Module reader

Expand description

MicroscopeReader — high-performance memory-mapped reader for the binary index.

Structs§

MicroscopeReader: High-performance memory-mapped reader for the Microscope index.
RadialResult: A single result from radial search.
ResultSet: ResultSet from radial search: primary hit + distance-weighted neighbors.

DataStore: Backing store for block data — either memory-mapped or decompressed in-memory.

append_emotion_log: Append an emotion vector to the emotion log (emotion_log.bin). Format: [u64 timestamp_ms] [f32; 21 emotion] [u32 text_len] [bytes text]
build_emotions_from_log: Build emotions.bin from the emotion log and main index. Reads emotion_log.bin and maps each entry to the closest main-index block.
contains_ci: Allocation-free, case-insensitive substring test.
emotional_similarity: Cosine similarity between two 21D emotion vectors.
format_emotion: Format an emotion vector as a human-readable string.
layer_color
load_emotion_lookup: Load emotions.bin and return a lookup closure. emotions.bin format: flat array of [f32; 21] per block index.
parse_imp_marker: Parse a leading (imp=N) marker from a layer entry.
print_append_result: Display a single append-log result entry.
read_append_log
relevance_score: Score how well a block answers a text query.
store_memory
store_memory_temporary: Store memory to append log and timeline only (NOT to layer files). Used for temporary/internal thoughts that should not persist through rebuilds.
store_memory_with_embedding: Store a memory whose embedding the caller has already computed.
store_memory_with_emotion: Variant with emotion vector.
store_memory_with_status
write_emotion: Write a single block’s emotion vector to emotions.bin. The file is grown to fit the block index if needed.
write_emotions_batch: Write many block emotions with a single read-modify-write.