AI Content Platform with Site and Search

Scenario

Use this architecture when building an enterprise-grade content platform that requires zero-touch content generation, instant public site deployment, and secure, AI-optimized internal search. It’s ideal for teams needing a Bailian-driven autonomous agent to curate Notion databases, publish marketing pages via Vercel, and index knowledge articles into an IDaaS-protected Elasticsearch cluster with PAI-trained semantic ranking.

Integration steps

1. Initialize Notion CMS & MCP Server: Create a Notion database and generate an Internal Integration token with content:read and content:write scopes. Deploy the notion-build-ai MCP server and configure mcp.json:

``json { "notion_token": "ntn_...", "database_id": "your_db_id", "server_url": "https://mcp.yourdomain.com" } ``

2. Configure Bailian AI Agent: In Alibaba Cloud Bailian, create an agent workflow. Attach the MCP server endpoint and set system prompts to output structured JSON matching your Notion schema. Enable M2M auth via IDaaS:

``bash aliyun bailian CreateAgent --Name content-manager --McpEndpoint https://mcp.yourdomain.com --AuthProvider idaas ``

3. Provision Infrastructure with Terraform: Deploy ECS, OSS, and an Elasticsearch cluster secured by IDaaS. Use the provided Terraform module:

``bash terraform apply -var="es_cluster_name=ai-search-prod" -var="idaas_client_id=$IDAAS_CLIENT" ``

4. Set Up Vercel Deployment Pipeline: Connect your Git repo to Vercel. Configure vercel.json for ISR and Cloudflare edge routing:

``json { "rewrites": [{ "source": "/api/content", "destination": "https://mcp.yourdomain.com/sync" }] } ``

5. Ingest & Rank with PAI + Elasticsearch: Train a semantic ranking model in PAI Studio. Export the model to OSS, then configure the ES ingest pipeline:

``json PUT _ingest/pipeline/ai-rank { "processors": [{ "pai_ranking": { "model_id": "oss://pai-models/v1", "field": "content_vector" } }] } ``

6. Wire Dual-Publish Webhook: In Notion, set up a webhook to trigger on page.created. The Bailian agent intercepts, pushes markdown to Vercel via POST /api/deploy, and streams JSON to ES via POST /_bulk?pipeline=ai-rank.

Architecture

Data originates from the Bailian AI agent, which writes structured content to Notion as the single source of truth. A webhook/MCP bridge triggers a parallel pipeline: (1) Vercel fetches updated Notion pages, runs SSG/ISR, and serves the public site behind Cloudflare’s WAF and edge cache; (2) Simultaneously, raw text and metadata are pushed to an IDaaS-secured Elasticsearch cluster. An ingest pipeline enriches documents with embeddings, which are then ranked by a PAI-trained model. All search queries are gated by IDaaS JWT validation.

Prerequisites

• Alibaba Cloud account with Bailian, IDaaS, ECS, Elasticsearch, PAI, and OSS enabled
• Notion workspace with Internal Integration token (content:read, content:write)
• Vercel project linked to a Git repository
• Cloudflare account with DNS and WAF configured
• Terraform CLI v1.5+ and Node.js runtime for MCP server

Common pitfalls

• Notion API Rate Limits: Bailian’s batch generation can exceed 3 req/sec. Implement exponential backoff in the MCP server and batch writes via POST /v1/pages with children arrays.
• IDaaS Token Expiry: ES connections drop when short-lived JWTs expire. Use IDaaS refresh tokens with a background cron job to rotate Authorization: Bearer headers in the ES client.
• PAI Model Version Drift: If the PAI model is retrained without updating the ES pipeline config, ranking scores degrade. Pin model_id in Terraform and use versioned OSS paths.
• Vercel ISR Cache Staleness: Cloudflare edge cache may serve outdated pages if Cache-Control isn’t explicitly set to s-maxage=0, stale-while-revalidate=300 on Notion webhook triggers.

Typical questions

• AI content pipeline feeding website and search platform
• Bailian manages Notion for both Vercel site and Elasticsearch
• single AI agent powers public site and enterprise search
• unified AI content system with Vercel and Elasticsearch
• 百炼AI同时管理网站内容和搜索知识库
• 一个AI智能体维护Vercel网站加Elasticsearch搜索
• Notion统一内容源对接Vercel和搜索平台
• AI驱动的内容加搜索一体化平台

Q: How does the Bailian AI agent manage Notion content to power both the public website and enterprise search? A: A Bailian AI agent autonomously manages all organizational content in Notion to simultaneously publish marketing pages to a Vercel-hosted public website and ingest knowledge articles into an IDaaS-secured Elasticsearch cluster. This unified AI-driven platform leverages PAI-trained ranking models to optimize search accuracy while maintaining secure access through IDaaS integration.