Build Q&A chat over unstructured text in minutes
Using no-code agents, you can create text-to-vector-database pipelines and prompt templating agents for a variety of LLMs.
![A animation of a chatbot powered by LangStream](/images/ls-chatbot-gif.gif)
Use the latest Gen AI technologies
LangStream integrates with LLMs from OpenAI, Google Vertex AI, and Hugging Face, Vector databases like Pinecone and DataStax, Python libraries like LangChain and LlamaIndex.
![Logos from OpenAI, Google Vertex AI, Hugging Face, Pinecone, DataStax, Milvus, LangChain, LlamaIndex](/images/langstream-logos-2.png)
Manage embeddings in your vector database
Create pipelines that automatically create and update embeddings in your vector database for unstructured data (PDFs, Word documents, HTML) using proprietary and open-source embeddings models.
![Abstract image showing vector embeddings for structured and unstructured data](/images/vectors.jpeg)
Code and deploy in Visual Studio Code
To make it easier to build LangStream apps, we have created a VS Code extension that includes app deployment, app updating, log analysis, and code snippets.
![Screen shot from LangStream Visual Studio Code Extension](/images/vs-code.png)
Declare an app and deploy it to dev or prod
In LangStream an application consists of pipelines defined using configuration files or Python. Deploy to a local development environment or to a production environment, powered by proven tech: Kubernetes and Kafka.
![Screenshot of the LangStream CLI](/images/ls-deploy-screen.gif)
Bring your existing data to the LLM
LangStream can tap into the data you already have when engineering a prompt to send to the LLM. Query SQL and no-SQL databases or use Kafka Connect connectors to access the data you already have.
![Logos for Kafka Connect, MongoDB, Cassandra, Postgres, Redis, MySQL, and Snowflake](/images/data-integration-logos.png)
Fix your day 2 problems with an event-driven architecture
LangStream is a streaming framework using an event-driven architecture meaning it is asynchronous, decoupled, fault tolerant, and scalable.
![Screenshot of the LangStream pipeline in the UI](/images/langstream-screen-shot.png)