Crate forgetless 

Forgetless

Context optimization for LLMs. Takes massive context, outputs optimized version that fits your token budget.

Quick Start

use forgetless::Forgetless;

let result = Forgetless::new(128_000)
    .add("system prompt + conversation + everything...")
    .add_file("document.pdf")
    .add_files(&["code.rs", "data.json"])
    .run()
    .await?;

// Send to your LLM
let response = your_llm.chat(&result.content).await?;

Re-exports

pub use core::config::Config;

pub use core::config::ForgetlessConfig;

pub use core::config::ScoringConfig;

pub use core::error::Error;

pub use core::error::Result;

pub use core::types::OptimizationStats;

pub use core::types::OptimizedContext;

pub use core::types::PolishedContext;

pub use core::types::ScoredChunk;

pub use core::types::ScoreBreakdown;

pub use builder::Forgetless;

pub use input::content::ContentInput;

pub use input::content::FileWithPriority;

pub use input::content::IntoContent;

pub use input::content::IntoFileContent;

pub use input::content::WithPriority;

pub use input::file::read_file_content;

pub use processing::chunking::Chunk;

pub use processing::chunking::ChunkConfig;

pub use processing::chunking::Chunker;

pub use processing::chunking::ContentType;

pub use processing::scoring::Priority;

pub use processing::token::TokenCounter;

pub use processing::token::TokenizerModel;

pub use ai::embeddings::cosine_similarity;

pub use ai::embeddings::embed_batch;

pub use ai::embeddings::embed_text;

pub use ai::embeddings::EmbeddingCache;

pub use ai::llm::LLMConfig;

pub use ai::llm::Quantization;

pub use ai::llm::LLM;

pub use ai::vision::describe_image;

pub use ai::vision::describe_image_with_prompt;

pub use ai::vision::init_vision;

pub use ai::vision::is_vision_ready;

Modules

ai

AI modules - embeddings, LLM, and vision

builder

Forgetless builder and optimization pipeline

core

Core types, configuration, and error handling

input

Input handling - content types and file parsing

processing

Content processing - chunking, scoring, and tokenization

Constants

VERSION

Library version