deepset AI Platform (deepset Cloud)
SaaS platform for building and operating RAG apps with pipelines, datasets, evals and governance, from free prototyping to enterprise deployments.
Overview
deepset AI Platform is a managed environment for creating, testing and scaling retrieval augmented generation applications. Teams organize work in workspaces, upload files, configure retrievers and generators, and ship pipelines to cloud runtimes with observability. Built in templates accelerate chat over documents and question answering, while a Playground lets product and data teams try prompts, filters and chunking strategies without writing boilerplate.
The platform tracks datasets, versions and lineage so experiments are reproducible and audit friendly. Integrated evaluations compare prompts, models and retrievers using automatic and rubric based metrics, and traces reveal where quality or latency degrades. Engineers integrate via SDKs and APIs, or deploy the same patterns in private cloud for compliance.
Pricing starts with a free Studio tier that includes a single user, limited pipeline hours and file caps, then expands to paid business and enterprise options with SSO, role based access control, private networking and higher quotas. The goal is to turn RAG from ad hoc scripts into a governed product lifecycle that multiple functions can collaborate on while keeping data within approved boundaries.
Key features
- Workspace model with roles and quotas: organize users data and pipelines across teams without ad hoc access
- Pipeline templates for RAG and chat: ship question answering with retrievers chunking and prompt patterns
- Playground for prompts and filters: iterate chunk size top k and citations to balance cost and quality
- Datasets and lineage tracking: version corpora ground truth and experiments for reproducibility
- Eval runs and reports: compare models prompts retrievers with automatic and rubric based scoring
- Tracing for latency and failures: inspect spans retriever hits and model calls to fix bottlenecks
- SDK and API integrations: connect apps CI and vector stores without bespoke glue code
- Private and hybrid deployments: align with security SSO VPC peering and audit requirements
Best for
- Spin up RAG chat over policy wikis and manuals
- Benchmark retrievers and prompts before rollout
- Operationalize document refresh and reindexing
- Instrument latency budgets and timeouts in prod
- Create gold datasets and regression evals
- Provide governed access for analysts and PMs
- Prove impact to stakeholders with eval reports
- Deploy private cloud for regulated workloads
Capabilities
Pipelines and templates
Start from RAG chat and QA blueprints, connect corpora, choose retrievers and prompts, then deploy with guardrails.
Evals and datasets
Create datasets and rubric based reports to compare prompts, models and retrievers before and after release.
Tracing and metrics
Trace spans and model calls, track latency and errors, and spot drift or coverage gaps early.
Access and deployments
Use SSO roles and private networking to meet policy while enabling collaboration across teams.
Frequently Asked Questions
How does pricing start for deepset AI Platform?
Public pages list a Free Studio tier for one user and limited pipeline hours, paid business and enterprise plans are available by quote with SSO and higher quotas.
Does it support evaluations out of the box?
Yes, you can set up automatic and rubric based evaluations to compare prompts models and retrieval settings with shareable reports.
Can I deploy privately for compliance?
Yes, enterprise options include private or hybrid deployments with role based access control and VPC connectivity.
Do I need to code to prototype RAG?
No, templates and a Playground let non specialists test chunking filters and prompts, SDKs and APIs support full integration when ready.
What integrations are available?
The platform connects with popular frameworks and vector stores and exposes SDKs and APIs for app and CI workflows.
