Support matrix

Snap Pack and Vendor

AgentCreator leverage Snaps to connect to various vector databases and LLMs, in addition to providing utilities to create your LLM apps. The following matrix lists the supported functionality by Snap Pack.


Component	Snap Pack	Vendor	Capabilities
LLMs	Amazon Bedrock LLM Snap Pack	Amazon Bedrock	Integrates with diverse Amazon Foundation Models (FMs) Supports prompt generation, chat completions, and embeddings Runs agentic workflows with the Amazon Bedrock Converse API Agent Snap Supports Managed Knowledge Base retrieval, agentic retrieval, and document ingestion for RAG Supports multimodal input
	Azure OpenAI LLM Snap Pack	Azure OpenAI	Integrates OpenAI's Large Language Models (LLMs) with the Azure OpenAI Service Supports prompt generation, chat completions, and embeddings Runs agentic workflows with the Azure AI Foundry Chat Completions Agent Snap Supports multimodal input
	OpenAI LLM Snap Pack	OpenAI	Integrates with OpenAI's Large Language Models (LLMs) within the SnapLogic Platform Supports prompt generation, chat completions, and embeddings Runs agentic workflows with the OpenAI Chat Completions Agent and OpenAI Responses API Agent Snaps Manages vector stores and vector store files Supports multimodal input
	Google GenAI LLM Snap Pack	Google GenAI LLM	Integrates with Google Gemini's Large Language Models (LLMs) within the SnapLogic Platform Supports prompt generation, text generation, and embeddings Runs agentic workflows with the Google Gemini API Agent and Google Vertex AI Gemini API Agent Snaps Supports multimodal input
Vector Databases	Pinecone Snap Pack	Pinecone	Queries, inserts, and updates records in the Pinecone database Processes the vector from the input document
	OpenSearch Snap Pack	OpenSearch	Performs similarity searches Supports Approximate Nearest Neighbor (ANN) queries Supports Range queries
	MongoDB Atlas Vector Search	MongoDB	Performs similarity searches Supports Approximate Nearest Neighbor (ANN) queries Supports Range queries
	AlloyDB Vector Search	AlloyDB	Performs similarity searches Supports Approximate Nearest Neighbor (ANN) queries Supports Range queries
	PostgreSQL Vector Search	PostgreSQL	Performs similarity searches Supports Approximate Nearest Neighbor (ANN) queries Supports Range queries
	Snowflake - Vector Search	Snowflake	Performs similarity searches Supports Approximate Nearest Neighbor (ANN) queries Supports Range queries
Utilities	LLM Utilities Snap Pack	SnapLogic	Enables custom knowledge-context chunks with defined properties. Facilitates embedding and stores result chunks in the vector database to optimize data organization in subsequent queries Generates tool definitions for multiple tool pipelines from a single Snap with the Multi Pipeline Function Generator
	HTML Parser		Parses HTML into text
	Markdown Parser		Parses Markdown into text
	HTML to Markdown Converter		Converts HTML into Markdown
	Adobe PDF Snap Pack		Extract: Extracts text, table, and figures from a PDF file. OCR: Adds a scannable layer on top of an image-based PDF file. Adobe Account: Authenticates with Adobe APIs using the specified Client ID and Client secret.
	Unstructured Snap Pack		Partition API: Extracts structured data from raw unstructured documents using the Unstructured API and transforms it into structured data. Unstructured API Account: Authenticates with Unstructured API using the specified credentials.
	Message Appender		Merges document streams into a single output stream
	PipeLoop		Enables looping on pipeline executions based on a stop condition or iteration limit, supporting agent execution loops.

Capabilities and Snaps

The following sections describe the AgentCreator capabilities and the Snaps that provide them.

Agentic workflows

Agent Snaps run an agent execution loop: they accept an initial request, a list of tools, and parameters, then call the LLM, invoke the returned tool pipelines, and repeat until the model has no more tools to call. Each LLM Snap Pack provides an Agent Snap.

Text embedding

Text embedding provides a vector (numeric representation) of the input data based on the embedding model and method. This supports a wide variety of output use cases, such as semantic search.

Text generation

Text generation takes an input prompt and generates text-based output — for example, summarizing input data or producing a RAG completion that draws on embedded vector content added to the input context window. The prompt generator Snaps support an advanced plain text mapping system using mustache templates, and the generated response can be fed into any text generation Snap through the messages format or a single user prompt.

Structured-output text generation

For integration use cases, structured output lets generated data flow through downstream systems. Most text generation Snaps support JSON mode, which returns output in a structured, accessible form rather than requiring you to parse it from a conversational response.

Multimodal input text generation

You can send non-text input to models that support it, with some limitations. OpenAI and Azure OpenAI support certain image input types; SnapLogic can process PDFs into images and extract text directly; Amazon Bedrock models may support document or image input; and Google Gemini supports multimodal input with a file-size limit (larger inputs can be uploaded through Google storage).

Managed Knowledge Base RAG

The Amazon Bedrock Managed Knowledge Base (MKB) Snaps ingest documents into, and retrieve context from, an AWS Managed Knowledge Base for production retrieval-augmented generation.

Advanced document parsing

When you need to parse full documents — including formats not handled internally by SnapLogic, such as Word, PowerPoint, or Rich Text Format files — use the document parsing Snap Packs.