> For the complete documentation index, see [llms.txt](https://prism-ai-1.gitbook.io/prismai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://prism-ai-1.gitbook.io/prismai/core-functionality/images-and-media.md). # Prism core framework
Prism Framework is designed to address complex tasks through its innovative **Prompt Manager** and **Memory Manager**, two foundational components developed to automate prompt generation and manage chat history efficiently. The framework offers a scalable and user-friendly **Multi-Agent System** with the following features: *** **1. Agent Planner** * **Core Role**: Decomposes complex tasks into actionable steps, generating efficient action plans. * **Four Agent Types**: * **Markets Agent**: Handles trading and market insights. * **Image Agent**: Focused on image analysis and visual data processing. * **Social Media Agent**: Monitors and analyzes social platforms like Discord, Telegram, and Twitter. * **Analysis Agent**: Supports data evaluation, reporting, and decision-making. * **Planning Techniques**: * **Single-Path Strategy**: Creates a single pathway for achieving tasks without exploring alternatives. * **Multi-Path Strategy**: Evaluates and selects optimal paths from multiple generated plans. *** **2. Connector** * **Purpose**: Provides seamless integration with external systems via APIs and communication tools. * **Core Features**: * **API Integration Layer**: Interfaces with blockchain systems (e.g., Ethereum, Solana, Base). * **Data Transformation Layer**: Converts and processes data to ensure compatibility across systems. * **Event-Driven Layer**: Synchronizes external data in real-time based on triggers. * **Workflow Engine**: Defines automated workflows for data processing and system interactions. * **UI Management Dashboard**: Offers an intuitive interface for managing connectors. * **Logging and Monitoring**: Tracks data flows, operations, and system performance for debugging and optimization. *** **3. Memory Manager** * **Functionality**: Supports chat history storage, information compression, and memory retrieval. * **Key Components**: * **Message History**: Maintains recent conversations for consistent dialogue context. * **Fact Memory**: Stores user-specific or context-based facts for personalized responses. * **Knowledge Base**: Houses static data for answering broader queries. * **Relationship Tracking**: Monitors user-agent interactions, including frequency and emotional tone, to enhance personalization. * **RAG Integration**: Employs vector search to retrieve contextually relevant memories or knowledge for dynamic responses. *** **4. Prompt Manager** * **Goal**: Automates prompt construction for coordinating tasks across multiple LLMs. * **Prompt Types**: * **System Prompt**: Defines agent roles, tasks, and responsibilities. * **Context Prompt**: Incorporates document, code, and tool-related contexts for precise task execution. * **Customized Prompt**: Tailored prompts for specific tasks, supporting structured outputs like JSON or code snippets. * **Automation**: * Configures prompts dynamically through role and task definitions. * Provides reusable context strategies, such as session and tool retrieval. * Enables personalized prompts for diverse agent requirements. *** **5. Component Ecosystem** * **Retrieval**: * **Doc Retrieval**: Leverages vector databases to manage document-based knowledge. * **Code Retrieval**: Enhances LLM capabilities with contextual code analysis and generation. * **Search Retrieval**: Integrates internet searches (e.g., DuckDuckGo) for real-time data augmentation. * **Tool Integration**: Supports the rapid registration of tools using a Python-based interface (e.g., LangChain). * **Action Execution**: Defines dynamic actions, including knowledge retrieval, tool usage, and code execution. *** **6. Memory Systems (Eliza Framework)** * **Purpose**: Enables agents to understand context, maintain long-term interactions, and adapt dynamically. * **Memory Types**: * **Message History**: Ensures continuity in ongoing conversations. * **Fact Memory**: Retains user-specific data for personalized experiences. * **Knowledge Base**: Provides static references for broader information needs. * **Relationship Tracking**: Builds rapport through emotional and historical interaction data. * **Trend Analysis**: Supports temporal analysis for market and transactional changes. *** This framework represents a modular and highly adaptable system, empowering agents to perform complex tasks with precision and scalability, catering to the dynamic needs of the crypto ecosystem. *** ***