ComfyUI_LiteLLM

AgentNode

Iterative Completion Node with Memory and Recursion Support This node performs iterative LLM completions with optional memory integration and recursion filtering. It processes multiple prompts across iterations, maintaining conversation state and applying custom processing filters. Features: - Multi-iteration processing with configurable max iterations - Memory provider integration for RAG-like functionality - Recursion filters for post-processing completions - Caching system for response persistence - Support for multiple prompts in a single run Note: Despite the name "AgentNode", this is primarily an iterative completion processor. Consider renaming to "IterativeCompletionNode" for clarity.

6 outputs

BasicRecursionFilterNode

Iterative Completion Enhancement Filter This node creates a recursion filter that enhances completions through multiple rounds of processing. It takes an initial completion and iteratively improves it using a customizable prompt template. The filter is designed to: - Expand on initial completions with deeper analysis - Consider user intent and context more thoroughly - Apply multiple rounds of enhancement up to max_depth Usage: Connect the output to AgentNode's recursion_filter input to enable iterative completion enhancement during agent processing. Note: Consider renaming to "CompletionEnhancementFilterNode" for clarity.

2 inputs · 2 parameters · 1 output

Completion Enhancement Filter

Compact description unavailable for this node.

Document Chunk Processor

Compact description unavailable for this node.

DocumentChunkRecursionFilterNode

Document Chunk Processing Filter This node creates a recursion filter that processes documents in chunks. It's designed for handling large documents by breaking them into manageable pieces and processing each chunk sequentially. The filter: - Chunks documents into specified sizes - Processes one chunk per agent iteration - Maintains state across chunk processing - Can be combined with other recursion filters Usage: Connect to AgentNode's recursion_filter input for document processing. Set max_iterations on AgentNode to match expected number of chunks. Note: Consider renaming to "DocumentChunkProcessorNode" for clarity.

2 inputs · 2 parameters · 1 output

DocumentProcessor

Compact description unavailable for this node.

Iterative Completion Agent

Compact description unavailable for this node.

LinuxMemoryDirectory

Compact description unavailable for this node.

QueryNode

- Display Name: "LightRAG Query Node" - Purpose: Queries the LightRAG knowledge graph with advanced retrieval strategies - Query Modes: - naive: Simple keyword-based retrieval - local: Local context-aware retrieval - global: Global knowledge graph analysis - hybrid: Combines multiple strategies for optimal results

1 input · 2 parameters · 1 output

tmp_dir

Compact description unavailable for this node.