The Sage Glossary

Dec 31, 2024

The Sage Glossary

Agents Everywhere: Exploring the Power of Sage

Sage isn’t just a platform; it’s an ecosystem — a marketplace, a hub for innovation, and a powerful tool for creators. We often emphasize its flexibility: the ability to build what you need, when you need it. While this concept is exciting, it can feel abstract. That’s why, in this article, we’re diving into the specific agents and utilities currently built into Sage. By the end, you’ll have a clear understanding of the tools at your disposal and the creative possibilities they unlock.


A World of Possibilities

Sage offers tools for everyone — from simple, user-friendly utilities to highly specialized agents designed for complex workflows. Whether you’re extracting data from websites, generating images and videos, integrating with platforms like Telegram, Slack, or X, or even deploying automatic coders, Sage provides a vast range of agents to help you innovate.

Some tools are straightforward and ideal for quick use, like web search crawlers or social media integrations. Others are advanced, catering to developers who want to design intricate workflows or build agent swarms for specific purposes. This flexibility ensures Sage can meet your needs, whether you’re a casual user or a technical expert.


Understanding the Frameworks

Sage’s tools leverage two primary frameworks: LangChain and LlamaIndex.

  • LangChain: A framework for building complex applications, workflows, and multi-functional agents. It’s ideal for creating sophisticated systems tailored to your needs.

  • LlamaIndex: A tool optimized for retrieving information quickly, making it perfect for search-oriented tasks and data retrieval.

These frameworks complement each other, allowing you to mix and match depending on your goals.


Prebuilt or Custom: Your Choice

One of Sage’s greatest strengths is its flexibility. You can:

  • Use prebuilt agents created by others, either free or paid.

  • Build custom workflows and agent swarms tailored to your needs.

  • Skip the technical setup entirely by interacting with agents via natural language in interfaces like Telegram.

Whether you’re a developer or a non-technical user, Sage adapts to your preferred level of engagement.


The Agent List

The following is a comprehensive list of agents and utilities available in Sage at the time of writing. While some may be highly specialized and technical, others are straightforward and easy to use. Whether you’re looking to build complex systems or simply explore what’s possible, this list is your gateway to the innovation Sage enables.

Let’s dive into the details and uncover the tools that make Sage a powerhouse for creation.


Specific Agents in LangChain

1. Airtable Agent

Purpose: Executes queries directly on Airtable databases to retrieve or manipulate data.

Key Features: Integrates with Airtable APIs to enable efficient querying of structured data stored in Airtable sheets. Often used for automating workflows and real-time reporting.

2. AutoGPT

Purpose: An autonomous agent capable of completing complex tasks without explicit step-by-step instructions.

Key Features: Implements chain-of-thought reasoning to iteratively break down tasks, execute subtasks, and adjust its approach dynamically. Ideal for scenarios where self-guidance and multi-step task completion are critical.

3. BabyAGI

Purpose: A task-driven autonomous agent designed to prioritize and dynamically generate new tasks based on the overarching goal or objective.

Key Features: Maintains an evolving task list, adjusts priorities in real-time, and optimizes workflows by aligning subtasks with the main objective. Frequently used in goal-oriented automation systems.

4. CSV Agent

Purpose: Processes CSV data to answer questions or perform specific analyses.

Key Features: Reads structured data from CSV files, interprets relationships between columns, and delivers insights or results in a user-friendly format. Useful for data analysis, reporting, or generating summaries.

5. Conversational Agent

Purpose: Powers interactive chat models through dynamic and contextually aware responses.

Key Features: Relies on predefined or fine-tuned chat prompts to simulate natural conversations, making it suitable for customer support, personal assistants, or interactive systems.

6. OpenAI Assistant Agent

Purpose: Uses the OpenAI Assistant API to intelligently select the appropriate tools or data for a given task.

Key Features: Dynamically identifies which tool or function to invoke based on user input and task requirements. Acts as an intermediary between users and complex multi-tool systems.

7. ReAct Agent for Chat Models

Purpose: Leverages ReAct (Reasoning and Acting) logic to handle conversational workflows.

Key Features: Optimized for chat models, enabling the agent to analyze user intent, take necessary actions, and generate accurate responses in dialogue-driven tasks.

8. ReAct Agent for LLMs

Purpose: Applies ReAct logic to determine and execute the best action for large language model tasks.

Key Features: Specifically tailored for LLMs, ensuring effective reasoning and action selection in non-conversational tasks such as data analysis or document summarization.

9. Tool Agent

Purpose: Implements function-calling capabilities to invoke external tools or APIs as needed.

Key Features: Identifies and selects the appropriate tools based on the task, retrieves data, and processes results efficiently. Ideal for multi-functional systems requiring tool integration.

10. XML Agent

Purpose: Designed for tasks involving XML data or requiring structured output in XML format.

Key Features: Excels in reasoning, parsing, and writing XML data. Commonly used in systems that handle configuration files, data exchange formats, or structured document generation.


Cache Types in LangChain

1. InMemory Cache

Purpose: Temporarily stores LLM responses in memory during runtime.

Key Features: Ideal for scenarios where low-latency response time is critical. However, the cache is ephemeral and cleared each time the program restarts, making it unsuitable for persistent storage.

2. InMemory Embedding Cache

Purpose: Caches embeddings generated by LLMs directly in memory.

Key Features: Eliminates the need to recompute embeddings for repeated queries, improving performance for tasks such as similarity search. Limited to in-session use as it is not persistent.

3. Memento Cache

Purpose: Utilizes Memento, a distributed serverless cache, to store LLM responses.

Key Features: Provides scalable and persistent caching for distributed systems, enabling efficient response sharing across multiple instances or processes.

4. Redis Cache

Purpose: Stores LLM responses in a Redis database for shared access across systems.

Key Features: Enables centralized caching for multi-processor or multi-server environments, improving efficiency and reducing redundant API calls. Ideal for use in production systems requiring persistent and scalable caching.

5. Redis Embeddings Cache

Purpose: Saves generated embeddings in Redis for persistent storage and quick retrieval.

Key Features: Speeds up tasks like document retrieval or vector similarity by reusing precomputed embeddings. Works well in distributed setups where multiple applications or services need access to embeddings.

6. Upstash Redis Cache

Purpose: A serverless Redis-based cache for storing LLM responses.

Key Features: Combines Redis’s speed with the scalability of serverless infrastructure. Suitable for use in cloud-native and event-driven architectures requiring efficient caching.


Chains in LangChain

1. GET API Chain

Purpose: Executes queries against APIs using HTTP GET requests.

Key Features: Retrieves data from external data sources or APIs, making it ideal for integrating with real-time or static datasets from the web.

2. OpenAPI Chain

Purpose: Dynamically selects and calls APIs defined by an OpenAPI specification.

Key Features: Automates API integration by parsing the OpenAPI schema and executing appropriate API calls, reducing manual configuration efforts.

3. POST API Chain

Purpose: Executes HTTP POST requests to send data to APIs.

Key Features: Facilitates data submission to external APIs, enabling interactions that require posting data or triggering specific workflows.

4. Conversation Chain

Purpose: Implements chat-specific chains optimized for conversational models.

Key Features: Often incorporates memory to maintain contextual awareness throughout the dialogue. Useful for building conversational agents or chatbots.

5. Conversational Retrieval QA Chain

Purpose: Provides conversational question-answering capabilities while integrating with retrieval systems.

Key Features: Incorporates chat history for contextually aware responses, making it effective for multi-turn interactions that involve document retrieval.

6. LLM Chain

Purpose: Executes queries directly against large language models (LLMs).

Key Features: Serves as the foundational chain for integrating LLMs, supporting text generation, question answering, and other tasks.

7. Multi Prompt Chain

Purpose: Dynamically selects the most suitable prompt from a set of predefined templates.

Key Features: Automates prompt selection based on the context or input, ensuring optimal performance for varied tasks.

8. Multi Retrieval QA Chain

Purpose: Automatically chooses the best vector store from multiple retrievers to answer questions.

Key Features: Enhances flexibility in document retrieval by intelligently selecting the appropriate resource for the query.

9. Retrieval QA Chain

Purpose: Answers questions based on documents retrieved from a knowledge base.

Key Features: Combines document retrieval with LLM-powered question answering to deliver accurate and contextually relevant responses.

10. SQL DB Chain

Purpose: Executes SQL queries to retrieve or manipulate data from relational databases.

Key Features: Translates natural language inputs into SQL commands, enabling users to interact with structured databases without SQL expertise.

11. Vectara QA Chain

Purpose: Performs question-answering tasks using the Vectara platform.

Key Features: Integrates Vectara’s AI-powered search capabilities to deliver accurate responses based on indexed content.

12. VectorDB QA Chain

Purpose: Supports question-answering workflows for vector databases.

Key Features: Combines vector-based similarity search with LLMs to provide precise answers, making it suitable for tasks like semantic search and knowledge retrieval.


Chat Models in LangChain

1. AWS ChatBedrock

Purpose: Leverages Amazon Web Services’ Bedrock large language model (LLM) through a chat-specific endpoint.

Key Features: Provides a scalable, secure, and reliable solution for deploying conversational AI using AWS infrastructure.

2. Azure ChatOpenAI

Purpose: Uses Microsoft Azure’s integration of OpenAI’s LLMs via a chat endpoint.

Key Features: Optimized for enterprises requiring Azure-specific security, compliance, and scalability for conversational applications.

3. ChatAnthropic

Purpose: Integrates Anthropic’s LLMs via a chat endpoint for natural language processing.

Key Features: Offers advanced conversational AI capabilities with a focus on safe and aligned language model outputs.

4. ChatBaiduWenxin

Purpose: Utilizes Baidu’s Wenxin chat LLM through a dedicated chat endpoint.

Key Features: Ideal for applications requiring Chinese language proficiency and localized AI solutions.

5. ChatCohere

Purpose: Employs Cohere’s conversational LLMs via a chat endpoint.

Key Features: Known for efficient language models with strong performance in semantic understanding and conversational tasks.

6. ChatFireworks

Purpose: Integrates Fireworks’ chat-based LLMs using a chat endpoint.

Key Features: Designed for highly interactive and natural conversations in various business scenarios.

7. ChatGoogleGenerativeAI

Purpose: Uses Google’s Gemini LLM for conversational AI through a chat endpoint.

Key Features: Advanced AI with multimodal capabilities, optimized for both text and visual data interactions.

8. ChatGooglePaLM

Purpose: Leverages Google’s PaLM model via the MakerSuite chat endpoint.

Key Features: Supports advanced generative text and conversational capabilities, making it suitable for creative and enterprise use cases.

9. ChatGoogleVertexAI

Purpose: Integrates VertexAI’s LLMs for conversational AI applications.

Key Features: Part of Google Cloud, offering enterprise-grade AI solutions with a focus on customization and deployment scalability.

10. ChatHuggingFace

Purpose: Uses Hugging Face’s LLMs through a chat-specific endpoint.

Key Features: Provides access to a wide range of pre-trained and fine-tuned models for natural language understanding and generation tasks.

11. ChatLocalAI

Purpose: Operates local language models like Llama.cpp and GPT4All using LocalAI.

Key Features: Enables private, on-premise deployment for conversational AI without relying on external APIs.

12. ChatMistralAI

Purpose: Leverages Mistral’s LLMs via a chat-specific endpoint.

Key Features: Focuses on high-performance, open-access language models for interactive applications.

13. ChatOllama

Purpose: Uses open-source LLMs through Ollama’s chat endpoints.

Key Features: Ideal for developers seeking customizable and transparent conversational AI solutions.

14. ChatOpenAI

Purpose: Integrates OpenAI’s chat-focused LLMs via dedicated endpoints.

Key Features: Supports a variety of conversational tasks, including chatbots, Q&A systems, and customer service automation.

15. ChatOpenAICustom

Purpose: Employs fine-tuned versions of OpenAI models via APIs compatible with chat endpoints.

Key Features: Customized for domain-specific applications or tailored use cases.

16. ChatTogetherAI

Purpose: Uses TogetherAI’s LLMs for conversational applications.

Key Features: Focuses on collaborative and efficient AI capabilities across a variety of conversational tasks.

17. GroqChat

Purpose: Utilizes Groq’s APIs in conjunction with the LPU inference engine to power conversational AI.

Key Features: Offers high-performance inference for low-latency conversational tasks.


Document Loaders in LangChain

1. API Loader

Purpose: Retrieves data from external APIs.

Key Features: Supports dynamic, real-time data extraction for integration into downstream tasks.

2. Airtable Loader

Purpose: Extracts data directly from Airtable tables.

Key Features: Useful for automating workflows or incorporating structured Airtable data into applications.

3. Apify Website Content Crawler

Purpose: Loads data scraped from websites using Apify’s crawler.

Key Features: Efficiently captures web content for analysis or integration with AI models.

4. Cheerio Web Scraper

Purpose: Extracts data from webpages.

Key Features: Lightweight and fast, suitable for web scraping tasks with basic HTML parsing.

5. Confluence Loader

Purpose: Loads data from Atlassian Confluence documents.

Key Features: Facilitates integration with organizational knowledge bases.

6. CSV Loader

Purpose: Reads and processes data from CSV files.

Key Features: Ideal for importing tabular data for analysis or downstream tasks.

7. Custom Document Loader

Purpose: Allows users to define custom methods for loading documents.

Key Features: Flexible and adaptable to unique data formats or storage locations.

8. Document Store Loader

Purpose: Retrieves data from pre-configured document storage systems.

Key Features: Supports integration with enterprise-grade document management systems.

9. DOCX File Loader

Purpose: Loads content from Microsoft Word documents.

Key Features: Extracts structured or unstructured text for further processing.

10. Figma Loader

Purpose: Imports data from Figma files.

Key Features: Useful for integrating design artifacts into AI workflows.

11. File Loader

Purpose: Generic loader that supports multiple file types (e.g., TXT, JSON, CSV, PDF).

Key Features: Simplifies handling heterogeneous data formats in a unified way.

12. FireCrawl

Purpose: Loads data from specified URLs.

Key Features: Automates crawling and retrieving web content for integration into workflows.

13. Folder with Files Loader

Purpose: Processes multiple files within a directory.

Key Features: Useful for bulk document ingestion from file systems.

14. GitBook Loader

Purpose: Extracts content from GitBook repositories.

Key Features: Facilitates knowledge management and integration with documentation platforms.

15. GitHub Loader

Purpose: Loads data from GitHub repositories.

Key Features: Useful for code analysis, documentation retrieval, or knowledge extraction.

16. JSON File Loader

Purpose: Processes data stored in JSON files.

Key Features: Efficiently handles structured data for analysis or AI workflows.

17. JSON Lines File Loader

Purpose: Reads data from JSON lines files.

Key Features: Optimized for handling large datasets stored in a line-delimited JSON format.

18. Notion Loader

Purpose: Retrieves data from Notion databases, folders, or pages.

Key Features: Integrates Notion’s structured and unstructured data into workflows.

19. PDF Loader

Purpose: Extracts text from PDF documents.

Key Features: Supports parsing of structured and unstructured content, including scanned documents.

20. Plain Text Loader

Purpose: Processes simple text files.

Key Features: Suitable for basic text processing tasks or data ingestion.

21. Playwright Web Scraper

Purpose: Uses Playwright to scrape web content.

Key Features: Handles dynamic websites with advanced JavaScript rendering.

22. Puppeteer Web Scraper

Purpose: Uses Puppeteer for web scraping tasks.

Key Features: Ideal for dynamic websites requiring headless browser rendering.

23. S3 Directory Loader

Purpose: Loads data from AWS S3 buckets.

Key Features: Seamless integration with cloud storage for scalable document ingestion.

24. SearchAPI Loader

Purpose: Retrieves data from real-time search engine results.

Key Features: Enables live data collection for dynamic or time-sensitive applications.

25. SerpAPI Loader

Purpose: Processes data from web search results using SerpAPI.

Key Features: Designed for high-precision search engine data extraction.

26. Spider Document Loaders

Purpose: Generic web scraping tool for document retrieval.

Key Features: Adapts to a variety of web formats for scalable scraping.

27. Unstructured File Loader

Purpose: Handles unstructured data from file paths.

Key Features: Parses and processes content regardless of data organization.

28. VectorStore to Document Loader

Purpose: Searches documents in vector stores and returns results with relevance scores.

Key Features: Combines vector-based retrieval with document extraction for efficient QA or search workflows.


Embeddings and LLMs in LangChain

These components serve as tools to generate embeddings for given text or act as wrappers for large language models (LLMs), enabling seamless integration with LangChain workflows.

1. AWS Bedrock

Purpose: Integrates Amazon’s Bedrock platform to generate embeddings or interact with large-scale language models.

Key Features: Scalable, secure, and reliable, designed for enterprise-grade AI applications.

2. Azure OpenAI

Purpose: Leverages Microsoft Azure’s integration of OpenAI models to provide embeddings and LLM capabilities.

Key Features: Ensures enterprise-level security, compliance, and accessibility within Azure’s ecosystem.

3. Cohere

Purpose: Generates embeddings and interacts with Cohere’s language models for natural language processing tasks.

Key Features: Known for semantic search, text classification, and other language-heavy applications.

4. Fireworks

Purpose: Utilizes Fireworks’ LLMs to generate embeddings or handle advanced NLP tasks.

Key Features: Offers high-performing, language-model-driven solutions for text-based workflows.

5. Google GenerativeAI

Purpose: Uses Google’s cutting-edge Gemini models for embeddings and generative tasks.

Key Features: Advanced multimodal capabilities for text, images, and other data types.

6. Google PaLM

Purpose: Employs Google’s PaLM models to generate embeddings and perform text-based reasoning tasks.

Key Features: Highly efficient, capable of handling complex queries and large datasets.

7. Google VertexAI

Purpose: Provides access to Google Cloud’s VertexAI LLMs for embeddings and language model integration.

Key Features: Scalable, customizable, and tailored for enterprise applications.

8. Hugging Face

Purpose: Integrates Hugging Face models to generate embeddings or run language model operations.

Key Features: Access to a vast library of pre-trained and fine-tuned models for various NLP tasks.

9. LocalAI

Purpose: Uses local LLMs, such as Llama.cpp and GPT4All, for embeddings and other tasks.

Key Features: Allows for private, on-premise AI applications without dependency on external APIs.

10. MistralAI

Purpose: Utilizes Mistral’s high-performance models for embeddings and LLM workflows.

Key Features: Open-access, lightweight models optimized for speed and efficiency.

11. Ollama

Purpose: Leverages Ollama’s open-source models for embedding generation or conversational AI tasks.

Key Features: Transparent, customizable, and suitable for developers needing open-source solutions.

12. OpenAI

Purpose: Accesses OpenAI’s suite of models, such as GPT-4, to generate embeddings or interact with LLMs.

Key Features: Highly versatile and widely adopted for advanced NLP and generative tasks.

13. Replicate

Purpose: Wraps Replicate’s models for embedding generation and language processing tasks.

Key Features: Focuses on ease of use and deployment for AI models.

14. TogetherAI

Purpose: Uses TogetherAI’s large language models for embeddings and broader NLP workflows.

Key Features: Supports collaborative and efficient AI-powered applications.

15. VoyageAI

Purpose: Employs VoyageAI’s language models for embedding tasks and other AI-driven applications.

Key Features: Designed for high-performance and enterprise-level implementations.


Memory in LangChain

Memory modules enable conversation-aware applications by storing and managing conversational context for agents and workflows.

1. Agent Memory

Purpose: Maintains state and context for agent flows.

Key Features: Ensures that conversations with agents are contextually aware and logically coherent.

2. Buffer Memory

Purpose: Stores chat messages in a database for retrieval during a session.

Key Features: Keeps track of complete conversation history to maintain context.

3. Buffer Window Memory

Purpose: Stores a fixed number of recent messages, with customizable memory sizes.

Key Features: Optimized for scenarios where only recent context is needed, reducing memory usage.

4. Conversation Summary Memory

Purpose: Summarizes conversations and stores them in memory for long-term use.

Key Features: Ideal for compressing large conversations into manageable summaries for downstream processing.

5. DynamoDB Chat Memory

Purpose: Stores conversation data in Amazon DynamoDB tables.

Key Features: Provides scalable, cloud-based memory storage integrated with AWS infrastructure.

6. MongoDB Atlas Chat Memory

Purpose: Stores conversational history in MongoDB Atlas.

Key Features: Enables efficient memory retrieval in MongoDB’s distributed, cloud-based environment.

7. Redis-Backed Chat Memory

Purpose: Summarizes conversations and stores them in Redis servers.

Key Features: Offers fast and scalable in-memory storage for real-time applications.

8. Upstash Redis-Backed Chat Memory

Purpose: Summarizes and stores conversation history in Upstash Redis servers.

Key Features: Serverless and highly scalable, designed for cloud-native applications.

9. Zep Memory

Purpose: Stores conversation summaries in Zep servers (available as cloud-based or open-source).

Key Features: Flexible and suitable for both proprietary and open-source workflows.


Moderation in LangChain

Moderation tools ensure that input and output content aligns with ethical guidelines, usage policies, or specific business requirements.

1. OpenAI Moderation

Purpose: Verifies that generated content complies with OpenAI’s usage policies.

Key Features: Useful for maintaining legal and ethical compliance in applications relying on OpenAI’s models.

2. Simple Prompt Moderation

Purpose: Checks input text against predefined deny lists to filter inappropriate or restricted content.

Key Features: Prevents undesirable content from being processed or sent to LLMs, ensuring safer interactions.


Output Parsers in LangChain

Output parsers define how an LLM’s responses are processed and formatted into usable structures.

1. CSV Output Parser

Purpose: Converts an LLM’s output into a list of values formatted as CSV data.

Use Case: Ideal for scenarios where structured tabular data is needed from natural language input.

2. Custom List Output Parser

Purpose: Parses an LLM’s output into a list of values using a user-defined custom format.

Use Case: Provides flexibility for specialized list formatting or domain-specific requirements.

3. Structured Output Parser

Purpose: Transforms LLM outputs into JSON structures for structured data representation.

Use Case: Useful for applications requiring clear and predictable JSON-formatted outputs.

4. Advanced Structured Output Parser

Purpose: Parses LLM outputs into predefined structures using Zod schemas for validation.

Use Case: Ensures that outputs conform to strict data models, enhancing reliability in workflows.


Prompts in LangChain

Prompt templates guide LLMs by providing structured, repeatable input formats for generating consistent outputs.

1. Chat Prompt Template

Purpose: Represents a schema tailored for creating dynamic prompts in chat-based applications.

Use Case: Ensures a consistent structure for conversational interactions in chatbot systems.

2. Few Shot Prompt Template

Purpose: Builds prompts with multiple examples to guide the LLM’s behavior for specific tasks.

Use Case: Enhances model accuracy for complex tasks by showcasing desired output examples.

3. Prompt Template

Purpose: A foundational schema for creating basic prompts that direct LLM responses.

Use Case: Ideal for straightforward tasks or applications requiring minimal context.


Record Managers in LangChain

Record managers track document writes and updates in vector databases, ensuring data persistence and traceability.

1. MySQL Record Manager

Purpose: Uses MySQL to log document interactions with vector databases.

Use Case: Suitable for applications needing robust, relational database tracking in scalable environments.

2. Postgres Record Manager

Purpose: Leverages PostgreSQL to maintain a record of document writes in vector databases.

Use Case: Preferred for systems requiring advanced querying and relational data management.

3. SQLite Record Manager

Purpose: Uses SQLite for lightweight tracking of document interactions with vector databases.

Use Case: Perfect for local, small-scale, or development-focused applications.


Retrievers in LangChain

Retrievers fetch relevant documents or data from various sources to assist LLMs in answering queries.

1. AWS Bedrock Knowledge Base Retriever

Purpose: Connects to AWS Bedrock’s API to retrieve relevant knowledge base data.

Use Case: Designed for applications utilizing AWS’s enterprise-grade AI infrastructure.

2. Cohere Rerank Retriever

Purpose: Orders documents based on semantic relevance to a query.

Use Case: Ideal for improving search result quality by prioritizing the most relevant items.

3. Custom Retriever

Purpose: Returns results based on user-defined rules or formats.

Use Case: Allows for specialized retrieval logic tailored to specific business needs.

4. Embeddings Filter Retriever

Purpose: Compresses document sets by removing unrelated content using embeddings.

Use Case: Useful for reducing noise in large datasets while maintaining query relevance.

5. HyDE Retriever

Purpose: Retrieves documents from a vector store using HyDE algorithms.

Use Case: Enhances retrieval accuracy in vector-based storage systems.

6. LLM Filter Retriever

Purpose: Filters returned documents by extracting only the parts relevant to the user’s query.

Use Case: Ensures precise answers by discarding irrelevant content.

7. Multi Query Retriever

Purpose: Generates multiple query variations to retrieve data from different perspectives.

Use Case: Useful for ambiguous queries or exploring multiple facets of a topic.

8. Prompt Retriever

Purpose: Stores and retrieves prompt templates by name and description for later use.

Use Case: Facilitates reuse and consistency in prompt engineering.

9. Reciprocal Rank Fusion Retriever

Purpose: Reranks search results by combining scores from multiple query generations.

Use Case: Improves ranking precision by integrating diverse query perspectives.

10. Similarity Score Threshold Retriever

Purpose: Filters results based on a minimum similarity score threshold.

Use Case: Ensures retrieved content is highly relevant to the input query.

11. Vector Store Retriever

Purpose: Uses vector stores to retrieve documents based on embeddings.

Use Case: Efficient for applications involving semantic search and large document collections.

12. Voyage AI Rerank Retriever

Purpose: Reranks documents from most to least semantically relevant using Voyage AI models.

Use Case: Improves document ranking precision for enhanced user experiences.


Text Splitters in LangChain

Text splitters process and edit large or unstructured text into smaller, digestible formats, making it easier for language models to analyze and work with them.

1. Character Text Splitter

Purpose: Splits text into smaller chunks based on character limits.

Use Case: Ideal for handling large bodies of text where dividing by word or sentence isn’t necessary.

2. Code Text Splitter

Purpose: Splits source code into manageable chunks.

Use Case: Useful for analyzing or processing code files without exceeding LLM token limits.

3. HTML to Markdown Text Splitter

Purpose: Converts HTML text into Markdown format and splits it into chunks.

Use Case: Ideal for transforming web content into a structured and readable Markdown format.

4. Markdown Text Splitter

Purpose: Splits Markdown-formatted text into smaller sections.

Use Case: Preserves Markdown structure while dividing text for easier processing.

5. Recursive Character Text Splitter

Purpose: Splits text by progressively larger structures (e.g., paragraphs, sentences, characters) if size constraints are exceeded.

Use Case: Ensures efficient splitting while preserving as much context as possible.

6. Token Text Splitter

Purpose: Splits text based on token counts instead of characters or words.

Use Case: Optimized for use with LLMs to prevent exceeding model-specific token limits.


Tools in LangChain

LangChain tools provide agents with specialized functionalities to enhance their abilities in tasks like search, file handling, API calls, and more.

1. BraveSearchAPI

Purpose: Enables real-time access to Brave search results.

Use Case: Alternative to Google Search for retrieving web data with privacy in mind.

2. Calculator

Purpose: Allows agents to perform basic arithmetic calculations.

Use Case: Simplifies numerical and mathematical problem-solving.

3. Chain Tool

Purpose: Enables agents to execute chains as part of their toolset.

Use Case: Useful for chaining multiple actions or processes into a single workflow.

4. Chatflow Tool

Purpose: Executes another chatflow as a tool within an existing chatflow.

Use Case: Helps in modularizing and combining conversational workflows.

5. Code Interpreter

Purpose: Executes Python code in a sandbox environment.

Use Case: Ideal for data analysis, visualizations, or dynamic programming within conversations.

6. Custom Tool

Purpose: Integrates user-defined tools into LangChain chatflows.

Use Case: Extends LangChain functionality to meet specific application requirements.

7. Exa Search

Purpose: Uses Exa Search API, a search engine designed specifically for LLMs.

Use Case: Tailored for semantic understanding and LLM-driven applications.

8. Google Custom Search

Purpose: Provides real-time access to Google search results via its API.

Use Case: Retrieves web data with a focus on specific search configurations.

9. Luma

Purpose: Allows agents to create videos using the Luma Dream Machine.

Use Case: Converts text inputs into video content for creative or visual projects.

10. OpenAPI Toolkit

Purpose: Loads and interacts with OpenAPI specifications.

Use Case: Enables dynamic API calls based on OpenAPI documentation.

11. Read File

Purpose: Allows agents to read files from local disk storage.

Use Case: Useful for extracting data or context from existing files.

12. Requests for GET and POST

Purpose: Executes HTTP GET and POST requests for interacting with web resources.

Use Case: Fetches or sends data to external web services in real-time.

13. Retriever Tool

Purpose: Enables agents to use a retriever as a tool for fetching relevant data.

Use Case: Retrieves specific documents or data relevant to a query.

14. SearchAPI

Purpose: Provides real-time access to Google search data via an API.

Use Case: Fetches the latest search results for dynamic data retrieval.

15. SearXNG

Purpose: Connects agents to SearXNG, a privacy-focused internet metasearch engine.

Use Case: Combines results from multiple search engines in a secure manner.

16. Serp API

Purpose: Provides access to real-time search engine results, including Google.

Use Case: Retrieves web data with high accuracy and relevance.

17. Slack

Purpose: Integrates with Slack’s API for interactions within Slack channels.

Use Case: Facilitates seamless communication and task execution in Slack.

18. Telegram

Purpose: Interacts with Telegram’s API for messaging and automation.

Use Case: Enables agents to send messages, manage chats, or automate workflows on Telegram.

19. Web Browser

Purpose: Allows agents to browse websites and extract information.

Use Case: Useful for accessing web pages not covered by APIs or search tools.

20. Write Files

Purpose: Gives agents the ability to write and save files to disk.

Use Case: Enables file creation or modification as part of a task.

21. X (Formerly Twitter)

Purpose: Integrates with X’s API to manage posts, likes, retweets, and replies.

Use Case: Automates social media interactions for marketing or engagement purposes.


Vector Stores in LangChain

Vector stores are databases or platforms designed to store and search large volumes of vector embeddings. These embeddings represent the numerical representations of text, images, or other data types, enabling similarity-based searches and retrievals for various AI and machine learning tasks.

1. Astra

Description: Uses DataStax Astra DB, a serverless vector database, to upsert (insert or update) embedded data and perform similarity or Maximum Marginal Relevance (MMR) search upon query.

Use Case: Suitable for managing AI workloads that require scalability and reliability in production environments, especially for mission-critical applications.

2. Chroma

Description: An open-source embedding database that allows users to upsert embedded data and perform similarity search.

Use Case: Ideal for projects that require an open-source, lightweight, and fast vector search system with high flexibility.

3. Document Store (Vector)

Description: A vector database designed for searching and retrieving documents based on vectorized representations.

Use Case: Used for managing large collections of text documents and performing similarity searches to identify the most relevant documents to a query.

4. Elasticsearch

Description: A distributed search and analytics engine that supports vector search and retrieval. Elasticsearch allows you to upsert embedded data and perform similarity searches using its built-in vector search capabilities.

Use Case: Widely used for search-heavy applications, especially those that require real-time analytics and scalability across multiple nodes.

5. Faiss

Description: Developed by Meta, Faiss is a library optimized for fast similarity search of dense vectors. It can upsert embedded data and perform similarity searches efficiently.

Use Case: Excellent for AI applications that need fast, large-scale vector searches and high performance for deep learning-based tasks.

6. In-Memory Vector Store

Description: A vector store that holds embeddings directly in memory, performing exact linear searches for similarity.

Use Case: Best for small to medium-sized datasets where speed is crucial, and data can be loaded entirely into memory for fast access.

7. Meilisearch

Description: A fast and easy-to-use search engine that supports hybrid search functionality, allowing both traditional keyword-based and vector-based similarity searches.

Use Case: Suitable for applications that require both textual search and semantic vector search in a single solution.

8. Milvus

Description: A highly advanced open-source vector database that supports both vector and scalar searches. It enables fast similarity search through massive datasets.

Use Case: Ideal for large-scale machine learning and AI projects that require high-speed searches across billions of data points.

9. MongoDB Atlas

Description: A fully managed cloud database solution that supports vector search via its integration with MongoDB Atlas. It upserts embedded data and performs similarity searches with ease.

Use Case: Best for projects that require a cloud-based, scalable database with built-in vector search support.

10. OpenSearch

Description: An open-source, distributed search and analytics suite that supports vector search. It enables users to upsert embedded data and perform similarity searches.

Use Case: Ideal for open-source applications that require high scalability and advanced search features with vector capabilities.

11. Pinecone

Description: A fully managed vector database designed for high-performance, real-time similarity search. Pinecone allows easy upsertion of embedded data and performs quick similarity searches.

Use Case: Often used in AI and machine learning pipelines that require fast, scalable vector searches with minimal latency.

12. Postgres (with pgvector)

Description: A relational database that uses the pgvector extension to support vector embeddings and similarity searches.

Use Case: Combines the benefits of a relational database with the ability to handle vector search queries, making it a good choice for mixed workloads.

13. Qdrant

Description: A scalable open-source vector database written in Rust, Qdrant supports efficient upsertion and similarity searches of embedded data.

Use Case: Designed for handling high-scale vector search queries with low-latency responses in AI and ML-based applications.

14. Redis

Description: An open-source, in-memory data structure store that supports vector search capabilities through modules. Redis allows quick similarity searches on embedded data.

Use Case: Best for applications that require ultra-fast, in-memory vector searches with low-latency responses.

15. SingleStore

Description: A distributed cloud relational database that supports high-speed vector searches and upsertion of embedded data.

Use Case: Suitable for real-time analytics and AI-driven applications requiring fast, scalable vector database operations.

16. Supabase (with pgvector)

Description: An open-source alternative to Firebase that integrates PostgreSQL with the pgvector extension for vector search capabilities.

Use Case: A fast, scalable backend solution for apps that need both relational database functionality and vector-based similarity search.

17. Upstash Vector

Description: A serverless data platform that supports vector searches and upsertion of embedded data.

Use Case: Perfect for serverless applications that need fast vector search capabilities without managing infrastructure.

18. Vectara

Description: A fully managed vector search engine built to support LLM-powered search. It upserts embedded data and performs high-performance similarity search.

Use Case: A specialized platform for LLM-driven applications that need optimized search and retrieval of vector data.

19. Weaviate

Description: A scalable, open-source vector database designed for machine learning applications. Weaviate supports both vector and hybrid searches for complex datasets.

Use Case: Ideal for large-scale AI-driven applications that require robust, semantic search capabilities across multiple data types.

20. Zep Collection

Description: A fast vector search solution that supports both open-source and cloud deployments. Zep provides high-performance upsertion and similarity search features for large datasets.

Use Case: Perfect for building scalable LLM applications that require high-speed search across vast document collections.


Chat Models (LlamaIndex)

These are specific large language models (LLMs) configured for use with LlamaIndex. These models can be utilized to build conversational agents that answer questions or engage in dialogue based on their integration with LlamaIndex.

1. AzureChatOpenAI

Description: A version of OpenAI’s chat-based LLM that has been specifically configured for use with Microsoft Azure’s infrastructure, providing LlamaIndex applications with conversational AI capabilities.

Use Case: Ideal for organizations using Microsoft Azure who want to leverage OpenAI’s powerful LLMs in a cloud-native, scalable environment.

2. ChatAnthropic

Description: A conversational LLM from Anthropic, optimized for LlamaIndex. This model excels in dialogue-based interactions, ensuring safe and coherent conversations.

Use Case: Suitable for applications that require nuanced, ethical conversational AI, especially where safety and bias mitigation are key concerns.

3. ChatMistral

Description: Mistral’s LLM designed for use within LlamaIndex, optimized for generating high-quality, human-like responses in conversational scenarios.

Use Case: Best for applications requiring real-time conversations with highly dynamic responses, especially in complex and technical domains.

4. ChatOllama

Description: An LLM designed by Ollama, tailored for use with LlamaIndex. This model supports both text generation and retrieval-augmented generation (RAG) scenarios.

Use Case: Perfect for applications needing efficient text generation with the ability to leverage external data sources through retrieval-augmented methods.

5. ChatOpenAI

Description: OpenAI’s chat model, specifically fine-tuned for integration with LlamaIndex. This provides a robust platform for building conversational AI that can respond contextually based on provided input.

Use Case: Ideal for integrating OpenAI’s powerful conversational abilities into LlamaIndex-powered applications, useful in customer service, virtual assistants, and automated content generation.

6. ChatTogetherAI

Description: A version of the TogetherAI LLM optimized for use in LlamaIndex, enabling conversational AI that can handle multi-turn dialogues and nuanced contexts.

Use Case: Best for applications requiring multi-turn conversations, such as interactive assistants and educational platforms, where context and continuity matter.

7. ChatGroq

Description: Groq’s LLM designed for LlamaIndex, specifically built to provide fast and efficient AI-driven conversations.

Use Case: Suitable for high-performance applications where processing speed and scalability are crucial, such as large-scale customer service automation.


Embeddings (LlamaIndex)

Embeddings are mathematical representations of text, typically used for similarity searches and machine learning tasks. These LlamaIndex-specific embeddings transform raw data into vectors that can be queried effectively.

1. Azure OpenAI Embeddings

Description: Embeddings generated by OpenAI’s models hosted on the Azure platform, tailored for use with LlamaIndex to enable vector-based retrieval and analysis.

Use Case: Ideal for projects where embeddings need to be generated within a Microsoft Azure ecosystem, offering cloud-scale performance and OpenAI’s powerful models.

2. OpenAI Embeddings

Description: Standard embeddings generated by OpenAI’s models, used within LlamaIndex to transform text into vectors for similarity search or data processing tasks.

Use Case: Suitable for applications needing high-quality, general-purpose embeddings for tasks like semantic search, document retrieval, and NLP-based feature extraction.


Engine (LlamaIndex)

Engines in LlamaIndex manage the underlying logic for querying, processing, and answering questions from data sources. They offer different capabilities based on the nature of the application.

1. Context Chat Engine

Description: An engine that provides context-aware question answering by retrieving relevant documents and maintaining memory of past interactions.

Dev (WIll not DM), [2024/12/30 11:06]

Use Case: Best for applications that need to engage in ongoing conversations, where context from previous interactions is required to formulate accurate responses (e.g., virtual assistants, knowledge management systems).

2. Simple Chat Engine

Description: A basic engine designed to handle straightforward conversational flows without advanced memory capabilities.

Use Case: Ideal for simple conversational agents where context doesn’t need to be retained across sessions, such as FAQ bots or transactional agents.

3. Query Engine

Description: A basic query engine that answers questions over static or dynamic data sources without the need for conversational memory.

Use Case: Useful for applications where a direct, one-time query response is needed from a database or document store, such as in knowledge retrieval or data-driven applications.

4. Sub Question Query Engine

Description: A more advanced query engine that breaks down complex queries into sub-questions, retrieves relevant data from various sources, and synthesizes a final response.

Use Case: Perfect for complex queries requiring multi-step reasoning or aggregation from multiple data sources, such as multi-source research or detailed technical queries.


Response Synthesized (LlamaIndex)

These tools help to synthesize or refine responses generated from various data sources, enhancing the quality and relevance of the answers provided by the engine.

1. Compact and Refine

Description: A method that condenses text chunks into the smallest possible format while maintaining the core information.

Use Case: Useful for summarizing large documents or content, especially when concise answers or compact summaries are required for time-sensitive applications.

2. Refine

Description: A tool that refines answers by iterating over retrieved text chunks, improving the answer’s accuracy and completeness.

Use Case: Best for applications requiring detailed, nuanced answers based on multiple data sources, such as in-depth customer support or technical troubleshooting.

3. Simple Response Builder

Description: A simple method to apply a query to multiple text chunks and gather responses, returning a collection of string responses.

Use Case: Ideal for answering individual queries over a large set of data or documents, where responses need to be synthesized from multiple sources (e.g., product comparisons, research results).

4. TreeSummarize

Description: A tool specifically designed for summarizing data in a tree-like structure, useful for organizing hierarchical data.

Use Case: Suitable for summarizing complex, structured data or documents with nested content, such as legal texts, technical documents, or organizational reports.


Tools (LlamaIndex)

Tools enable LlamaIndex to interact with various data sources, execute queries, and manage workflows efficiently.

1. QueryEngine Tool

Description: A tool that invokes the QueryEngine to process and respond to user queries.

Use Case: Essential for applications that need to integrate a query engine into a larger workflow, enabling seamless data retrieval and interaction within a conversational AI system.

Vector Stores (LlamaChain)

Vector stores manage embeddings and support similarity searches. They store vectors representing text, making it easier to retrieve semantically similar information based on query vectors.

1. Pinecone

Description: A managed vector database designed for real-time similarity searches, used to upsert embedded data and perform efficient similarity searches.

Use Case: Ideal for scalable, production-level applications requiring fast, accurate similarity searches across large datasets.

2. SimpleStore

Description: A local vector store that allows users to upsert embeddings to local file paths and perform similarity searches without relying on cloud services.

Use Case: Useful for smaller-scale projects or situations where local, non-cloud storage of vector data is preferred.


Utilize (Generic)

These tools provide various utility functions that can be used to customize workflows, execute custom logic, and manage variables during execution.

1. Custom JS Function

Description: Executes custom JavaScript functions within the workflow, allowing for tailored logic execution.

Use Case: Perfect for users who need to integrate specific custom logic or operations that aren’t provided by default tools.

2. Get Variable

Description: Retrieves previously saved variables, making it possible to reference data that was set earlier in the flow.

Use Case: Useful for tracking and reusing data within a workflow, ensuring that state is preserved across different parts of the application.

3. IfElse Function

Description: Executes a decision tree logic where the flow branches based on conditions (similar to traditional if-else statements in programming).

Use Case: Critical for branching workflows where different actions are needed depending on specific conditions or user inputs.

4. Set Variable

Description: Defines and stores variables for later retrieval, enabling the flow to maintain state between operations.

Use Case: Used for tracking data across different stages of a process, ensuring that important values are maintained throughout the workflow.

5. Sticky Note

Description: Adds sticky notes to workflows, which are often used for annotations, reminders, or non-executing data.

Use Case: Useful for documentation or workflow management where notes or clarifications are necessary without impacting the workflow logic.


Web3 Tools

These tools and platforms are designed for various aspects of the Web3 ecosystem, ranging from token creation to decentralized finance (DeFi) and storage solutions.

1. Pump.fun Deployer

Description: A platform enabling the creation and launch of new cryptocurrencies (coins) on Pump.fun.

Use Case: Ideal for developers and entrepreneurs looking to create and deploy custom tokens for various blockchain applications or projects.

2. Imgnai

Description: A tool that leverages powerful image and media generation capabilities for creating custom digital assets, including artwork, NFTs, and other media.

Use Case: Useful for Web3 projects needing high-quality visual content, such as digital artists, NFT creators, and marketing teams.

3. Camelot

Description: A decentralized exchange (DEX) platform that allows users to trade various cryptocurrencies with advanced liquidity and trading features.

Use Case: Best suited for users engaging in DeFi and crypto trading, offering a seamless and efficient environment for swapping assets.

4. Nuklai

Description: Provides a suite of powerful infrastructure tools designed for Web3 and decentralized applications (dApps).

Use Case: Perfect for developers needing scalable infrastructure for building and deploying decentralized applications on blockchain networks.

5. Pinata IPFS Uploader

Description: A platform for uploading, managing, and sharing files on the InterPlanetary File System (IPFS), a decentralized storage protocol.

Use Case: Used for Web3 projects that require decentralized file storage, such as NFTs, dApp assets, or any type of digital content that needs to be immutably stored.

6. NFT Deployer and Minter

Description: A tool for deploying and minting Non-Fungible Tokens (NFTs) on blockchain networks.

Use Case: Essential for NFT creators, artists, and projects wanting to mint unique digital assets and deploy them on the blockchain.

7. Uniswap

Description: A popular decentralized exchange (DEX) that allows users to swap ERC-20 tokens directly from their wallets without relying on centralized exchanges.

Use Case: Used for trading tokens in the Ethereum ecosystem, providing users with liquidity pools and automated market-making features.

8. 1inch

Description: A decentralized exchange aggregator that helps users find the best token swap rates across multiple DEXes by routing transactions through different platforms.

Use Case: Ideal for users seeking the most efficient and cost-effective token swaps, helping maximize value by aggregating liquidity.

9. Native Token

Description: A Web3 tool that facilitates the wrapping, transferring, and checking of balances for native tokens across different blockchain ecosystems.

Use Case: Used for managing native blockchain assets such as ETH, BTC, or any native chain token, enabling easy transfers and balance checks.

10. ERC20 Token

Description: A tool that manages ERC-20 tokens, allowing for transfers, approvals, and balance checks within Ethereum-based wallets.

Use Case: Primarily used for managing ERC-20 token interactions in dApps and DeFi protocols, enabling seamless token management.

11. CrossChain Bridge

Description: A tool that facilitates the transfer of tokens across different blockchain networks, enabling cross-chain interoperability.

Use Case: Ideal for users who need to move assets between various blockchain ecosystems, such as Ethereum, Binance Smart Chain, and others.

12. Coingecko

Description: Provides real-time cryptocurrency data, including prices, market cap, volume, and historical data, sourced from a wide range of exchanges.

Use Case: Used by traders, developers, and analysts to track crypto market trends, prices, and other metrics for making informed decisions.

13. DexScreener

Description: A tool that aggregates data from decentralized exchanges (DEXes), providing users with real-time market data and insights.

Use Case: Perfect for traders looking to track market movements on various DEX platforms and analyze trends for better trading strategies.


Upcoming Partners

These are Web3 platforms or projects that are expected to launch soon on Sage, each offering unique features and services for the ecosystem:

1. *****

Description: A secure liquidity vault system designed for Web3 applications.

Use Case: Likely to be used for storing and managing liquidity in a secure, decentralized way for DeFi platforms or dApps.

2. *****

Description: A leading memecoin project.

Use Case: Likely focused on creating a popular, community-driven cryptocurrency, ideal for fun or social media-driven projects.

3. *****

Description: A platform focused on yield optimization within the decentralized finance (DeFi) space.

Use Case: Perfect for DeFi users seeking to maximize returns on their assets through efficient yield farming strategies.

4. *****

Description: A leverage trading platform within the DeFi ecosystem.

Use Case: Targeted at advanced traders looking for leverage options to amplify their positions in cryptocurrency markets.

5. *****

Description: A self-learning and improving agent platform for Web3 applications.

Use Case: Likely aimed at creating AI-driven agents that can autonomously evolve and improve their functions over time, potentially used in customer support or automated trading.

6. *****

Description: A platform for job creation and talent searching within the Web3 space.

Use Case: Ideal for Web3 projects and companies looking to find skilled developers, marketers, and other talent specifically for blockchain-based initiatives.

7. *****

Description: An AI-powered decentralized finance platform.

Use Case: Likely focused on using artificial intelligence to enhance DeFi protocols, optimize transactions, or improve liquidity management.

8. *****

Description: An AI agent trading platform.

Use Case: Likely used for algorithmic trading or AI-driven investment strategies, allowing users to automate their trades based on AI decision-making.

9. *****

Description: A platform for uncensored coding agents.

Use Case: Likely to offer decentralized development tools, allowing developers to create and deploy code without restrictions or censorship.

10. *****

Description: A decentralized storage platform for permanent data storage on the blockchain.

Use Case: Ideal for decentralized applications (dApps), ensuring that data stored is immutable and permanently accessible without relying on traditional cloud storage solutions.

Start building with Sage

Start building with Sage