Vellum LLM Platform API

aid: vellum:llm-platform
name: Vellum LLM Platform API
description: The Vellum REST API exposes prompts, workflows, evaluations, datasets, document indexes,
  deployments, and execution endpoints so developers can run versioned LLM pipelines from their own backends
  and capture logs and metrics for production monitoring.
humanURL: https://docs.vellum.ai
baseURL: https://api.vellum.ai
tags:
- Prompts
- Workflows
- Evaluations
- Datasets
- Documents
- Deployments
- Executions
- Monitoring
properties:
- type: Documentation
  url: https://docs.vellum.ai
- type: APIReference
  url: https://docs.vellum.ai/api-reference
- type: GettingStarted
  url: https://docs.vellum.ai/welcome/getting-started
- type: SignUp
  url: https://app.vellum.ai/signup
- type: SDK
  url: https://github.com/vellum-ai/vellum-client-python
- type: SDK
  url: https://github.com/vellum-ai/vellum-client-typescript
- type: GitHubOrganization
  url: https://github.com/vellum-ai
- type: Pricing
  url: https://www.vellum.ai/pricing
- type: Blog
  url: https://www.vellum.ai/blog
features:
- name: Prompt Engineering Workbench
  description: Versioned prompts, side-by-side model comparisons, and structured prompt variables.
- name: Workflows Builder
  description: Visual builder for multi-step LLM pipelines including branching, tools, and RAG nodes.
- name: Evaluation Suites
  description: Run dataset-driven evals with built-in and custom metrics for prompts and workflows.
- name: Dataset Management
  description: Store labeled test cases and production examples to drive evaluations and fine-tuning.
- name: Document Indexes / RAG
  description: Managed document ingestion, embeddings, and retrieval for grounded generation.
- name: Deployments and Versioning
  description: Promote prompts and workflows through environments with rollback and traffic splits.
- name: Production Monitoring
  description: Logs, traces, latency, cost, and quality metrics for every execution.
- name: Multi-Provider Routing
  description: Vendor-neutral access to OpenAI, Anthropic, Google, Mistral, and open models.
- name: SDKs for Python and TypeScript
  description: First-class SDKs for invoking prompts, workflows, and datasets from application code.
- name: Self-Hosted Option
  description: Deploy Vellum into the customer's own cloud for compliance-sensitive workloads.
useCases:
- name: Build Production LLM Apps
  description: Iterate on prompts and workflows, then deploy versioned endpoints into apps.
- name: Evaluate LLM Quality
  description: Use datasets and evals to measure regressions across models and prompt variants.
- name: Build Agents
  description: Compose tools, retrieval, and conditional logic via the visual workflow builder.
- name: RAG Pipelines
  description: Ingest documents, index them, and query through Vellum's managed retrieval layer.
- name: Observe and Debug
  description: Trace production runs to debug failures and improve quality over time.
integrations:
- name: OpenAI
- name: Anthropic
- name: Google
- name: Mistral
- name: Cohere
- name: AWS Bedrock
- name: Azure OpenAI
- name: Pinecone
- name: Snowflake
- name: LangChain
- name: LlamaIndex
authentication:
- type: API Key
  description: Workspace API keys passed via the `X_API_KEY` header authenticate REST and SDK calls.