Struct ImproveParams 

pub struct ImproveParams<'a> {

    pub dataset_name: String,
    pub session_ids: Option<Vec<String>>,
    pub node_name: Option<Vec<String>>,
    pub owner_id: Uuid,
    pub tenant_id: Option<Uuid>,
    pub feedback_alpha: f64,
    pub extraction_tasks: Option<Vec<String>>,
    pub enrichment_tasks: Option<Vec<String>>,
    pub data: Option<String>,
    pub build_global_context_index: bool,
    pub run_in_background: bool,
    pub llm: Arc<dyn Llm>,
    pub storage: Arc<dyn StorageTrait>,
    pub graph_db: Arc<dyn GraphDBTrait>,
    pub vector_db: Arc<dyn VectorDB>,
    pub embedding_engine: Arc<dyn EmbeddingEngine>,
    pub ontology_resolver: Arc<dyn OntologyResolver>,
    pub db: Option<Arc<DatabaseConnection>>,
    pub session_store: Option<Arc<dyn SessionStore>>,
    pub session_manager: Option<Arc<SessionManager>>,
    pub add_pipeline: Option<&'a AddPipeline>,
    pub checkpoint_store: Option<Arc<dyn CheckpointStore>>,
    pub cognify_config: &'a CognifyConfig,
}

Parameters for improve.

All fields are required at construction time — Default is intentionally not derived because several fields (owner_id, the engine handles, and cognify_config) have no sensible default value. This forces every caller to think about each dependency. Callers that omit optional behavior should pass None explicitly for the Option<...> fields.

LIB-04 (Decision 8) introduced this struct to replace the previous 18 positional parameters. E-05 (this commit) extended it with three v2 power-user fields — extraction_tasks, enrichment_tasks, data — matching Python’s ImprovePayloadDTO field-for-field. They are pure-data fields and currently informational: the orchestrator does not yet branch on them, but they are accepted by the constructor so callers (especially the HTTP layer) can plumb the raw payload through without dropping fields.

Fields

dataset_name: String

Dataset name to operate on (Stage 2 persistence + Stage 4 lookup).

session_ids: Option<Vec<String>>

Session ids that drive Stages 1, 2, and 4. None or empty skips them.

node_name: Option<Vec<String>>

Optional graph node-name filter applied to the memify (Stage 3) pass.

owner_id: Uuid

Owner UUID under which graph/session reads and writes are scoped.

tenant_id: Option<Uuid>

Optional tenant UUID for multi-tenant deployments.

feedback_alpha: f64

Mixing factor for feedback weight propagation (Stage 1).

extraction_tasks: Option<Vec<String>>

Optional list of extraction-task identifiers (Python parity: extraction_tasks: Optional[List[str]]). Currently informational — reserved for future power-user overrides matching Python’s ImproveKwargs.extraction_tasks.

enrichment_tasks: Option<Vec<String>>

Optional list of enrichment-task identifiers (Python parity: enrichment_tasks: Optional[List[str]]). Currently informational.

data: Option<String>

Optional inline text payload (Python parity: data: Optional[str]). Currently informational; reserved for future power-user overrides.

build_global_context_index: bool

When true and not running in background, build the global context index (graph summary) after Stage 3.

Mirrors Python’s build_global_context_index parameter. Default false (opt-in) — matches Python parity.

run_in_background: bool

When true, treat this as a background run: skips stages that require the prior stage to have completed synchronously (e.g. the global context index and the sync-graph stage).

Background dispatch is handled by the host (HTTP server or CLI); this flag is used only for stage-skipping logic parity with Python.

llm: Arc<dyn Llm>

LLM handle (used by Stage 2 cognify-of-session-text).

storage: Arc<dyn StorageTrait>

File storage handle.

graph_db: Arc<dyn GraphDBTrait>

Graph database handle.

vector_db: Arc<dyn VectorDB>

Vector database handle.

embedding_engine: Arc<dyn EmbeddingEngine>

Embedding engine handle.

ontology_resolver: Arc<dyn OntologyResolver>

Ontology resolver handle.

db: Option<Arc<DatabaseConnection>>

Metadata DB connection. Required for Stage 4 (dataset lookup).

session_store: Option<Arc<dyn SessionStore>>

Session backing store. Required for Stages 1 and 2.

session_manager: Option<Arc<SessionManager>>

Session manager. Required for Stages 1 and 4.

add_pipeline: Option<&'a AddPipeline>

Add pipeline (borrowed). Required for Stage 2.

checkpoint_store: Option<Arc<dyn CheckpointStore>>

Checkpoint store. Required for Stage 4.

cognify_config: &'a CognifyConfig

Borrowed cognify configuration used by Stage 2 persistence.