handler. Parse decrypted configuration through the shared JSON accessor and fail closed on corrupt or non-object configuration.
Serve runtime-extension vendor .mjs shims with a JavaScript MIME type in all shipped nginx vhosts, without intercepting plugin API assets.
Exclude SQLite files from updater tree snapshots so the dedicated database backup is not duplicated.
Set VERSION to 1.10.0, update the changelog, and regenerate the API surface, controller-boundary baseline, and migration inventory without raising any test ceilings.

Validation

70 targeted backend tests passed locally, covering AI connections and SDK wire behavior, credential isolation, corrupt configuration, API-key access and role enforcement, typed errors, inventories, authorization checks, nginx MIME configuration, and panel version resolution.
Full GitHub CI runs on the final PR head gate the merge. The release workflow rechecks backend, frontend, and extensions before building and publishing artifacts.

The backend dependency is pinned to prompture[openai,anthropic]==1.11.0; upgrades must install the updated requirements and run migration 098_ai_provider_connections.

Stock nginx (1.18-1.24 on Debian/Ubuntu) has no `.mjs` entry in mime.types, so the runtime-extension vendor shims the import map points at came back as application/octet-stream. Browsers refuse to execute an ES module with that type, so every runtime-loaded extension (WordPress, Git, Mail, Fail2ban, ...) failed on fresh installs with "Failed to fetch dynamically imported module: blob:..." (Chrome) or "'application/octet-stream' is not a valid JavaScript MIME type" (Safari). Add a location block to the shipped vhosts (and the container frontend config) that forces text/javascript for /serverkit-vendor/*.mjs. The regex is anchored to that folder: a bare `\.mjs$` would also capture the /api/v1/plugins/<slug>/assets/dist/index.mjs route and 404 every bundle. update.sh already re-copies these vhosts, so existing installs pick it up on their next update. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NWc84WDsWR7M55Mn7qtpie

…endor shim MIME The tree backup rsync copied backend/instance/serverkit.db even though the same database had just been written as the pre-upgrade copy, and rollback only ever restores from that copy. On a 450 MB database that made every update cost ~900 MB of backups, which is how small VPSes kept filling up. Exclude the SQLite files from the tree snapshot. Retention is unchanged (SERVERKIT_BACKUP_RETENTION, default 3, plus the 15% disk budget). Add a test that every shipped nginx vhost forces a JavaScript MIME type for /serverkit-vendor/*.mjs and that the regex cannot capture the /api/v1/plugins/<slug>/assets/... route. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NWc84WDsWR7M55Mn7qtpie

Copilot

🟡 Changes recommended

There are a couple of confirmed issues in the updated code (UI state handling and an inaccurate updater comment) that should be corrected before approval.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR upgrades the AI assistant configuration model from a single global provider/key to named, encrypted provider connections with per-chat affinity, and includes operational fixes for nginx .mjs MIME handling and update backup size.

Changes:

Add backend connection catalog/save/probe plumbing (Prompture descriptor–driven) plus DB migration to pin conversations to a connection and widen model IDs.
Update frontend settings + chat drawer to manage/select connections and models, and adjust i18n to the new ai.connections namespace.
Fix shipped nginx vhosts to serve /serverkit-vendor/*.mjs as JavaScript and reduce updater backup footprint by excluding SQLite DB files from rsync tree snapshots.

File summaries

File	Description
scripts/update.sh	Exclude SQLite DB files from rsync install-tree snapshot to avoid double backup size.
ROADMAP.md	Document implemented “Prompture Provider Connections” feature and scope notes.
nginx/sites-available/serverkit.conf	Force correct JS MIME for `/serverkit-vendor/*.mjs` shims.
nginx/sites-available/serverkit-insecure.conf	Force correct JS MIME for `/serverkit-vendor/*.mjs` shims.
frontend/src/styles/components/_ai-assistant.scss	Add styling for the in-chat connection/model selector row.
frontend/src/services/api/ai.js	Add API methods for connection CRUD and probe endpoints.
frontend/src/i18n/locales/ar.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/bn.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/de.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/en.json	Add `ai.connections.*` strings and remove old AI settings strings no longer used.
frontend/src/i18n/locales/es.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/fr.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/id.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/it.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/ko.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/pl.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/pt.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/ru.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/th.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/tr.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/vi.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/zh-Hans.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/i18n/locales/zh-Hant.json	Remove obsolete AI settings keys; rely on new namespace fallbacks.
frontend/src/contexts/AIContext.jsx	Track available connections + selected connection/model and send on first message.
frontend/src/components/settings/AISettingsTab.jsx	Replace legacy provider/key UI with the new AIConnectionsSettings panel.
frontend/src/components/settings/AIConnectionsSettings.jsx	New connection editor UI with provider-driven fields + discover/test/save flows.
frontend/src/components/ai/ConnectionSelector.jsx	New pre-chat selector to choose connection and optionally override model.
frontend/src/components/ai/ChatDrawer.jsx	Render the new connection selector before chat start.
frontend/nginx.conf	Force correct JS MIME for `/serverkit-vendor/*.mjs` shims (frontend container vhost).
docs/AI_PROVIDERS.md	New documentation for configuring providers/gateways and connection semantics.
CHANGELOG.md	Note nginx `.mjs` MIME fix for runtime-loaded extensions.
backend/tests/test_vendor_shim_mime.py	New tests asserting vhosts include correct `.mjs` location block and regex scope.
backend/tests/test_ai_connections.py	New tests for connection behavior, probes, migration, and driver boundary rules.
backend/requirements.txt	Pin Prompture + include OpenAI/Anthropic extras for supported drivers.
backend/migrations/versions/098_ai_provider_connections.py	New migration adding connection table, FK, indexes, and model_name widening + legacy import.
backend/app/services/ai_service.py	Route provider/model selection through saved connections and inject configured driver on resume.
backend/app/services/ai_connections.py	New service for provider catalog, draft validation, secret handling, driver build, and probe requests.
backend/app/models/ai.py	Add AiProviderConnection model; add conversation connection_id and widen model_name.
backend/app/api/ai.py	Add connection CRUD/probe endpoints and plumb connection/model into conversation creation/chat routes.

Review details

Files reviewed: 39/39 changed files
Comments generated: 2
Review effort level: Lite

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Keep connection persistence behind the service boundary, use typed errors and shared JSON parsing, and refresh reviewed inventories. The new GET connection catalog accepts API keys so API clients managing connections can discover IDs without exposing credentials; existing JWT-only routes stay unchanged. Add regression coverage and release notes.

jhd3197 and others added 4 commits September 8, 2026 00:25

feat(ai): expose Prompture providers and saved connections

569b052

feat(ai): use Prompture 1.11 native gateway streaming

add2f71

Copilot AI lite review requested due to automatic review settings September 8, 2026 05:23

Copilot started reviewing on behalf of jhd3197 September 8, 2026 05:24 View session

Copilot AI reviewed Sep 8, 2026

View reviewed changes

Comment thread frontend/src/components/settings/AIConnectionsSettings.jsx

Comment thread scripts/update.sh

jhd3197 added 2 commits September 8, 2026 01:30

fix: resolve credential draft and backup fallback review [skip version]

5a4036e

jhd3197 merged commit b962614 into main Sep 8, 2026
66 checks passed

Navigation Menu

Uh oh!

Named AI provider connections, plus nginx and updater fixes - #141

Named AI provider connections, plus nginx and updater fixes#141
jhd3197 merged 6 commits into
mainfrom
dev

jhd3197 commented Sep 8, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

jhd3197 commented Sep 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Validation

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

jhd3197 commented Sep 8, 2026 •

edited

Loading