Results
The builder returns OptimizedContext. Its fields map directly to the exported structs in core/types.rs, including ScoredChunk, ScoreBreakdown, and OptimizationStats.
Optimized context#
result.content
result.chunks
result.total_tokens
result.stats
result.compression_ratio()| Field | Type | Meaning |
|---|---|---|
| content | String | The final grouped prompt string. |
| chunks | Vec<ScoredChunk> | Every chunk that survived selection. |
| total_tokens | usize | Final token count for the output string. |
| stats | OptimizationStats | Diagnostics for the run pipeline. |
Score breakdown#
| Field | Meaning |
|---|---|
| algorithmic | The position-based heuristic currently written into the score breakdown. |
| semantic | Embedding similarity against the query. |
| llm | Reserved compatibility field; currently unused. |
| priority_boost | Contribution from chunk priority. |
| recency_factor | Conversation-style recency factor used during chunk scoring. |
Stats#
| Field | Meaning |
|---|---|
| input_tokens | Total token count before budget selection. |
| output_tokens | Tokens after optimization. |
| chunks_processed | How many chunks entered scoring. |
| chunks_selected | How many chunks survived selection. |
| processing_time_ms | Runtime duration in milliseconds. |
| compression_ratio | Input tokens divided by output tokens. |
Results pages explain how Forgetless selects chunks, reports scoring diagnostics, tracks compression ratios, and returns the final optimized context payload for downstream LLM calls.
Related#
Configuration
Configure Forgetless token budgets, scoring weights, chunk behavior, local model options, and feature toggles to tune LLM context for your workload.
Tokenizer
Count LLM tokens in Rust with cl100k_base or a custom chars-per-token mode, plus truncation helpers and image token estimation to plan context budgets.
HTTP Server
Run the optional Forgetless HTTP server built on Axum: a multipart API for language-agnostic context optimization with a health endpoint and file priorities.