* Fix OpenRouter support by using laravel/ai default provider config.
PR #216 patched Lab::OpenRouter into every agent match, but laravel/ai
already resolves config('ai.default') — including openrouter — when agents
omit provider(). Those matches also forced unknown providers to Gemini and
BrandAnalyzerRunner checked a non-existent services.openrouter key.
Remove the duplicated provider() overrides, gate availability on
ai.providers.*.key, and document OPENROUTER_API_KEY.
Co-authored-by: Paulo Castellano <hello@paulocastellano.com>
* chore(deps): bump laravel/ai to v0.10.3 for native OpenRouter support
v0.5.1's OpenRouter driver only covered text/embeddings via the legacy
Prism gateway. v0.10.3 ships a native OpenRouter gateway with image,
audio (TTS/STT), and web search support, and drops prism-php/prism as
a dependency. ai-sdk-development skill docs refreshed via boost:update
to match the installed version.
* fix: honor AI_IMAGE_PROVIDER instead of hardcoding OpenAI's gpt-image-2
AiImageClient always passed model: 'gpt-image-2' to Image::of()->generate(),
so AI_IMAGE_PROVIDER silently did nothing for any provider other than
OpenAI — gemini/xai/openrouter would fail against an OpenAI-only model id
and quietly fall back to a stock photo. Usage recording was hardcoded to
provider 'openai' too, so credits were billed against the wrong model
whenever a different provider actually ran.
Drop the hardcoded model so generation falls through to the SDK's own
config('ai.default_for_images') + per-provider default model, and read
the actual provider/model back off the response's meta for usage
recording and source_meta instead of assuming OpenAI.
* refactor: extract AiImageClient into single-purpose steps, fix uncaught exception
generate() built the prompt, called the SDK, and unpacked the response all
in one block, with bytes extraction happening after the try/catch — so a
response with an empty images collection threw an uncaught RuntimeException
from ImageResponse::firstImage() instead of returning null as documented.
Split into cleanKeywords(), buildPrompt(), resolveBrandContext(), and
toResult(), and moved response unpacking inside the try block so any
malformed response is treated as a failure like everything else. Added a
regression test with an empty-images fake response.
* fix: drop hardcoded default_text_model, resolve model per provider
default_text_model was the only per-modality model override in ai.php —
image, audio, transcription, embeddings, and reranking all just pick a
provider and let it use its own default model. Text had a config-pinned
model on top, forced into every agent's model() and into every usage
log's model field regardless of which provider actually ran. Switching
AI_TEXT_PROVIDER (e.g. to openrouter) kept sending OpenAI's model id to
whichever provider ended up handling the request.
Removed model() from all six agents so laravel/ai resolves the model
from the active provider's own default (OpenAI's default is already
'gpt-5.4', so no behavior change there). Usage-recording call sites now
read the actual provider/model back off the response's meta instead of
assuming config('ai.default')/default_text_model. StreamPostContent
needed the then() callback since broadcast()'s StreamableAgentResponse
doesn't expose meta directly.
* refactor: drop AiConfiguration wrapper, use data_get() for array reads
AiConfiguration was a one-line static helper used by only two call
sites, with no laravel/ai equivalent to lean on (confirmed AiManager
and the Provider base class expose no isConfigured()/hasKey() check —
the package's model is try-then-catch, not pre-flight checks). Inlined
the filled(config(...)) check directly into HandleInertiaRequests and
BrandAnalyzerRunner instead of keeping a class around one line of logic.
Also swapped direct array-key reads for data_get() per project
convention across every file touched by the recent AI provider/model
fixes (agents' budget arrays, RunGenerateNode/StreamPostCreation's
humanizer merge, RegeneratePostMediaImage's baseContext/copy/rendered
access). Write/assignment sites (`$x['key'] = ...`) are left as-is —
data_get() only reads.
* feat: enhance AI configuration with new providers and options
Added strict types declaration and updated Azure OpenAI API version. Introduced new 'bedrock' provider configuration with AWS credentials and role assumptions. Enhanced existing providers with additional options, including image deployment for Azure and OpenAI, and updated URLs for Gemini and Ollama. Added support for an 'openai-compatible' driver to broaden integration capabilities.
* fix: remove trailing newline in AI configuration file
* feat: add support for OpenRouter and ElevenLabs API keys in configuration
Updated the production Docker Compose file and example environment file to include commented-out entries for OPENROUTER_API_KEY and ELEVENLABS_API_KEY. This enhances the configuration options for AI providers, allowing for easier integration of additional services.
* feat: allow per-provider model overrides for every AI modality
Adds a `models` array to each provider block, wired to env vars, so
self-hosted operators can pin a specific text/image/audio/transcription/
embeddings/reranking model instead of relying on the package's built-in
default for that provider. Only added for the modalities each provider
actually implements (verified against laravel/ai's Provider classes).
Azure is left untouched — it resolves models via deployment names
(AZURE_OPENAI_DEPLOYMENT etc.), not raw model strings, which was already
wired before this change.
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* Add workspace MCP settings and token access controls.
Ship MCP settings UI, OAuth revoke/list helpers, Passport deploy wiring,
and workspace.token:mcp gating so assistants can connect without pulling
in welcome/onboarding from the parent epic.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Type MCP client config shapes instead of string checks.
Encode http/config-root on each advanced client and tighten primary
client ids so snippet generation does not branch on magic strings.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Polish MCP settings follow-ups from review.
Translate Ukrainian MCP copy, deep-link ChatGPT into connector
creation, drop an unused asset and revoke arg, and assert PATs are
rejected on the MCP endpoint.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Harden MCP connected clients, revoke scope, and OAuth consent.
List recoverable sessions with live refresh tokens, revoke only PATs,
throttle registration alone, and block viewers from authorizing MCP.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Simplify MCP OAuth route throttling to a single middleware group.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Allow workspace viewers read-only MCP access with web policy writes.
Mirror the web app: MCP connects on view + OAuth mcp:use, write tools
enforce createPost/update/delete/manageAccounts/manageTeam, and demotion
to Viewer keeps grants. Cover role denials, consent, and disconnect.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Harden MCP tool authz with shared workspace helpers.
Route ApiKey tools through AuthorizesMcpTool, fail closed on null user
or policy argument, and resolve the current workspace before mutating.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Drop redundant string casts on validated request data.
Enum::from and validated() fields are already strings, so the casts
add noise without changing behavior.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Show only the current user's MCP connections in settings.
Match API keys privacy: list and disconnect your own OAuth clients,
not teammates' across the account.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Cover LoadWorkspaceFromToken gaps and harden AuthorizesMcpTool tests.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Drop redundant is_string guard before UpdatePostTool find.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Refactor AppSidebar to always show MCP link and simplify route middleware definition in ai.php. The MCP link is now consistently displayed regardless of the current workspace state, and the route middleware syntax has been streamlined.
* Refresh MCP connected clients with Inertia usePoll.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Bump laravel/mcp to 0.9.1 and add the TryPost server icon.
Requires laravel/boost 2.5 for the Icon attribute; expose images/trypost/icon.png on TryPostServer.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Drop no-op ReflectionClass import in TryPostServerTest.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Replace the additive LINKEDIN_EXTRA_SCOPES approach with a single
full-override env var per flow, exploded into an array at the config
layer (config/trypost.php -> platforms.linkedin{,-page}.scopes). This
keeps env values as plain comma-separated strings, lets self-hosters
override the entire set in one place, and removes the controller-side
scope-merge logic.
- config/trypost.php: explode LINKEDIN_SCOPES / LINKEDIN_PAGE_SCOPES
into the scopes arrays (deprecated r_basicprofile stays out of the
personal default)
- LinkedInController: drop resolveScopes(), read config scopes directly
- LinkedInPageController: drop the hardcoded $scopes property, read
config scopes at both call sites
- tests: drive the connect scope assertions from config overrides
- .env.example, docker/.env.docker.example: document LINKEDIN_SCOPES
and LINKEDIN_PAGE_SCOPES
Make r_basicprofile opt-in via LINKEDIN_EXTRA_SCOPES so self-hosted users
unblock by default and ops with legacy/enterprise products keep working.
Why
---
LinkedIn rejects OAuth authorize requests with a generic "Bummer,
something went wrong" page when an app asks for a scope it can't grant.
`r_basicprofile` is a legacy scope deprecated in 2018; new LinkedIn dev
apps don't have it, so every self-hosted user hits the rejection
immediately on `/connect/linkedin`.
The two products LinkedIn actually grants to standard apps today are:
- Sign In with LinkedIn using OpenID Connect → `openid profile email`
- Share on LinkedIn → `w_member_social`
That set is enough for the connect flow. The only piece of data
`r_basicprofile` was buying us is `/v2/me`'s `vanityName` (pretty
`linkedin.com/in/<slug>`). `fetchVanityName()` already handles HTTP
failure gracefully (returns null), and the only downstream consumer —
`LinkedInPagePublisher`'s post-URL builder — already falls back to a
numeric `linkedin.com/feed/update/<id>` URL when `$account->username`
is null.
Backward compatibility
----------------------
Ops with legacy or enterprise LinkedIn products approved on their dev
app (so they DO have `r_basicprofile`) can opt back in via env:
LINKEDIN_EXTRA_SCOPES=r_basicprofile
`LinkedInController::resolveScopes()` merges this comma-separated list
into the default scope array. The connect flow's `Socialite::scopes()`
call then includes the legacy scope, preserving the pre-PR behaviour
end-to-end (including `vanityName` lookup).
Net effect for users without `r_basicprofile`:
- Connect flow works (was previously rejected by LinkedIn).
- Posts publish exactly the same way.
- Generated post URLs use the numeric form instead of the vanity slug.
Tests
-----
- `linkedin connect requests the default scope set when LINKEDIN_EXTRA_SCOPES is unset`
- `linkedin connect appends LINKEDIN_EXTRA_SCOPES to the default scope set`
- Existing `splits comma-separated approvedScopes` fixture updated to
match the new default set.
Add /app/ (Pusher protocol) and /apps/ (REST API) location blocks
with Upgrade/Connection headers so browsers can connect to the
in-container Reverb server via the same port 80/443 endpoint.
- Add libwebp-dev + libavif-dev and --with-webp --with-avif to GD
configure (fixes imagewebp() undefined at runtime)
- Merge wayfinder-gen into asset-build stage: PHP + Node together
so @laravel/vite-plugin-wayfinder can invoke artisan during build
- Pass VITE_* as build args and ENV so they bake into the bundle
- Stub APP_URL=http://localhost so artisan boot doesn't crash on null
- mkdir -p storage/framework/{cache/data,sessions,views} storage/logs
bootstrap/cache before artisan runs (dirs stripped by .dockerignore)
- Production stage: rm vendor + bootstrap/cache/*.php before swapping
to prod-only deps (prevents missing-provider errors from dev cache)
- Switch production chown from app:app to www-data:www-data to match
Alpine php-fpm default pool user
Two adjustments from the PR review.
Stripe placeholders (STRIPE_KEY, STRIPE_SECRET, STRIPE_WEBHOOK_SECRET,
and the eight price IDs) only matter when SELF_HOSTED=false. The
Docker stack defaults to SELF_HOSTED=true, so keeping the block in
.env.docker.example just adds noise for first-time contributors.
upload_max_filesize and post_max_size go from 100M to 1G in both
php.dev.ini and php.prod.ini to accept full-length video uploads.
Bumping nginx client_max_body_size to match — otherwise the proxy
would cut the request off before PHP ever sees it.
Introduces a self-contained Docker build under docker/ that boots the
full Laravel + Vite + Reverb + Horizon + scheduler stack inside a
single container.
The Dockerfile exposes two targets sharing a common system-base layer
(PHP 8.4-FPM Alpine + Postgres/Redis/intl/sockets/redis extensions):
- dev: bind-mount source at runtime, runs Vite via supervisord, hot
reloads PHP via opcache.validate_timestamps=1, UID/GID build args
align container writes with the host user.
- production: ships the prebuilt application — composer --no-dev,
npm run build + build:ssr, wayfinder TS pre-generated, OpCache
hardened, fixed UID 1000.
Sidecar configs (nginx, php.{dev,prod}.ini, supervisord.{dev,prod}.conf,
entrypoint, postgres-init for the test DB) live next to the Dockerfile
so the build context is self-describing. The entrypoint is idempotent
and handles APP_KEY generation, migrations, storage:link, Passport
keys, Wayfinder regen, and dependency reinstall on every boot.
