/**
* Recommended scoring configuration presets
*
* These presets provide tuned weight configurations for different use cases.
* Users can select a preset or use it as a starting point for customization.
*/
import type { ScoringConfig } from "./types";
/**
* Balanced preset (default)
* - Equal emphasis on semantic similarity and contextual factors
* - Good for general-purpose memory retrieval
*/
export const BALANCED_PRESET: Partial<ScoringConfig> = {
weights: {
semantic: 0.5,
recency: 0.2,
confidence: 0.15,
frequency: 0.15,
},
boostRules: [
{
name: "recent-boost",
condition: { type: "recency", maxDays: 7 },
factor: 1.2,
},
{
name: "high-importance-boost",
condition: { type: "importance", minValue: 0.8 },
factor: 1.15,
},
{
name: "stale-penalty",
condition: { type: "staleness", minDays: 180 },
factor: 0.8,
},
],
};
/**
* Semantic-focused preset
* - Prioritizes semantic similarity over other factors
* - Best for: Finding conceptually related memories regardless of age
* - Use case: Research, exploring related concepts, finding solutions
*/
export const SEMANTIC_FOCUSED_PRESET: Partial<ScoringConfig> = {
weights: {
semantic: 0.7,
recency: 0.1,
confidence: 0.1,
frequency: 0.1,
},
boostRules: [
{
name: "high-importance-boost",
condition: { type: "importance", minValue: 0.8 },
factor: 1.2,
},
],
};
/**
* Recency-focused preset
* - Prioritizes recently accessed/created memories
* - Best for: Active development sessions, current context awareness
* - Use case: Working on recent features, debugging current issues
*/
export const RECENCY_FOCUSED_PRESET: Partial<ScoringConfig> = {
weights: {
semantic: 0.3,
recency: 0.5,
confidence: 0.1,
frequency: 0.1,
},
recencyDecay: {
type: "exponential",
halfLifeDays: 14, // Shorter half-life for faster decay
},
boostRules: [
{
name: "very-recent-boost",
condition: { type: "recency", maxDays: 3 },
factor: 1.3,
},
{
name: "recent-boost",
condition: { type: "recency", maxDays: 7 },
factor: 1.15,
},
{
name: "stale-penalty",
condition: { type: "staleness", minDays: 90 },
factor: 0.7,
},
],
};
/**
* Frequency-focused preset
* - Prioritizes frequently accessed memories
* - Best for: Finding commonly referenced patterns and decisions
* - Use case: Discovering team conventions, frequently used solutions
*/
export const FREQUENCY_FOCUSED_PRESET: Partial<ScoringConfig> = {
weights: {
semantic: 0.35,
recency: 0.15,
confidence: 0.15,
frequency: 0.35,
},
frequencyNormalization: {
method: "log",
maxCount: 50, // Lower threshold for "frequently accessed"
coldStartScore: 0.3, // Lower score for new items
},
boostRules: [
{
name: "high-frequency-boost",
condition: { type: "frequency", minAccessCount: 10 },
factor: 1.25,
},
{
name: "high-importance-boost",
condition: { type: "importance", minValue: 0.8 },
factor: 1.1,
},
],
};
/**
* Architecture-focused preset
* - Prioritizes architectural decisions and patterns
* - Best for: Understanding system design and high-level decisions
* - Use case: Onboarding, architecture reviews, design discussions
*/
export const ARCHITECTURE_FOCUSED_PRESET: Partial<ScoringConfig> = {
weights: {
semantic: 0.45,
recency: 0.15,
confidence: 0.25,
frequency: 0.15,
},
boostRules: [
{
name: "architecture-boost",
condition: { type: "memoryType", types: ["architecture", "decision"] },
factor: 1.3,
},
{
name: "pattern-boost",
condition: { type: "memoryType", types: ["pattern"] },
factor: 1.2,
},
{
name: "high-importance-boost",
condition: { type: "importance", minValue: 0.7 },
factor: 1.15,
},
],
};
/**
* All available presets mapped by name
*/
export const SCORING_PRESETS = {
balanced: BALANCED_PRESET,
"semantic-focused": SEMANTIC_FOCUSED_PRESET,
"recency-focused": RECENCY_FOCUSED_PRESET,
"frequency-focused": FREQUENCY_FOCUSED_PRESET,
"architecture-focused": ARCHITECTURE_FOCUSED_PRESET,
} as const;
export type ScoringPresetName = keyof typeof SCORING_PRESETS;
/**
* Get a scoring config preset by name
*/
export function getScoringPreset(
name: ScoringPresetName,
): Partial<ScoringConfig> {
return SCORING_PRESETS[name];
}
/**
* List all available preset names
*/
export function listScoringPresets(): ScoringPresetName[] {
return Object.keys(SCORING_PRESETS) as ScoringPresetName[];
}
