Inferable
Added by nadeesha-cabralInferable helps developers build LLM-based agentic automations faster with a delightful developer experience.
Description
Key features include:
- Distributed tool calling architecture that separates agent runtime from function execution
- Support for multiple programming languages including NodeJS, Go, and .NET
- On-premise execution ensuring sensitive data stays within your infrastructure
- Private networking capabilities with no inbound connections required
- Built-in observability and monitoring tools
- Managed LLM operations with intelligent model routing
- Function registration and execution within your own environment
- Support for existing REST and GraphQL APIs
Pros and cons of Inferable
Pros of Inferable
- Developer-first AI automation platform that integrates with existing codebase.
- Supports multiple languages including Node.js, Golang, and C#, with plans to support more.
- Offers distributed orchestration for reliable AI workflow execution.
- Open-source and self-hostable for full control over data and compute.
- Provides extensive observability into workflows and function calls.
- Adapts to existing infrastructure and allows bringing in custom models.
Cons of Inferable
- Limited availability of supported languages with some in beta or coming soon.
- Primarily uses specific AI models with limited flexibility in choosing providers.
- Integration with custom models in the managed runtime is limited to enterprise customers.
- Potential complexity in managing AI workflows and ensuring data privacy.
FAQ about Inferable
Where does my data reside when using Inferable?
Your data remains within your infrastructure at all times. Inferable employs an on-premise execution model where all compute and data processing occurs within your environment. The Inferable SDK only sends function outputs to the control plane, not the execution context or sensitive data like your environment variables and secrets.
Does Inferable require incoming connections to my infrastructure?
No. Inferable uses a long-polling approach where all connections are initiated from your infrastructure to the Inferable control plane. There's no need for inbound connections or open ports, significantly reducing your security attack surface.
What happens to my data when it's processed by AI models?
When using Inferable's AI capabilities: - Data is never stored by the AI model, beyond the context of your specific run - The AI model only has access to what your functions explicitly return, unless it's masked(). - The control plane never has access to data being processed by your functions - All model providers guarantee zero data retention - We exclusively work with model providers that can provide zero data retention guarantees
How long do you retain my data?
All data is sharded by the cluster. Data sent to the model is never retained by the model provider. We use AWS Bedrock for our model providers, which guarantees zero data retention. Data sent to the control plane will be retained until you delete the cluster or delete the run.
Share this tool
Tags: