Module assembler 

Context assembler: orders evaluated entries and applies token budgets.

After entries are activated and evaluated, the assembler sorts them by slot and priority, then fits them within token budgets. The result is an ordered list of AssembledBlocks that the host application inserts into the LLM prompt.

Slots

Entries declare which slot they target. Slots represent functional depth in the context window, from deep background to immediate foreground. The host application’s ContextDefinition declares which slots exist and where they appear in the prompt.

┌─────────────────────────────┐
│  preamble                   │  ← world axioms, framing rules
│  foundation                 │  ← core lore, character backstory
│  context                    │  ← situational scene info
│  reference                  │  ← supporting material, style guides
│  framing                    │  ← scene mood, active mode
│  guidance                   │  ← behavior rules, quest objectives
│  emphasis                   │  ← active effects, high-attention
│  [chat messages...]         │
│  immediate                  │  ← urgent reminders near chat
│  at_depth(N)                │  ← injected N msgs from end
│  aftermath                  │  ← post-chat notes, summaries
└─────────────────────────────┘

Structs

AssembledBlock

A fully assembled context block, ready for prompt insertion.

ContextAssembler

GuesstimationTokenizer

A rough tokenizer that estimates ~4 characters per token.

TokenBudget

Token budget tracking for assembled context.

Enums

Slot

Where in the prompt an entry’s content should be inserted.

Traits

Tokenizer

Trait for estimating token counts of text content.