Struct RawDocument 

pub struct RawDocument {

    pub source: String,
    pub source_id: String,
    pub title: String,
    pub summary: Option<String>,
    pub content: String,
    pub url: Option<String>,
    pub last_modified: i64,
    pub metadata: HashMap<String, String>,
    pub memory_type: Option<String>,
    pub domain: Option<String>,
    pub source_agent: Option<String>,
    pub confidence: Option<f32>,
    pub confirmed: Option<bool>,
    pub stability: Option<String>,
    pub supersedes: Option<String>,
    pub pending_revision: bool,
    pub entity_id: Option<String>,
    pub quality: Option<String>,
    pub is_recap: bool,
    pub enrichment_status: String,
    pub supersede_mode: String,
    pub structured_fields: Option<String>,
    pub retrieval_cue: Option<String>,
    pub source_text: Option<String>,
}

A raw document fetched from any source, ready for chunking and embedding.

Fields

source: String

Source identifier (“gmail”, “notion”, “local_files”, etc.)

source_id: String

Unique ID within the source (message ID, page ID, file path)

title: String

Document title (filename, subject line, page title)

summary: Option<String>

LLM-generated summary (stored separately from chunk content)

content: String

Plain text content

url: Option<String>

Deep link back to the source (URL, file path)

last_modified: i64

Unix timestamp of last modification

metadata: HashMap<String, String>

Additional metadata

memory_type: Option<String>

Memory category: “preference”, “decision”, “fact”, “goal”, “relationship”

domain: Option<String>

Domain context: “work”, “personal”, “health”, or “project:”

source_agent: Option<String>

Which AI agent stored this memory (e.g. “claude-code”, “chatgpt”)

confidence: Option<f32>

Confidence score (0.0-1.0) assigned by the storing agent

confirmed: Option<bool>

Whether a human has confirmed this memory

stability: Option<String>

Stability tier: “new”, “learned”, or “confirmed”

supersedes: Option<String>

source_id of the memory this entry supersedes (version chain)

pending_revision: bool

Whether this is a pending revision awaiting human approval (Protected tier supersede)

entity_id: Option<String>

Link to a knowledge graph entity (nullable, cascade handled manually)

quality: Option<String>

Quality assessment: “low”, “medium”, “high” (NULL = unassessed)

is_recap: bool

Whether this memory is a recap/summary of other memories

enrichment_status: String

Deprecated: enrichment status is now derived from the enrichment_steps table. This field is ignored on INSERT. Kept for API compatibility with downstream consumers.

supersede_mode: String

How superseded content is handled: “hide” (default) or “archive” (visible but muted)

structured_fields: Option<String>

JSON object with type-specific structured fields (e.g. {“claim”: “…”, “context”: “…”})

retrieval_cue: Option<String>

LLM-generated question this memory answers – embedded for vector search

source_text: Option<String>

Original prose content, preserved when structured_fields are promoted to primary content

Trait Implementations

impl Clone for RawDocument

fn clone(&self) -> RawDocument

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RawDocument

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

Formats the value using the given formatter.

impl Default for RawDocument

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for RawDocument

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for RawDocument

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

where __S: Serializer,

Serialize this value into the given Serde serializer.