Retrieval Mechanisms

Reeflect provides sophisticated retrieval mechanisms to find the most relevant memories for any context.

Retrieval Methods

Method	Description	Use Cases
retrieve()	Direct retrieval by ID	When you know the exact memory to retrieve
query()	Filtered query by parameters	Searching by namespace, tags, or other filters
search()	Semantic similarity search	Finding memories similar to a query text
enhance_prompt()	Injecting memories into a prompt	Adding context to LLM queries
reason()	Reasoning using relevant memories	Generating insights from memory corpus

Semantic Search

The most powerful retrieval method is semantic search, which finds memories based on meaning rather than exact matches:

# Simple semantic search
results = memory.search(
    query="What are the user's color preferences?",
    namespace="user_preferences",
    limit=5
)

# Advanced semantic search with filters
results = memory.search(
    query="What are the user's food allergies?",
    namespace="user_health",
    filter_params={
        "min_importance": 0.7,
        "tags": ["allergy", "medical"],
        "created_after": datetime(2023, 1, 1)
    },
    limit=10,
    min_similarity=0.7
)

Relevance Scoring

Reeflect uses a sophisticated relevance scoring system that combines multiple factors:

score = (
    semantic_similarity * 0.6 +
    memory.importance * 0.2 +
    recency_score * 0.1 +
    usage_score * 0.1
)

You can customize the relevance weights to prioritize different factors:

from reeflect.core.retrieval import RelevanceScorer

# Rerank search results with custom weights
custom_results = RelevanceScorer.rerank_results(
    results,
    weights={
        "similarity": 0.4,
        "importance": 0.3,
        "recency": 0.2,
        "usage": 0.1
    }
)

Next Steps

Now that you understand how to retrieve memories, learn about Model Adapters to see how Reeflect integrates with various LLM providers.