Platform For AI:RAG and web search chatbot with DeepSeek

Prerequisites

This template combines web search for real-time data with RAG retrieval from domain-specific knowledge bases.

• Register at the SerpApi website and obtain an API key (free tier: 100 searches/month).

• Select a vector database:

  • Faiss: For testing, no setup required

  • Milvus: For production with larger datasets. Create a Milvus instance first

• Upload the RAG knowledge base corpus to OSS.

Deploy models (optional)

Skip this step if you already have OpenAI API-compatible model services.

Go to QuickStart > Model Gallery and deploy models for these scenarios:

• Select large-language-model for Scenarios. In this example, DeepSeek-R1 is used. For more information, see Deploy DeepSeek-V3 and DeepSeek-R1 models.

  

• Select embedding for Scenarios. In this example, bge-m3 embedding model is used.

  

Create connections

Create LLM service connection

1. Go to LangStudio and select a workspace.

2. On the Connection > Model Service tab, click New Connection.

3. Create a General LLM Model Service connection.

Key parameters:

Create embedding model service connection

Follow the same steps as above and select General Embedding Model Service type.

Create SerpApi connection

1. On the Connection > Custom Connection tab, click New Connection.

2. Configure the api_key obtained in Prerequisites.

Create knowledge base index

Create a knowledge base index to parse, chunk, and vectorize the corpus into a vector database. For complete configuration details, see Manage knowledge base indexes.

Create and run application flow

1. On the LangStudio Application Flow tab, click Create Application Flow. Select the Chatbot with RAG and Web Search template.

   

2. Click Select Runtime in the upper-right corner and select an existing runtime. If no runtime exists, click Create Runtime on the Runtime tab.

   

   Note: Start the runtime before parsing Python nodes or viewing additional tools.

   VPC configuration: When using Milvus, configure the same VPC as the Milvus instance, or ensure VPC interconnection. When using FAISS, no VPC configuration is required.

3. Configure key nodes:

   

   • Knowledge Retrieval:

     • Index Name: Select the index created earlier

     • Top K: Number of matching results to return

   • Serp Search:

     • SerpApi Connection: Select the connection created earlier

     • Engine: Supports Bing, Google, Baidu, Yahoo, etc. See the SerpApi website for details

   • LLM:

     • Model Configuration: Select the connection created earlier

     • Chat History: Whether to use chat history as input

   For more information about each node, see Node components.

4. Click Run in the upper-right corner to execute the application flow. For common issues, see LangStudio FAQ.

   

5. Click View Logs below the generated answer to view trace details or topology.

   

Deploy application flow

On the application flow development page, click Deploy in the upper-right corner to deploy as an EAS service. Key parameters:

• Resource Information > Instances: For testing, set to 1. For production, configure multiple instances to avoid single points of failure.

• VPC: SerpApi requires internet access. Configure a VPC with internet access capability. See Service internet access. When using Milvus, ensure VPC connectivity with the Milvus instance.

For more deployment details, see Deploy and call application flows.

Call the service

After deployment, test the service on the Online Debugging tab of the EAS service details page.

For more calling methods (such as API calls), see Call the service.

References

• LangStudio FAQ

• LangStudio pricing

• Manage connections

• Manage knowledge base indexes

• Develop application flows

• Deploy and call application flows