LLMStack
LLMStack is a no-code platform that enables users to build generative AI apps, chatbots, and agents using their own data. With LLMStack, us...
Last verified:
What is LLMStack?
LLMStack is an open-source, no-code/low-code platform for building generative AI agents, workflows, chatbots, and applications by chaining multiple LLMs together. It supports all major model providers including OpenAI, Cohere, Stability AI, Hugging Face, Anthropic's Claude, and more, allowing users to easily combine different models to create powerful AI applications. The platform enables text generation, question answering, text summarization, translation, media generation, classification, and complex multi-step AI workflows.
Key features include a visual no-code builder with a web-based UI for creating AI apps without coding experience, built-in vector database for storing and augmenting LLMs with user data, HTTP API with streaming and non-streaming support for integrating AI experiences into existing applications, multi-tenant organization support with granular permissions, and the ability to connect to external data sources like CSV, PDF, URLs, YouTube, Google Drive, Notion, websites, and direct uploads. LLMStack also supports agents that can autonomously perform tasks using processors as tools, jobs for batch processing with CSV inputs, Slack and Discord integrations for triggering AI chains, and app versioning with testing capabilities.
LLMStack is designed for businesses, developers, and anyone wanting to build AI applications without coding expertise. It targets organizations needing to create custom chatbots, AI assistants for workflow automation, AI agents for day-to-day tasks, and AI applications for various business purposes. The platform supports both cloud deployment via Promptly (their hosted enterprise-grade service) and on-premise deployment on user infrastructure, making it suitable for companies with security or compliance requirements. It's particularly useful for teams needing multi-agent systems, data-integrated AI applications, and workflow automation.
The platform includes built-in processors for OpenAI's ChatGPT, image generation, Stability's image generation, and more, with support for creating custom processors. Features like connections for storing encrypted credentials for external services (databases, APIs, LinkedIn, Hubspot), variables for dynamic values, app sharing with viewer/collaborator roles, history tracking for all app runs, and a playground for testing processors before using them in apps make it a comprehensive end-to-end solution for AI application development.
LLMStack pricing
Pricing model: Free
LLMStack is open-source and free to install locally using pip install llmstack. For cloud deployment, Promptly offers a free tier that includes the full platform with AI Assistant for 3 users, 10k metrics, 50GB logs, and 50GB traces. It is free to get started with Promptly at trypromptly.com without local installation. The platform supports both cloud deployment via Promptly (fully managed, enterprise-grade service) and on-premise deployment on user infrastructure. Prices may vary based on usage volume and selected features for paid plans. Contact sales is available for custom enterprise pricing.
LLMStack pros
- No-code builder enables building AI apps without coding experience
- Supports all major model providers (OpenAI, Cohere, Stability AI, Hugging Face, Claude)
- Chain multiple LLMs together for complex AI applications
- Built-in vector database for data augmentation
- Import data from diverse sources (CSV, PDF, URLs, YouTube, Google Drive, Notion)
- HTTP API with streaming and non-streaming support
- Multi-tenant organization support with granular permissions
- On-premise deployment option for security/compliance
- Cloud hosting available via Promptly
- Agents that autonomously perform tasks using processors as tools
- Jobs page for batch processing with CSV inputs
- Slack and Discord integrations for triggering AI chains
- App versioning and testing capabilities
- Playground for testing processors before app deployment
- History tracking for all app and playground runs
- Connections for encrypted external service credentials
- Variables for dynamic values in processors
- Granular permission model with viewer and collaborator roles
- Public app sharing or restricted access options
- Default admin account with easy local installation via pip
LLMStack cons
- Requires WSL2 on Windows for local installation
- Default admin password (promptly) must be changed after first login
- Needs Docker images downloaded on first start (can take minutes)
- API requires token authentication from Settings page
- Model provider keys must be added manually in Settings
- Configuration file location may be unclear for new users
- Port and settings require config file edits for server deployment
- Streaming API adds complexity for implementation
- Session management requires tracking session_id for chat history
- Custom processor creation requires additional development work
Frequently asked questions about LLMStack
What is LLMStack?
LLMStack is an open-source, no-code/low-code end-to-end platform that allows you to build powerful AI applications using LLMs. You can build AI applications that generate text, answer questions, summarize text, translate text, generate media, classify text, and more. It comes with a no-code builder, built-in vector database, HTTP API access, and supports deployment to cloud or on-premise.
What model providers does LLMStack support?
LLMStack supports all major model providers including OpenAI, Cohere, Stability AI, Hugging Face, Anthropic's Claude, and more. You can easily use these models to build powerful apps and chain multiple LLMs together to build complex generative AI applications.
How do I install LLMStack locally?
You can install LLMStack locally using pip install llmstack. Once installed, start it using llmstack command. It will automatically open your browser to the login page at http://localhost:3000. On Windows, you must use WSL2 (Windows Subsystem for Linux) to install LLMStack.
What is Promptly?
Promptly is a hosted cloud version of LLMStack. It is a fully managed, enterprise-grade service that allows you to build powerful AI applications without worrying about infrastructure, scalability, etc. It is free to get started with Promptly at trypromptly.com without requiring local installation.
What data sources can I import into LLMStack?
LLMStack supports a wide variety of data sources including Web URLs, Sitemaps, PDFs, Audio, PPTs, Google Drive, Notion imports, CSV, TXT, PDF, DOCX, PPTX files, YouTube, and direct uploads. The platform preprocesses and vectorizes your data automatically and stores it in the included vector database.
How do I access LLMStack apps via API?
LLMStack provides a HTTP API at POST /api/apps/<app_uuid>/run that allows you to programmatically interact with the platform. All API requests must be authenticated using a token found in the Settings page, passed in the Authorization header with 'Token:' prefix. The API supports both streaming and non-streaming modes.
What can I build with LLMStack?
You can build chatbots that answer questions and generate text, AI assistants that automate workflows and business processes, AI agents that help with day-to-day tasks, and AI applications that generate text, answer questions, summarize text, translate text, and classify text. You can also create multi-agent systems and connect AI chains to Slack or Discord.
Is LLMStack multi-tenant?
Yes, LLMStack is multi-tenant. You can create multiple organizations and add users to them. Users can only access the data and AI chains that belong to their organization. Datasources can also be shared with other users of your LLMStack installation within the context of an organization.
What are processors in LLMStack?
Processors are the basic building blocks in LLMStack that provide functionality to process input from users or previous processors in a chain, take some action, and optionally generate a response. LLMStack comes with built-in processors like OpenAI's ChatGPT, Image Generation, and Stability's Image Generation. You can also create your own processors.
How do I deploy LLMStack on a server?
When deploying LLMStack on a server, update allowed_hosts and csrf_trusted_origins in the ~/.llmstack/config file to include your server's hostname. You can add provider keys (OpenAI, Cohere, Stability) to the Settings page or add default keys for all users in the config file. The platform will download required Docker images on first start.