Getting Started

Install the crate, choose a token budget, attach content and files, then call `run()` to produce an optimized context block.

Install the crate

1cargo add forgetless

The current crate line in the repository is 0.1.1.

1[dependencies]
2forgetless = "0.1.1"

Optional features

1# HTTP server
2cargo add forgetless --features server
3
4# Apple Silicon GPU inference
5cargo add forgetless --features metal
6
7# NVIDIA GPU inference
8cargo add forgetless --features cuda
9
10# CPU acceleration
11cargo add forgetless --features accelerate
12cargo add forgetless --features mkl

Build the first pipeline

1use forgetless::{Config, Forgetless, WithPriority};
2
3let result = Forgetless::new()
4    .config(Config::default().context_limit(48_000))
5    .add(WithPriority::critical(system_prompt))
6    .add(conversation_history)
7    .add_file("spec.pdf")
8    .add_file("screenshot.png")
9    .query("What changed in the API contract?")
10    .run()
11    .await?;
12
13println!("{}", result.content);
14println!("compression: {:.1}x", result.compression_ratio());

Attach content and files with priorities

1use forgetless::{FileWithPriority, Forgetless, WithPriority};
2
3let result = Forgetless::new()
4    .add_pinned("You are preparing a release summary.")
5    .add(WithPriority::high(recent_messages))
6    .add(WithPriority::low(archived_notes))
7    .add_file(FileWithPriority::critical("system.md"))
8    .add_file(FileWithPriority::high("design-review.pdf"))
9    .add_files(["src/lib.rs", "src/builder.rs"])
10    .run()
11    .await?;

Add in-memory bytes

1use forgetless::{Forgetless, Priority};
2
3let png_bytes = std::fs::read("diagram.png")?;
4
5let result = Forgetless::new()
6    .add_bytes(b"temporary notes", "text/plain")
7    .add_bytes_p(&png_bytes, "image/png", Priority::High)
8    .run()
9    .await?;

In the current builder implementation, non-text byte payloads are turned into placeholder content records rather than being parsed as rich binary inputs.

Runtime configuration

Option	Default	What it changes
context_limit	128,000	Maximum output token budget.
vision_llm	false	Loads the local SmolVLM vision model for image descriptions.
context_llm	false	Loads the local SmolLM2-135M model for post-selection polishing.
chunk_size	512	Target chunk size passed into the chunker.
parallel	true	Stored on config; current file reads still use parallel processing.
cache	true	Stored on config; the current global embedding cache is not toggled by this flag.

1use forgetless::{Config, Forgetless};
2
3let result = Forgetless::new()
4    .config(
5        Config::default()
6            .context_limit(96_000)
7            .chunk_size(256)
8            .vision_llm(true)
9            .context_llm(true)
10            .parallel(true)
11            .cache(true),
12    )
13    .add_file("diagram.png")
14    .run()
15    .await?;

Read the result

1result.content
2result.total_tokens
3result.chunks
4
5result.stats.input_tokens
6result.stats.output_tokens
7result.stats.chunks_processed
8result.stats.chunks_selected
9result.stats.processing_time_ms
10
11result.compression_ratio()

Introduction

Builder