Plugins Overview

Plugins are the building blocks of Tensorify workflows. Each plugin encapsulates a specific capability - from AI model calls to database operations.

Plugin Categories

AI & LLM

Connect to leading AI models and services.

| Plugin | Description | | ------------- | ---------------------------------- | | OpenAI | GPT-4, GPT-3.5, embeddings, DALL-E | | Anthropic | Claude 3 models | | Pinecone | Vector database for RAG |

Data Processing

Transform, filter, and manipulate data.

| Plugin | Description | | ------------------ | ---------------------------------------- | | JSON Transform | Python expressions for data manipulation | | CSV Dataset | Load and process CSV files | | PDF Reader | Extract text from PDF documents | | Web Scraper | Fetch and parse web pages | | Text Splitter | Chunk text for RAG pipelines |

Machine Learning

Build and train neural networks.

| Plugin | Description | | -------------------------- | ------------------------- | | MNIST / CIFAR | Built-in datasets | | DataLoader | Batch data for training | | Linear / Conv2D / ReLU | Neural network layers | | Training Loop | Automated training epochs | | Save Model | Export trained models |

Databases

Connect to popular databases.

| Plugin | Description | | -------------- | ---------------------------- | | PostgreSQL | SQL queries and operations | | MongoDB | Document database operations |

Integrations

Connect with external services.

| Plugin | Description | | ------------------- | ------------------------- | | Slack | Send messages to channels | | SendGrid | Send emails | | HTTP Request | Call any REST API | | Webhook Trigger | Receive HTTP requests |

Logic & Control Flow

Add conditional logic and loops.

| Plugin | Description | | --------------- | -------------------- | | Conditional | If/else branching | | Loop | Iterate over lists | | Branch | Multi-path workflows |

Using Plugins

Adding to Canvas

1. Open the plugin sidebar (left panel)
2. Search or browse for your plugin
3. Drag it onto the canvas
4. Click to configure settings

Configuring Plugins

Each plugin has unique settings. Common configurations include:

• API Keys - Credentials (stored securely)
• Model Selection - Choose AI models
• Input Mapping - How to use incoming data
• Output Format - How to structure results

Connecting Plugins

Plugins communicate through handles:

• Input handles (left) - Receive data from upstream
• Output handles (right) - Send data downstream

Drag from output to input to connect nodes.

Plugin Requirements

Each plugin specifies its Python dependencies. When you export:

• Dependencies are aggregated into requirements.txt
• Secrets are referenced as environment variables
• GPU requirements are detected automatically

Creating Custom Plugins

---

Explore specific categories:

• AI Plugins
• Data Plugins
• Integration Plugins