Update Ml Settings

PUT

/api/creative/lore/ml/settings

• DivinityEngine - Local development
• DivinityEngine - Localhost alternative

Update per-project ML settings. Persisted in project settings JSON.

Request Body ^required

LoreMLSettings

Complete per-project ML settings. All values have sensible defaults.

object

pipeline

LoreMLSettingsPipeline

Pipeline-level training settings.

object

minEntriesForTraining

Minentriesfortraining

Min entries to enable ML training

integer

default: 5 >= 1

retrainGrowthThreshold

Retraingrowththreshold

Corpus growth % to trigger retrain

number

default: 0.1 >= 0.01 <= 1

anomaly

LoreMLSettingsAnomaly

Anomaly detection settings.

object

weights

LoreMLSettingsAnomalyWeights

Signal weights for multi-signal anomaly detection (must sum to ~1.0).

object

isolationForest

Isolationforest

number

default: 0.25 <= 1

orphan

Orphan

number

default: 0.1 <= 1

completeness

Completeness

number

default: 0.15 <= 1

tagIsolation

Tagisolation

number

default: 0.1 <= 1

sizeOutlier

Sizeoutlier

number

default: 0.1 <= 1

staleness

Staleness

number

default: 0.1 <= 1

kgeOutlier

Kgeoutlier

number

default: 0.1 <= 1

spectralIsolation

Spectralisolation

number

default: 0.1 <= 1

minEntries

Minentries

Minimum entries required to run anomaly detection

integer

default: 5 >= 2

reasonThreshold

Reasonthreshold

Min signal score to generate a reason string

number

default: 0.3 <= 1

logThreshold

Logthreshold

Composite score threshold for logging flagged anomalies

number

default: 0.5 <= 1

orphanDivisor

Orphandivisor

Connections needed for 0.0 orphan score

number

default: 10 >= 1

stalenessWindowDays

Stalenesswindowdays

Days behind sibling for max staleness score

integer

default: 180 >= 7

zScoreDivisor

Zscoredivisor

Z-score divisor for size outlier normalization

number

default: 3 >= 1

isolationForestContamination

Isolationforestcontamination

IsolationForest contamination parameter

number

default: 0.05 >= 0.01 <= 0.5

similarity

LoreMLSettingsSimilarity

Similarity detection settings.

object

suggestedLinkThreshold

Suggestedlinkthreshold

Min similarity for suggesting a link

number

default: 0.5 <= 1

nearDuplicateThreshold

Nearduplicatethreshold

Min similarity for near-duplicate detection

number

default: 0.85 <= 1

missingEntryMinMentions

Missingentryminmentions

Min mentions for a missing entry suggestion

integer

default: 3 >= 1

maxSuggestedLinks

Maxsuggestedlinks

Max suggested links returned

integer

default: 100 >= 10

maxNearDuplicates

Maxnearduplicates

Max near-duplicates returned

integer

default: 50 >= 10

maxMissingEntries

Maxmissingentries

Max missing entries returned

integer

default: 50 >= 10

clustering

LoreMLSettingsClustering

Clustering settings.

object

maxClusters

Maxclusters

Maximum clusters allowed

integer

default: 20 >= 2

minClusters

Minclusters

Minimum clusters floor

integer

default: 2 >= 2

tagTfidfMaxFeatures

Tagtfidfmaxfeatures

Max TF-IDF features for tag encoding

integer

default: 200 >= 10

maxCategories

Maxcategories

Max categories for one-hot encoding

integer

default: 50 >= 5

labelTfidfMaxFeatures

Labeltfidfmaxfeatures

Max TF-IDF features for cluster label generation

integer

default: 500 >= 50

labelMinTokenLength

Labelmintokenlength

Min characters for a word to appear in cluster labels

integer

default: 4 >= 2

contradiction

LoreMLSettingsContradiction

Contradiction detection settings.

object

minMentions

Minmentions

Min mentions for contradiction check

integer

default: 2 >= 2

similarityThreshold

Similaritythreshold

Snippet embedding similarity below this = potential contradiction

number

default: 0.45 <= 1

severityThreshold

Severitythreshold

Below this = medium severity (else low)

number

default: 0.3 <= 1

maxMentionsCompared

Maxmentionscompared

Max mentions compared per entity

integer

default: 50 >= 5

maxContradictionsTotal

Maxcontradictionstotal

Stop after finding this many

integer

default: 200 >= 10

maturity

LoreMLSettingsMaturity

Maturity tier thresholds (each 0.0-1.0).

object

completeThreshold

Completethreshold

number

default: 0.9 <= 1

matureThreshold

Maturethreshold

number

default: 0.75 <= 1

developingThreshold

Developingthreshold

number

default: 0.5 <= 1

draftThreshold

Draftthreshold

number

default: 0.25 <= 1

coverage

LoreMLSettingsCoverage

Coverage analysis settings.

object

typeGapThreshold

Typegapthreshold

Below this % of avg = gap

number

default: 0.2 <= 1

minCooccurrence

Mincooccurrence

Min co-occurrence count to include a tag pair

integer

default: 2 >= 1 <= 20

topTagsLimit

Toptagslimit

Max top tags to return

integer

default: 30 >= 5 <= 200

topPairsLimit

Toppairslimit

Max co-occurring pairs to return

integer

default: 50 >= 5 <= 500

health

LoreMLSettingsHealth

Health scoring settings.

object

contradictionDecayRate

Contradictiondecayrate

Exponential decay rate for contradiction ratio. Higher = harsher.

number

default: 3 >= 0.1 <= 20

anomalyDecayRate

Anomalydecayrate

Exponential decay rate for anomaly ratio. Higher = harsher.

number

default: 2 >= 0.1 <= 20

consistencyWeight

Consistencyweight

Consistency dimension weight

number

default: 1 <= 5

depthWeight

Depthweight

Depth dimension weight

number

default: 1 <= 5

breadthWeight

Breadthweight

Breadth dimension weight

number

default: 1 <= 5

interconnectednessWeight

Interconnectednessweight

Interconnectedness dimension weight

number

default: 1 <= 5

maturityWeight

Maturityweight

Maturity dimension weight

number

default: 1 <= 5

coherenceWeight

Coherenceweight

KGE coherence dimension weight

number

default: 1 <= 5

coherenceTargetScore

Coherencetargetscore

Target avg_positive_score for 100% coherence (when MRR unavailable)

number

default: 5 >= 0.1 <= 50

semanticCoverageWeight

Semanticcoverageweight

Semantic connectivity weight in interconnectedness blend

number

default: 50

relationshipCoverageWeight

Relationshipcoverageweight

Relationship participation weight in interconnectedness blend

number

default: 30

fulfillmentWeight

Fulfillmentweight

Link fulfillment weight in interconnectedness blend

number

default: 20

entityImportance

LoreMLSettingsEntityImportance

Entity importance tier percentile thresholds.

object

sTierPercentile

Stierpercentile

number

default: 0.05 <= 1

aTierPercentile

Atierpercentile

number

default: 0.2 <= 1

bTierPercentile

Btierpercentile

number

default: 0.5 <= 1

tags

LoreMLSettingsTags

Tag analysis settings.

object

frequencyFloor

Frequencyfloor

Min tag frequency to be considered common

integer

default: 2 >= 1

tfidfMaxFeatures

Tfidfmaxfeatures

Max TF-IDF features for tag analysis

integer

default: 5000 >= 100

maxTagClusters

Maxtagclusters

Maximum tag clusters to discover

integer

default: 10 >= 2

validation

LoreMLSettingsValidation

Cross-reference validation settings.

object

secretLeakageWordThreshold

Secretleakagewordthreshold

Min shared secret words to flag leakage

integer

default: 3 >= 1

secretLeakageSimilarityThreshold

Secretleakagesimilaritythreshold

Embedding similarity threshold for semantic leakage

number

default: 0.85 <= 1

embedding

LoreMLSettingsEmbedding

Embedding/ChromaDB settings.

object

maxEmbedChars

Maxembedchars

Max characters sent to embedding model

integer

default: 6000 >= 500

kge

LoreMLSettingsKGE

Knowledge Graph Embedding (ComplEx) training settings.

object

embeddingDim

Embeddingdim

ComplEx embedding dimension

integer

default: 128 >= 16 <= 512

epochs

Epochs

Training epochs

integer

default: 200 >= 10 <= 2000

learningRate

Learningrate

Adam learning rate

number

default: 0.01 > 0 <= 1

negativeSamples

Negativesamples

Negative samples per positive triple

integer

default: 10 >= 1 <= 100

margin

Margin

Margin for margin-based loss

number

default: 1 > 0 <= 10

batchSize

Batchsize

Training batch size

integer

default: 256 >= 16 <= 4096

spectral

LoreMLSettingsSpectral

Spectral graph embedding settings.

object

nComponents

Ncomponents

Number of spectral dimensions

integer

default: 32 >= 4 <= 256

fusion

LoreMLSettingsFusion

Embedding fusion weight settings. Controls how semantic, KGE, and spectral signals blend.

object

semanticWeight

Semanticweight

Weight for semantic (ChromaDB) embeddings

number

default: 0.5 <= 1

kgeWeight

Kgeweight

Weight for KGE (ComplEx) embeddings

number

default: 0.3 <= 1

spectralWeight

Spectralweight

Weight for spectral graph embeddings

number

default: 0.2 <= 1

nli

LoreMLSettingsNLI

NLI (Natural Language Inference) contradiction detection settings.

object

contradictionThreshold

Contradictionthreshold

Min probability to flag

number

default: 0.5 <= 1

highConfidenceThreshold

Highconfidencethreshold

High-severity threshold

number

default: 0.8 <= 1

batchSize

Batchsize

NLI inference batch size

integer

default: 16 >= 1 <= 128

maxNliPairs

Maxnlipairs

Max candidate pairs to evaluate per run

integer

default: 2000 >= 100 <= 50000

classification

LoreMLSettingsClassification

Relationship type classification settings (Layer 2).

object

autoThreshold

Autothreshold

Auto-assign above this confidence

number

default: 0.8 <= 1

suggestThreshold

Suggestthreshold

Suggest above this confidence

number

default: 0.5 <= 1

enabled

Enabled

Enable relationship classification

boolean

default: true

propagation

LoreMLSettingsPropagation

Impact propagation settings (Layer 3).

object

maxDepth

Maxdepth

Max hops in impact chain

integer

default: 3 >= 1 <= 10

minStrength

Minstrength

Min strength to include in chain

number

default: 0.1 <= 1

strengthDecay

Strengthdecay

Strength multiplier per hop

number

default: 0.7 >= 0.1 <= 1

Responses

200

Successful Response

LoreMLSettings

Complete per-project ML settings. All values have sensible defaults.

object

pipeline

LoreMLSettingsPipeline

Pipeline-level training settings.

object

minEntriesForTraining

Minentriesfortraining

Min entries to enable ML training

integer

default: 5 >= 1

retrainGrowthThreshold

Retraingrowththreshold

Corpus growth % to trigger retrain

number

default: 0.1 >= 0.01 <= 1

anomaly

LoreMLSettingsAnomaly

Anomaly detection settings.

object

weights

LoreMLSettingsAnomalyWeights

Signal weights for multi-signal anomaly detection (must sum to ~1.0).

object

isolationForest

Isolationforest

number

default: 0.25 <= 1

orphan

Orphan

number

default: 0.1 <= 1

completeness

Completeness

number

default: 0.15 <= 1

tagIsolation

Tagisolation

number

default: 0.1 <= 1

sizeOutlier

Sizeoutlier

number

default: 0.1 <= 1

staleness

Staleness

number

default: 0.1 <= 1

kgeOutlier

Kgeoutlier

number

default: 0.1 <= 1

spectralIsolation

Spectralisolation

number

default: 0.1 <= 1

minEntries

Minentries

Minimum entries required to run anomaly detection

integer

default: 5 >= 2

reasonThreshold

Reasonthreshold

Min signal score to generate a reason string

number

default: 0.3 <= 1

logThreshold

Logthreshold

Composite score threshold for logging flagged anomalies

number

default: 0.5 <= 1

orphanDivisor

Orphandivisor

Connections needed for 0.0 orphan score

number

default: 10 >= 1

stalenessWindowDays

Stalenesswindowdays

Days behind sibling for max staleness score

integer

default: 180 >= 7

zScoreDivisor

Zscoredivisor

Z-score divisor for size outlier normalization

number

default: 3 >= 1

isolationForestContamination

Isolationforestcontamination

IsolationForest contamination parameter

number

default: 0.05 >= 0.01 <= 0.5

similarity

LoreMLSettingsSimilarity

Similarity detection settings.

object

suggestedLinkThreshold

Suggestedlinkthreshold

Min similarity for suggesting a link

number

default: 0.5 <= 1

nearDuplicateThreshold

Nearduplicatethreshold

Min similarity for near-duplicate detection

number

default: 0.85 <= 1

missingEntryMinMentions

Missingentryminmentions

Min mentions for a missing entry suggestion

integer

default: 3 >= 1

maxSuggestedLinks

Maxsuggestedlinks

Max suggested links returned

integer

default: 100 >= 10

maxNearDuplicates

Maxnearduplicates

Max near-duplicates returned

integer

default: 50 >= 10

maxMissingEntries

Maxmissingentries

Max missing entries returned

integer

default: 50 >= 10

clustering

LoreMLSettingsClustering

Clustering settings.

object

maxClusters

Maxclusters

Maximum clusters allowed

integer

default: 20 >= 2

minClusters

Minclusters

Minimum clusters floor

integer

default: 2 >= 2

tagTfidfMaxFeatures

Tagtfidfmaxfeatures

Max TF-IDF features for tag encoding

integer

default: 200 >= 10

maxCategories

Maxcategories

Max categories for one-hot encoding

integer

default: 50 >= 5

labelTfidfMaxFeatures

Labeltfidfmaxfeatures

Max TF-IDF features for cluster label generation

integer

default: 500 >= 50

labelMinTokenLength

Labelmintokenlength

Min characters for a word to appear in cluster labels

integer

default: 4 >= 2

contradiction

LoreMLSettingsContradiction

Contradiction detection settings.

object

minMentions

Minmentions

Min mentions for contradiction check

integer

default: 2 >= 2

similarityThreshold

Similaritythreshold

Snippet embedding similarity below this = potential contradiction

number

default: 0.45 <= 1

severityThreshold

Severitythreshold

Below this = medium severity (else low)

number

default: 0.3 <= 1

maxMentionsCompared

Maxmentionscompared

Max mentions compared per entity

integer

default: 50 >= 5

maxContradictionsTotal

Maxcontradictionstotal

Stop after finding this many

integer

default: 200 >= 10

maturity

LoreMLSettingsMaturity

Maturity tier thresholds (each 0.0-1.0).

object

completeThreshold

Completethreshold

number

default: 0.9 <= 1

matureThreshold

Maturethreshold

number

default: 0.75 <= 1

developingThreshold

Developingthreshold

number

default: 0.5 <= 1

draftThreshold

Draftthreshold

number

default: 0.25 <= 1

coverage

LoreMLSettingsCoverage

Coverage analysis settings.

object

typeGapThreshold

Typegapthreshold

Below this % of avg = gap

number

default: 0.2 <= 1

minCooccurrence

Mincooccurrence

Min co-occurrence count to include a tag pair

integer

default: 2 >= 1 <= 20

topTagsLimit

Toptagslimit

Max top tags to return

integer

default: 30 >= 5 <= 200

topPairsLimit

Toppairslimit

Max co-occurring pairs to return

integer

default: 50 >= 5 <= 500

health

LoreMLSettingsHealth

Health scoring settings.

object

contradictionDecayRate

Contradictiondecayrate

Exponential decay rate for contradiction ratio. Higher = harsher.

number

default: 3 >= 0.1 <= 20

anomalyDecayRate

Anomalydecayrate

Exponential decay rate for anomaly ratio. Higher = harsher.

number

default: 2 >= 0.1 <= 20

consistencyWeight

Consistencyweight

Consistency dimension weight

number

default: 1 <= 5

depthWeight

Depthweight

Depth dimension weight

number

default: 1 <= 5

breadthWeight

Breadthweight

Breadth dimension weight

number

default: 1 <= 5

interconnectednessWeight

Interconnectednessweight

Interconnectedness dimension weight

number

default: 1 <= 5

maturityWeight

Maturityweight

Maturity dimension weight

number

default: 1 <= 5

coherenceWeight

Coherenceweight

KGE coherence dimension weight

number

default: 1 <= 5

coherenceTargetScore

Coherencetargetscore

Target avg_positive_score for 100% coherence (when MRR unavailable)

number

default: 5 >= 0.1 <= 50

semanticCoverageWeight

Semanticcoverageweight

Semantic connectivity weight in interconnectedness blend

number

default: 50

relationshipCoverageWeight

Relationshipcoverageweight

Relationship participation weight in interconnectedness blend

number

default: 30

fulfillmentWeight

Fulfillmentweight

Link fulfillment weight in interconnectedness blend

number

default: 20

entityImportance

LoreMLSettingsEntityImportance

Entity importance tier percentile thresholds.

object

sTierPercentile

Stierpercentile

number

default: 0.05 <= 1

aTierPercentile

Atierpercentile

number

default: 0.2 <= 1

bTierPercentile

Btierpercentile

number

default: 0.5 <= 1

tags

LoreMLSettingsTags

Tag analysis settings.

object

frequencyFloor

Frequencyfloor

Min tag frequency to be considered common

integer

default: 2 >= 1

tfidfMaxFeatures

Tfidfmaxfeatures

Max TF-IDF features for tag analysis

integer

default: 5000 >= 100

maxTagClusters

Maxtagclusters

Maximum tag clusters to discover

integer

default: 10 >= 2

validation

LoreMLSettingsValidation

Cross-reference validation settings.

object

secretLeakageWordThreshold

Secretleakagewordthreshold

Min shared secret words to flag leakage

integer

default: 3 >= 1

secretLeakageSimilarityThreshold

Secretleakagesimilaritythreshold

Embedding similarity threshold for semantic leakage

number

default: 0.85 <= 1

embedding

LoreMLSettingsEmbedding

Embedding/ChromaDB settings.

object

maxEmbedChars

Maxembedchars

Max characters sent to embedding model

integer

default: 6000 >= 500

kge

LoreMLSettingsKGE

Knowledge Graph Embedding (ComplEx) training settings.

object

embeddingDim

Embeddingdim

ComplEx embedding dimension

integer

default: 128 >= 16 <= 512

epochs

Epochs

Training epochs

integer

default: 200 >= 10 <= 2000

learningRate

Learningrate

Adam learning rate

number

default: 0.01 > 0 <= 1

negativeSamples

Negativesamples

Negative samples per positive triple

integer

default: 10 >= 1 <= 100

margin

Margin

Margin for margin-based loss

number

default: 1 > 0 <= 10

batchSize

Batchsize

Training batch size

integer

default: 256 >= 16 <= 4096

spectral

LoreMLSettingsSpectral

Spectral graph embedding settings.

object

nComponents

Ncomponents

Number of spectral dimensions

integer

default: 32 >= 4 <= 256

fusion

LoreMLSettingsFusion

Embedding fusion weight settings. Controls how semantic, KGE, and spectral signals blend.

object

semanticWeight

Semanticweight

Weight for semantic (ChromaDB) embeddings

number

default: 0.5 <= 1

kgeWeight

Kgeweight

Weight for KGE (ComplEx) embeddings

number

default: 0.3 <= 1

spectralWeight

Spectralweight

Weight for spectral graph embeddings

number

default: 0.2 <= 1

nli

LoreMLSettingsNLI

NLI (Natural Language Inference) contradiction detection settings.

object

contradictionThreshold

Contradictionthreshold

Min probability to flag

number

default: 0.5 <= 1

highConfidenceThreshold

Highconfidencethreshold

High-severity threshold

number

default: 0.8 <= 1

batchSize

Batchsize

NLI inference batch size

integer

default: 16 >= 1 <= 128

maxNliPairs

Maxnlipairs

Max candidate pairs to evaluate per run

integer

default: 2000 >= 100 <= 50000

classification

LoreMLSettingsClassification

Relationship type classification settings (Layer 2).

object

autoThreshold

Autothreshold

Auto-assign above this confidence

number

default: 0.8 <= 1

suggestThreshold

Suggestthreshold

Suggest above this confidence

number

default: 0.5 <= 1

enabled

Enabled

Enable relationship classification

boolean

default: true

propagation

LoreMLSettingsPropagation

Impact propagation settings (Layer 3).

object

maxDepth

Maxdepth

Max hops in impact chain

integer

default: 3 >= 1 <= 10

minStrength

Minstrength

Min strength to include in chain

number

default: 0.1 <= 1

strengthDecay

Strengthdecay

Strength multiplier per hop

number

default: 0.7 >= 0.1 <= 1

422

Validation Error

HTTPValidationError

object

detail

Detail

Array<object>

ValidationError

object

loc

required

Location

Array

msg

required

Message

string

type

required

Error Type

string

input

Input

ctx

Context

object