Crate bamboo_compression 

Token budget management for LLM conversations.

This module provides token budget management to prevent context window overflow while maintaining conversation continuity. It implements a hybrid approach combining rolling summary with recent message windowing.

Key Components

• types: Core types like TokenBudget, BudgetStrategy, PreparedContext
• counter: Token counting via BPE tokenizer (tiktoken) or heuristic estimation
• segmenter: Message segmentation preserving tool-call atomicity
• limits: Model context window limits registry
• preparation: Context preparation with budget enforcement
• summarizer: Conversation summarization for context preservation

Re-exports

pub use compression_tooling::active_messages_for_budget;

pub use compression_tooling::apply_compression_plan;

pub use compression_tooling::build_compression_plan_with_summary;

pub use compression_tooling::build_forced_compression_plan_with_summary;

pub use compression_tooling::build_summary_prompt;

pub use compression_tooling::compression_summary_message;

pub use compression_tooling::context_window_usage_percent;

pub use compression_tooling::estimate_context_compression_exposure;

pub use compression_tooling::normalized_trigger_percent;

pub use compression_tooling::summary_source_messages;

pub use compression_tooling::CompressionPlan;

pub use compression_tooling::CompressionPlanError;

pub use compression_tooling::ContextCompressionExposure;

pub use counter::HeuristicTokenCounter;

pub use counter::TiktokenTokenCounter;

pub use counter::TokenCounter;

pub use limits::create_budget_for_model;

pub use limits::ModelLimitsRegistry;

pub use preparation::estimate_prompt_cache_savings;

pub use preparation::prepare_hybrid_context;

pub use segmenter::MessageSegmenter;

pub use summarizer::HeuristicSummarizer;

pub use summarizer::LlmSummarizer;

pub use summarizer::Summarizer;

pub use summarizer::SummaryManager;

pub use summarizer::SummaryMode;

pub use summarizer::SummaryTrigger;

pub use types::BudgetError;

pub use types::PreparedContext;

Modules

compression_tooling

counter

Token counting for budget management.

limits

Model context window limits registry.

preparation

Context preparation for budget management.

segmenter

Message segmentation for budget management.

summarizer

Conversation summarization for rolling context management.

types

Core types for token budget management.

Structs

TokenBudget

Token budget configuration for a conversation.

TokenUsageBreakdown

Detailed token usage breakdown.

Enums

BudgetStrategy