Struct MemoryConfig 

pub struct MemoryConfig {

    pub enabled: bool,
    pub auto_inject: bool,
    pub auto_retain: bool,
    pub retain_llm: bool,
    pub retain_llm_always: bool,
    pub retain_every_n: usize,
    pub retain_max_facts: usize,
    pub max_index_lines: usize,
    pub max_index_bytes: usize,
    pub recall_top_k: usize,
    pub recall_min_score: f32,
    pub recall_token_budget: usize,
    pub embed_dim: usize,
    pub scope: String,
    pub embed_backend: String,
    pub code_inject: bool,
    pub code_top_k: usize,
    pub code_min_score: f32,
    pub subagent_banks: bool,
    pub auto_index: bool,
    pub auto_index_max_files: usize,
    pub auto_index_max_chunks: usize,
    pub session_inject: bool,
    pub session_top_k: usize,
    pub session_min_score: f32,
    pub consolidate: bool,
    pub consolidate_max: usize,
}

Cross-session semantic / auto memory (Claude-style index + local recall).

Fields

enabled: bool

Master switch. Default on (cheap hashing embedder; no ONNX).

auto_inject: bool

Auto-inject top-k semantic hits for the current user message.

auto_retain: bool

Post-turn auto-retain (heuristic and/or LLM). Default on.

retain_llm: bool

Also call a small LLM to extract durable facts when heuristic finds none (or always when retain_llm_always). Default on.

retain_llm_always: bool

Run LLM retain even if heuristic already found facts.

retain_every_n: usize

Run retain every N user turns (1 = every turn).

retain_max_facts: usize

Max facts retained per turn.

max_index_lines: usize

Max lines of MEMORY.md loaded into every session (Claude parity: 200).

max_index_bytes: usize

Max bytes of MEMORY.md loaded into every session (Claude parity: 25 KiB).

recall_top_k: usize

Max recalled facts injected per turn.

recall_min_score: f32

Minimum cosine similarity for a fact to be recalled.

recall_token_budget: usize

Token budget for recalled facts (chars ≈ tokens × 4).

embed_dim: usize

Hashing embedder dimension (stored BLOB width).

scope: String

user (data_dir, default) or project (.whycodes/memory, git-shareable).

embed_backend: String

hash (default) or onnx (MiniLM; needs --features onnx).

code_inject: bool

Inject code-index hits when available.

code_top_k: usize

code_min_score: f32

subagent_banks: bool

Give subagents their own memory bank (project::agent_name).

auto_index: bool

On session start, ensure a code index exists (skip if already indexed).

auto_index_max_files: usize

Max source files to walk when auto-indexing.

auto_index_max_chunks: usize

Max chunks when auto-indexing.

session_inject: bool

Inject related past-session turns.

session_top_k: usize

session_min_score: f32

consolidate: bool

Cap the fact bank after retain.

consolidate_max: usize

Trait Implementations

impl Clone for MemoryConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for MemoryConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for MemoryConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for MemoryConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for MemoryConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.