api-server

name

api-server

description

Nuxt 3 server API routes, middleware, agents, background queues, MCP server, voice pipelines, and server utilities. Use when creating new API endpoints, modifying server middleware, working with the agent runtime, adding server utils, or debugging API errors. Covers route conventions (GET/POST/PUT/DELETE), auth middleware, error handling, and Nitro server architecture. Do NOT use for background job processing (use processor skill) or database schema changes (use database skill).

API & Server

Nuxt 3 server engine (Nitro + H3) powering the Minds AI platform. Handles authentication, AI agent orchestration, real-time streaming, background job processing, MCP protocol, voice pipelines, and third-party integrations.

Architecture

• Framework: Nitro server engine with H3 request handlers
• Auth: Supabase Auth (cookie + Bearer token) + API key auth (minds_ prefix, legacy aox_) + OAuth 2.1 (MCP/ChatGPT)
• Database: Prisma ORM with PostgreSQL (singleton client in server/utils/prisma.ts)
• AI: Vercel AI SDK (streamText/generateObject/generateText) with OpenAI, Anthropic, Google providers
• Observability: Langfuse via OpenTelemetry (@langfuse/tracing + @langfuse/otel)
• Queues: BullMQ (Redis-backed) with nuxt-processor module (defineQueue/defineWorker)
• Real-time: SSE via H3 createEventStream, WebSockets via crossws (defineWebSocketHandler)
• Prompt management: Langfuse prompts with local .md file fallbacks, TTL-based cache in langfuse-prompts.ts

Middleware Chain

All middleware lives in server/middleware/. Global middleware (.global.ts suffix) runs on every request. Named middleware runs only on matching paths.

File	Scope	Purpose
domain-redirect.global.ts	Global	301 redirect art-of-x.com -> getminds.ai
cors-native.global.ts	Global, /api/	CORS headers for Capacitor/Ionic native apps (capacitor://, ionic://)
api-observability.global.ts	Global	Initializes Langfuse OTel via @langfuse/tracing v4 SDK (startActiveObservation/updateActiveTrace), creates session cookie (minds_langfuse_session, reads legacy aox_langfuse_session), injects event.context.trace() helper
auth-bearer.ts	/api/	Extracts Bearer token from Authorization header, validates via Supabase, sets event.context.user and event.context._token
api-auth.ts	/api/v1/ and /v1/	V1 API auth: tries internal auth (x-internal-auth + x-user-id from localhost), then Supabase session, then API key (minds_* / legacy aox_* with hash verification), then OAuth token. Throws 401 if no valid auth (except api-keys and google-chat endpoints)
api-subdomain.ts	api.getminds.ai / api.art-of-x.com host	Routes API subdomain requests: allows /v1/, /mcp, /.well-known/, /oauth/; redirects GET /v1/* to /api/v1/* (307); rewrites non-GET /v1/* internally
portfolio-auth.ts	/api/portfolio/	Resolves user's team membership, attaches event.context.userId and event.context.teamId

API Route Structure

All routes use Nuxt file-based routing in server/api/. HTTP method is the file suffix (.get.ts, .post.ts, etc.).

server/api/auth/ — Authentication

• check-email.post — Check if email exists
• confirm-email.post — Confirm email verification
• confirm-password-reset.post — Complete password reset
• create-profile.post — Create user profile after signup
• custom-signup.post — Custom signup flow
• reset-password.post — Initiate password reset

server/api/spark/ — Spark (AI persona) management

• index.get / index.post / index.delete — List, create, bulk delete sparks
• [id].get / [id].put / [id].patch / [id].delete — CRUD single spark
• messages.post — Main chat endpoint: streams AI response via SSE (uses runAgentCompletion, subject to chat rate limiting — 30/min authenticated, 10/min public)
• create-from-input.post — Create spark from natural language input
• collect-data.post / collect-data-demo.post — Trigger data collection (dispatches auto-spark queue job)
• generate-profile.post / generate-prompt-component.post / profile-image.post — AI-powered generation
• progress/[sparkId]/stream.get — SSE stream for pipeline progress (uses progressEventBus)
• pipeline-status/[sparkId].get / collection-status/[sparkId].get — Polling status endpoints
• share.put / shared/[shareId].get — Public link sharing
• favorites.get / favorites.post / favorites.delete — Favorite sparks
• search-sources.post — Search for training sources
• [id]/chat-history.get / [id]/chat-history.post — Chat history CRUD
• [id]/invite/index.post / [id]/members/* / [id]/invitations/* — Collaboration
• [id]/sms-number.get / [id]/sms-number.post — Twilio phone number assignment
• [id]/regenerate-embeddings.post — Dispatches regenerate-embeddings queue job
• [id]/profile-image*.ts — Profile image upload (signed URL flow)

server/api/flows/ — Flow (multi-spark conversation) management

• index.get / index.post / index.delete — List, create, bulk delete
• [id]/index.get / [id]/index.put / [id]/index.delete — CRUD single flow
• [id]/messages/index.get / [id]/messages/index.post — Get/send messages (POST streams AI via SSE with rate limiting — 30/min per user)
• [id]/messages/summary.post — Generate conversation summary
• [id]/stream.get — SSE endpoint for real-time flow updates (uses broadcast.ts connection registry)
• [id]/sparks/* — Manage sparks within a flow
• [id]/ideas/* — Idea CRUD with comments, images, feedback, promotion
• [id]/guided/* — Guided curator: keyword generation, selection, curator agent
• [id]/board.ts / [id]/board/image.post — Board state management
• [id]/cancel.post / [id]/typing.post — Cancel generation, typing indicators
• [id]/invite/* / [id]/members/* — Flow collaboration
• shared/[shareId].get / shared/[shareId]/stream.get — Public shared flows

server/api/v1/ — Versioned Public API (API key / OAuth auth via api-auth.ts middleware)

• api-keys/index.get / index.post / [keyId].delete — API key management (requires subscription)
• auth/me.get — Current user info
• sparks/index.get / index.post — List/create sparks
• sparks/[sparkId].put / [sparkId].delete — Update/delete spark
• sparks/[sparkId]/completion.post — Chat completion (streams JSON response with citations)
• sparks/[sparkId]/knowledge/* — Knowledge items CRUD and patterns
• sparks/[sparkId]/regenerate-prompt.post — Regenerate system prompt
• google-chat/webhook.post — Google Chat bot integration
• user/shareable-sparks.get — List shareable sparks

server/api/billing/ — Stripe integration

• checkout.post / portal.post / session.get / subscription.get — Checkout, portal, session info
• academic.post — Academic plan application
• webhook.post — Stripe webhook handler (raw body, signature verification via readRawBody)

server/api/cron/ — Scheduled tasks (all protected by Bearer {CRON_SECRET})

• cleanup-stuck-jobs.post — Fixes sparks stuck in queued/running status (every 15 min)
• cleanup-unused-numbers.post — Releases Twilio numbers unused for 2 weeks (daily)
• renew-calendar-webhooks.post — Renews Google Calendar webhook channels expiring within 2 days (daily)
• watch-knowledge-items.post — Checks watched links for content changes, dispatches watch-knowledge-items BullMQ job
• backfill-voice-profiles.post — Generates voiceProfiles for sparks missing them (cursor-based pagination, 50/batch)
• validate-voice-pool.post — Validates all Fish Audio voice IDs in the curated pool, reports invalid/expired voices
• demo-analytics.post — Refreshes demo analytics data via BullMQ (every 4 hours)
• reclone-voices.post — Triggers voice cloning for public sparks without voice clones, searches YouTube for interview videos (supports sparkIds filter, includeRetry, dryRun mode)
• classify-spark-voices.post — Migration: classifies voice gender for existing sparks
• migrate-voice-webhooks.post — Migration: configures voice webhooks on existing phone numbers

Other route groups

• server/api/team/ — Team CRUD, invitations, members, integrations
• server/api/user/ — Profile, preferences, number lookup, delete-all
• server/api/portfolio/ — Portfolio items CRUD, file access, processing status, retrigger
• server/api/content/ — Content processing (items, messages, uploads)
• server/api/notifications/ — Push notification registration, preferences, test
• server/api/integrations/google-calendar/ — OAuth flow, status, disconnect
• server/api/public/ — Unauthenticated: demo flow, public sparks, widget config, room info
• server/api/chat/ — Chat history and recent messages
• server/api/email/ — Email preferences, digest cron, test send
• server/api/health/ — Health checks (general, content-processor, data-collector)
• server/api/webhooks/google-calendar.post — Google Calendar push notifications
• server/api/voice/ — transcribe.post (Deepgram STT), webhook.post (Twilio voice webhook)
• server/api/sms/webhook.post / server/api/whatsapp/webhook.post — Twilio SMS/WhatsApp inbound
• server/api/trainer/message.post — Trainer agent chat endpoint
• server/api/evaluate/reliability.post — Evaluator agent for response reliability scoring
• server/api/patterns/ — Pattern analysis endpoints
• server/api/competencies/ — Competency framework data
• server/api/widgets/spark.get — Widget embed configuration
• server/api/admin/ — Demo analytics, spark analytics, user analytics
• server/api/uploads/ — Signed URL generation, upload confirmation
• server/api/geo/currency.get — Geo-based currency detection
• server/api/generated-images/ — Proxy for generated image assets
• server/api/preview/og.get — Open Graph preview generation
• server/api/context/search-sources.post — Context source search
• server/api/oauth/validate-token.post — OAuth token validation
• server/api/debug/auth-check.get — Debug auth state check

Non-API Routes (server/routes/)

• mcp.ts — MCP protocol handler at /mcp (StreamableHTTP transport, rate limiting, circuit breakers)
• bull-board.ts + bull-board/[...].ts — Bull Board UI at /bull-board (Basic auth via runtime config)
• oauth/authorize.get / callback.get / token.post / register.post — OAuth 2.1 + PKCE for MCP clients (ChatGPT, Claude Desktop, Cursor)
• .well-known/oauth-authorization-server.get / oauth-protected-resource.get / openid-configuration.get — OAuth/OIDC discovery
• api/voice/stream.ts — WebSocket: Twilio Media Streams voice bridge (mulaw 8kHz <-> PCM 24kHz)
• api/voice/browser-stream.ts — WebSocket: Browser voice mode (PCM16 24kHz, JSON control messages)
• v1/sparks/[sparkId]/completion.post — V1 completion (alternate route path)
• widget/spark.html.get — Embeddable widget HTML
• sitemap.xml.ts — Dynamic sitemap generation
• admin/auth.ts / callback.ts / config.yml.ts — Admin CMS (Decap CMS) auth

Handler Patterns

Authenticated GET with router params

export default defineEventHandler(async (event) => {
  const user = await requireAuthenticatedUser(event) // from server/utils/auth.ts
  const id = getValidatedRouterParam(event, 'id', 'Spark ID') // UUID validation
  const spark = await prisma.spark.findUnique({ where: { id } })
  if (!spark) throw createError({ statusCode: 404, statusMessage: 'Spark not found' })
  return spark
})

POST with body and observability tracing

export default defineEventHandler(async (event) => {
  const user = await requireAuthenticatedUser(event)
  const body = await readBody(event)
  if (!body.sparkId) throw createError({ statusCode: 400, statusMessage: 'sparkId is required' })
  return await event.context.trace('spark-message', async () => {
    // AI agent call inside trace
    return await runAgentCompletion({ event, ... })
  }, { sparkId: body.sparkId, userId: user.id, agentType: 'spark' })
})

SSE streaming response

export default defineEventHandler(async (event) => {
  setHeader(event, 'Content-Type', 'text/event-stream; charset=utf-8')
  setHeader(event, 'Cache-Control', 'no-cache, no-store, must-revalidate, no-transform')
  setHeader(event, 'Connection', 'keep-alive')
  setHeader(event, 'X-Accel-Buffering', 'no')
  setHeader(event, 'Content-Encoding', 'none')
  // Use createEventStream or manual write via event.node.res.write()
})

Webhook with raw body (Stripe pattern)

export default defineEventHandler(async (event) => {
  const body = await readRawBody(event, 'utf8')
  const sig = getRequestHeader(event, 'stripe-signature')
  const stripeEvent = stripe.webhooks.constructEvent(body, sig, webhookSecret)
  // handle event...
})

Cron job with secret verification

export default defineEventHandler(async (event) => {
  const authHeader = getHeader(event, 'authorization')
  if (authHeader !== `Bearer ${useRuntimeConfig().cronSecret}`) {
    throw createError({ statusCode: 401, statusMessage: 'Unauthorized' })
  }
  // run cron logic...
})

Agent System (server/agents/)

The agent runtime (runtime.ts) provides runAgentCompletion() — the central function for all AI interactions. It:

1. Fetches prompt config from Langfuse (model, temperature, maxTokens, provider)
2. Creates the AI provider (OpenAI, Anthropic, or Google via Vercel AI SDK)
3. Calls streamText() with tools, system prompt, messages, and abort signal
4. Records generation observations to Langfuse

Agent	File	Purpose
runtime	runtime.ts	Core agent runner: runAgentCompletion() with multi-provider support, tool execution, Langfuse tracing
spark	spark.ts	Generates spark system prompts from Langfuse templates with temporal context
flow	flow.ts	Generates flow context prompts and moderator prompts for multi-spark conversations
idea	idea.ts	Generates idea context prompts for idea refinement within flows
trainer	trainer.ts	Trainer agent with DB query, RAG, portfolio, and web search tools
evaluator	evaluator.ts	Scores AI response reliability (0-100) against the spark's system prompt
guidedCurator	guidedCurator.ts	Curates guided flow sessions, suggests spark participants
content-filter	content-filter.ts	LLM-based content relevance filtering (multi-provider)
pattern-analyzer	pattern-analyzer.ts	Analyzes content for thinking/behavioral patterns using frameworks

Tool Orchestrator (server/utils/tool-orchestrator.ts)

getToolsForContext() returns the appropriate tool set based on context:

• SPARK_CHAT_TOOLS: RAG, image generation, web search, display image, document processing, link analysis
• TRAINER_CHAT_TOOLS: RAG, web search, link analysis, document processing, image display + portfolio management, training content, system prompt editing, user preferences
• FLOW_CHAT_TOOLS: Similar to spark tools
• IDEA_AGENT_TOOLS: RAG, idea CRUD, comments, images, endorsement

Tool implementations live in server/utils/tools/: rag.ts, web.ts, image.ts, db.ts, chat.ts, portfolio.ts, file.ts, spark.ts, idea.ts, user.ts.

Queue System (BullMQ)

Queues are defined in server/queues/ using defineQueue({ name }) from nuxt-processor. Workers are in server/workers/ using defineWorker({ name, processor, options }). All use Redis connection from environment.

Queue	Worker	Concurrency	Purpose
auto-spark	auto-spark.ts	2	Data collection pipeline: scrapes sources, generates embeddings/patterns, classifies voice, attempts voice cloning from YouTube
regenerate-embeddings	regenerate-embeddings.ts	1	Batch re-embeds all portfolio items + existing embeddings using text-embedding-3-large (1536 dims). Handles SIGTERM gracefully
reprocess-portfolio-item	reprocess-portfolio-item.ts	1	Re-processes a single portfolio item (embeddings + patterns)
demo-analytics	demo-analytics.ts	1	Runs demo analytics aggregation
watch-knowledge-items	watch-knowledge-items.ts	1	Checks watched portfolio items for content changes (social profiles via Apify, links via Tavily), queues reprocessing if updated
whatsapp-sync	whatsapp-sync.ts	2	Syncs spark profile to Twilio WhatsApp sender, polls for ONLINE status
concept-feedback	(uses conceptFeedbackQueue.ts util)	—	Queues concept feedback processing

Worker Prisma Pattern: Workers must lazy-load Prisma via const { prisma } = await import('~/server/utils/prisma') to prevent DATABASE_URL initialization crashes at startup. Top-level imports are not supported in workers.

Bull Board UI: Available at /bull-board with Basic auth (runtimeConfig.bullboard.username/password). Configured in server/handlers/bull-board.ts.

Job dispatch pattern:

import autoSparkQueue from '~/server/queues/auto-spark'
await autoSparkQueue.add('collect', { sparkId, userId, entityNames, ... })

MCP Server (server/mcp/)

Model Context Protocol server for external AI clients (ChatGPT, Claude Desktop, Cursor).

• Transport: StreamableHTTP via @modelcontextprotocol/sdk, session-based (session ID in headers)
• Auth: Bearer API key (minds_* / legacy aox_*) or OAuth 2.1 with PKCE
• Route: /mcp (handled by server/routes/mcp.ts)
• Tools: list_my_ai_personas, create_ai_persona_or_digital_twin, talk_to_ai_persona, check_ai_persona_training_progress
• Resources: ai-persona-widget (sparkWidget.ts — unified widget for creation and chat modes, loaded via widgetLoader.ts)
• Middleware: Rate limiting (server/mcp/middleware/rateLimit.ts), CORS (config.ts)
• Utilities: Circuit breakers, audit logging, metrics, error codes, fuzzy matching, token/cache management, validation
• Standalone: Session-based HTTP server (main.ts), stateless HTTP server (http.ts), stdio (stdio.ts + stdioServer.ts)

Voice Pipeline

Real-time voice conversations via STT -> LLM -> TTS pipeline.

• Orchestrator (server/utils/voice/orchestrator.ts): Manages single voice session lifecycle
• STT: Deepgram streaming (server/utils/voice/deepgram-stt.ts)
• TTS: Fish Audio streaming (server/utils/voice/fish-audio-tts.ts)
• Twilio bridge (server/routes/api/voice/stream.ts): WebSocket for phone calls (mulaw 8kHz <-> PCM 24kHz conversion via audio-convert.ts)
• Browser bridge (server/routes/api/voice/browser-stream.ts): WebSocket for browser voice mode (PCM16 24kHz)
• Voice cloning (server/utils/voice-cloning/): YouTube audio extraction, speaker identification, voice cloning
• Voice classifier (server/utils/voice-classifier.ts): Classifies spark voice gender, assigns default voice IDs

Server Utilities Catalog (server/utils/)

Utility	Purpose
auth.ts	getAuthenticatedUser(), requireAuthenticatedUser(), ensureUserProfile() — Bearer + cookie auth, skips minds_/aox_ API key tokens
spark-auth.ts	determineSparkAccess() — granular access control (owner, collaborator, public, session)
validation.ts	validateAuthenticatedUser(), validateEmail(), validateUUID(), getValidatedRouterParam(), handleApiError()
prisma.ts	Singleton PrismaClient with global caching in dev, no-op proxy client during Nitro prerender
subscription-guard.ts	requireSubscription() — enforces plan check (premium/lite/academic/team)
broadcast.ts	In-memory SSE connection registry for flows (addFlowConnection/removeFlowConnection/broadcastMessageToFlow)
progress-event-bus.ts	In-memory pub/sub for spark pipeline progress updates (SSE)
langfuse-prompts.ts	Prompt loading from Langfuse with TTL cache + local .md fallback. MAX_CACHE_SIZE pattern
tool-orchestrator.ts	getToolsForContext() — returns tool sets by context type with Langfuse observation recording
ai-helpers.ts	isReasoningModel() — checks if model supports reasoning
embeddings.ts	OpenAI embedding creation (text-embedding-3-large)
content-processor.ts	Processes portfolio items: extracts text, generates embeddings + patterns
stripe.ts	Stripe client, plan resolution, checkout session creation
email.ts	Email sending (daily digest, notifications)
twilio.ts	Twilio phone number management, SMS/WhatsApp/voice webhook handling
image-generation.ts	AI image generation (DALL-E, Flux)
image-storage.ts	Supabase Storage image upload/management
image-analysis.ts	Vision API image analysis
face-detection.ts	Face detection for profile images
social-scraper.ts	Social media profile scraping (Apify)
social-content-processor.ts	Processes scraped social content into portfolio items
seo-keywords.ts	SEO keyword extraction and analysis
competitor-analysis.ts	Competitive analysis for sparks
chat-rate-limiter.ts	checkChatRateLimit() — sliding-window rate limiting for chat endpoints (per-user or per-IP, operation-type specific limits: spark_chat/flow_chat 30/min, trainer/idea 20/min, unauthenticated 10/min; returns 429 with Retry-After and X-RateLimit-* headers)
context-compression.ts	Compresses conversation context to stay within token limits
context-documents.ts	Builds context document summaries for guided flows
flow-conversation.ts	Flow message handling, tool metadata parsing
flow-moderator.ts	Determines next speaker in multi-spark flows
flow-name-generator.ts	AI-generated flow names
spark-creation.ts	Spark creation logic (from input, from demo)
spark-profile.ts	Spark profile generation (system prompt, description)
spark-shaping.ts	Spark shaping utilities
spark-avatar.ts	Profile image generation for sparks
sparks.ts	Spark query utilities
generate-prompt-components.ts	Generates individual prompt components (bio, style, expertise)
generateEditSummary.ts	Summarizes edits to spark system prompts
frameworks.ts	Thinking framework aspects/sub-aspects (loaded from server/data/frameworks.json)
sparkThinkingAspects.ts	Builds thinking aspects section for spark prompts
ideaFeedback.ts	Idea feedback processing
conceptFeedbackQueue.ts	BullMQ-based concept feedback processing
board-state.ts	Flow board state management
demo-flow.ts	Demo flow creation and management
demo-analytics.ts	Demo analytics aggregation
analytics.ts	Event tracking (PostHog/Langfuse)
crypto.ts	hashKey(), verifyKey() — API key hashing
validateOAuthToken.ts	OAuth token validation
locale.ts	Language detection, user locale, language instructions
geo.ts	Geo-based currency detection
push.ts	Web push notification sending
team.ts	Team creation/management utilities
user-preferences.ts	User preference storage
citations.ts	Maps RAG/web sources to citation objects
tool-labels.ts	Localized tool display labels
pipeline-status.ts	updateSparkPipelineStatus() — writes pipeline status to DB
terminal-progress.ts	Terminal progress display for CLI tools
portfolio-processor.ts	Portfolio item processing pipeline
rag.ts	RAG retrieval utilities
google-calendar-*.ts	Google Calendar OAuth, sync, webhook, enrichment, event updates
google-chat.ts	Google Chat bot message handling
plugins/convoy.ts	Convoy A/B testing integration (gated by env vars)
widgets/	Embeddable widget generation (Vue app, styles, sphere graph)
data-collection/	Full data collection pipeline: search (Tavily, YouTube/Apify), content processing, chunking, embedding generation, pattern extraction, quality filtering, spark generation
voice-cloning/	Voice cloning pipeline: YouTube audio extraction, speaker identification, clone voice
voice/	Voice session orchestrator, Deepgram STT, Fish Audio TTS
safety-filters.ts	Input sanitization (sanitizeInput()) and output validation (validateOutput()) for prompt injection/leakage detection
agent-errors.ts	Error classification (classifyError()), retry logic (withRetry()), per-tool timeout config (getToolTimeout())
conversation-memory.ts	LLM-based rolling conversation summaries with DB persistence
observability/	Langfuse initialization, generation recording (recordGenerationObservation), error event recording (recordErrorEvent), demo funnel tracking (recordDemoEvent), evaluation queueing

Key Environment Variables

• DATABASE_URL / DIRECT_URL — PostgreSQL connection
• SUPABASE_URL / SUPABASE_ANON_KEY / SUPABASE_SERVICE_ROLE_KEY — Supabase
• OPENAI_API_KEY / OPENAI_ORG_ID — OpenAI
• ANTHROPIC_API_KEY — Anthropic
• GOOGLE_API_KEY — Google AI
• LANGFUSE_PUBLIC_KEY / LANGFUSE_SECRET_KEY — Langfuse observability
• STRIPE_SECRET_KEY / stripeWebhookSecret — Stripe billing
• CRON_SECRET — Cron job authorization
• TWILIO_ACCOUNT_SID / TWILIO_AUTH_TOKEN — Twilio voice/SMS
• DEEPGRAM_API_KEY — Deepgram STT
• FISH_AUDIO_API_KEY — Fish Audio TTS
• CONVOY_PROXY_URL / CONVOY_SECRET — Convoy A/B testing (optional)
• bullboard.username / bullboard.password — Bull Board UI auth (runtime config)
• demoFlowOwnerId — Demo flow owner user ID (runtime config)

Nitro Configuration (from nuxt.config.ts)

• experimental.wasm: true / experimental.websocket: true — Enable WASM and WebSocket support
• externals.external: ['@prisma/client', '.prisma/client'] — Exclude Prisma from bundling
• Route rules: Disable buffering for /api/flows/*/messages, /api/flows/*/stream, /api/public/flow-messages, /mcp
• /api/billing/webhook has bodyParser: false for raw Stripe signature verification

Related Files

• server/api/ — All API route handlers
• server/routes/ — Non-API routes (MCP, Bull Board, OAuth, WebSocket, widget, sitemap)
• server/middleware/ — Server middleware (auth, CORS, observability, subdomain)
• server/agents/ — AI agent implementations (runtime, spark, flow, idea, trainer, evaluator, etc.)
• server/utils/ — Server utilities (auth, DB, AI, integrations, pipelines)
• server/utils/tools/ — AI tool implementations (RAG, web, image, DB, portfolio, etc.)
• server/utils/voice/ — Voice pipeline (orchestrator, Deepgram STT, Fish Audio TTS)
• server/utils/voice-cloning/ — Voice cloning pipeline
• server/utils/data-collection/ — Data collection pipeline (search, content, embeddings, patterns)
• server/utils/observability/ — Langfuse initialization and recording
• server/utils/widgets/ — Embeddable widget generation
• server/queues/ — BullMQ queue definitions
• server/workers/ — BullMQ worker implementations
• server/handlers/ — Shared handlers (Bull Board)
• server/mcp/ — MCP server (tools, resources, config, middleware, utilities)
• server/data/frameworks.json — Thinking framework definitions
• nuxt.config.ts — Nitro/route rules configuration