Platform > Integration Layer > AI Connectors
AI connectors.
Connect to eight leading AI + LLM providers as native workflow nodes - choose the right model for each task and chain them with any data source or automation.
CAPABILITY OVERVIEW
Bring any AI model into your workflows.
Rayven provides native connectors for leading AI and LLM platforms, each operating as a node in the visual workflow builder.
Every AI connector accepts dynamic inputs from upstream workflow nodes - passing documents, structured data or operational context to the model and routing its output to downstream logic, storage or user interfaces. AI becomes a functional step in any operational process, not a standalone tool requiring separate integration.
Connect the right model for each task: document extraction, anomaly detection, structured data generation, summarisation, or decision support.
Inbound connections include:
+ more.
KEY CAPABILITIES
What AI Connectors give you.
OpenAI
GPT-4 and GPT-4o family models as workflow nodes. Accepts static or dynamic prompts, reads files directly from FTP paths + outputs structured JSON, summaries or classification results.
Anthropic Claude
Ultra-long context windows, strong reasoning + safe outputs. Best suited for large document processing, regulated industry applications + workflows requiring careful, detailed analysis.
Google Gemini
Enterprise-grade multimodal model with strong performance on text, images + structured data. Preferred for GCP-integrated environments + workflows involving mixed data types.
Cohere Command R+
Optimised for retrieval-augmented generation (RAG) + enterprise search. Best for extracting structured information from large document sets, internal knowledge bases + unstructured data stores.
Mistral, Meta Llama + open-source
Open-weight models for private or on-premise AI deployments. Mistral 7B is cost-efficient for edge use cases. Meta Llama supports full platform control + custom model training.
Microsoft Copilot + IBM Granite
Copilot is GPT-backed and ideal for Microsoft 365 workflows. IBM Granite is built for enterprise governance + regulated industries requiring explainability, compliance + sensitive data handling.
HOW IT CONNECTS: EXPLAINER
Where AI Connectors fit in the Rayven Platform stack.
AI connector nodes sit in the Integration Layer as bidirectional connections to external AI models.
They operate within Rayven workflows alongside any other node type:
-
Upstream: data from IoT devices, files, APIs, databases or forms is passed to the AI model as context or input The AI model processes the input + returns structured output
-
Downstream: output routes to storage, dashboards, automation logic, external APIs or user interfaces
Multiple AI nodes can chain sequentially or in parallel within the same workflow - enabling agent-style orchestration without a separate AI infrastructure.
USE CASES
How AI Connectors get used.
Industrial AI document processing
Maintenance reports uploaded to an SFTP folder are passed by a Rayven workflow to a Claude node for structured data extraction. Results are written to a Primary Table + a work order is triggered if a critical defect is identified - fully automated from upload to action.
Real-time data enrichment for BI
Customer records ingested via CRM connector are passed to a Cohere Command R+ node for classification and tagging. Enriched records are written to a dashboard table - improving BI accuracy without manual data preparation.
Partner-built AI-powered client portal
An MSP embeds a Rayven conversational analytics interface into a white-label client portal. The interface uses an LLM node to answer questions about client operational data in natural language - delivered as the partner's own AI product.
AI Connectors FAQs:
Which AI/LLM providers does Rayven support natively?
OpenAI, Anthropic Claude, Google Gemini, Cohere Command R+, Mistral/7B, Meta Llama, Microsoft Copilot + IBM Granite. A custom API node connects to any additional LLM endpoint not listed. See all Integration Layer connector options.
Can I use multiple AI models in the same workflow?
Yes. Multiple AI nodes can run sequentially or in parallel within the same workflow - for example, using one model to extract data and another to classify or summarise the output. Explore the Execution Layer for workflow design options.
How do I pass data from other systems to an AI model?
Any upstream workflow node output can be passed to an AI connector as input. File contents, database records, API responses + sensor readings can all serve as AI model context within the same workflow chain. See all Integration Layer data sources.
What output formats do AI connectors return?
AI connectors return structured text or JSON by default. Downstream JavaScript or Advanced Function nodes can parse, transform + route the output to storage, dashboards or external APIs. Learn how the Data Layer stores AI outputs.
Can AI connectors read files directly?
Yes. AI connector nodes can read files from an FTP path directly - passing document content to the model without a separate extraction step. Useful for PDF processing, report analysis + document classification. See File Uploads for ingestion options.
Which model should I use for which task?
Claude for long document processing + regulated industries. Cohere Command R+ for RAG + document extraction. OpenAI GPT-4 for general reasoning + structured output. Gemini for multimodal + GCP. Mistral/Llama for private or edge deployments. See all AI Connector options.
Can AI outputs trigger automated actions in the same workflow?
Yes. AI node outputs feed directly into a Conditional Filter, Rule Builder or output node. If an AI model flags an anomaly or classifies a record as high-priority, the same workflow can trigger an alert, update a database or fire an API call. See Control + Automation.
Is there a custom API option for AI providers not in the list?
Yes. The API node connects to any LLM endpoint that accepts HTTP POST requests - covering any provider not listed as a native connector, including self-hosted models. See all Integration Layer connection options.
Are AI connector calls logged for audit purposes?
All workflow executions are stored in Cassandra with UID + timestamp indexing. AI connector inputs + outputs are part of the workflow payload history, accessible via the Inspect Data tab. See Audit Trails + Logs.
Can AI connectors work with on-premise or private deployments?
Yes. Models supporting private deployment - Meta Llama, Mistral + self-hosted endpoints - connect via the custom API node, enabling AI processing within private cloud or on-premise Rayven deployments. See Deployment + Architecture options.
Also in the Integration Layer:
Pre-Built IT/OT Connectors
150+ ready-to-use connectors across CRM, ERP, BI, AI/LLM services + industrial systems.
File Uploads
Ingest files directly via FTP, FTPS, SFTP + AWS S3, or through manual uploads within the platform.
Forms + Manual Input
Capture structured data from human input via configurable form widgets, feeding directly into workflows.
IoT Devices + Protocols
Native support for MQTT, Modbus, LoRaWAN, SNMP + Raw UDP for edge and industrial device connectivity.
API Endpoints
Expose processed platform data to external systems via authenticated GET endpoints + inbound POST hooks.
Custom Integrations
Build bespoke two-way integrations via HTTP request nodes, REST API adapters + custom webhook endpoints.
Streaming Data Connectors
High-frequency ingestion via AWS Kinesis, MQTT + AMQP into Cassandra time-series storage.
Join the Shift
Discover the easy way to do something new.
Book a free 30 minute assessment with our team and we'll scope your project, needs + what a solution might look like.