
Developers send a list of chat messages to a model and get a reply, for example to add a chat feature to an app.
Developers send feedback on a generated response, for example a thumbs down for an incomplete answer.
Data scientists create embedding vectors for input text with an embedding request.
AI agents look up exact request and response fields, such as the details of a chat generation response, before they write code.
The Salesforce Models REST API connects an app to large language models (LLMs) for chat, embeddings, text generation and feedback. Requests go through the Einstein Trust Layer. Salesforce publishes the OpenAPI 3.0 file for this API in the salesforce/einstein-platform repository on GitHub, under the Apache 2.0 license. The schemas in this entry include ChatGenerationsRequest, ChatMessage, GenerationRequest, EmbeddingRequest and FeedbackRequest. Other schemas carry trust scan and localization details. This entry covers the Models REST API, not the sObject data API. Developers can use this entry to check request and response shapes before they call the Models API from code or Apex.
ChatGenerationsRequest, ChatGenerationsResponse, ChatMessage, ChatMessageRequest, GenerationRequest, GenerationResponse, GenerationRepresentation, GenerationDetails, EmbeddingRequest, EmbeddingResponse, Embeddings, FeedbackRequest
Query the library's CoreModels project for request and response shapes before writing Models API calls. Check model names in Salesforce's current guide. Sample prompt: "Using the Salesforce Models REST API schema in the Schematica Library, which fields of a ChatGenerationsRequest hold the messages and the localization?" Read via MCP: get_project_summary with graphProjectId b37534c4c84446d8943819b0991ce5e0.
Publisher: Salesforce. License: Apache-2.0 (https://github.com/salesforce/einstein-platform/blob/main/LICENSE). Version: v1. Format: OpenAPI (listed here as JSON Schema until the library format list includes it). Described by the Schema Librarian on 2026-10-23; auto-cataloged.