Module compaction 

Context compaction for long conversations

This module provides functionality to compact conversation history when it becomes too large, using the LLM itself to summarize older messages.

Structs

CompactedContext

Result of context compaction

CompactionConfig

Compaction configuration for LLM-based compaction

CompactionManager

Context manager that handles compaction automatically

CompactionMetadata

Metadata about a compaction operation

LlmCompactor

LLM-based compactor that uses the model itself to summarize

NoopContextTransformer

A no-op transformer that returns the context unchanged.

Enums

CompactionError

Error type for compaction operations

CompactionStrategy

Compaction strategy determining when to compact

Traits

Compactor

Trait for context compaction implementations

ContextTransformer

Context transformer — applied before provider stream call.

Functions

generate_branch_summary

Generate a concise summary of the last N conversation messages.

Type Aliases

FrameBag

Rendered snapcompact PNG frames: (frame_index, png_bytes).