4.4 Get Recommendations

Leora offers two distinct recommendation modes:

4.4.1 User-Based Recommendations

Retrieve personalized suggestions based solely on a user’s historical interactions and profile features.

Endpoint

Query Parameters

• user_id (string, required)

• top_k (integer, default: 10)

• filters (JSON string, optional)

• explanations (boolean, default false)

Response

{
  "user_id": "U1234",
  "recommendations": [
    {"item_id": "I5678", "score": 0.982},
    {"item_id": "I9101", "score": 0.965}
  ],
  "explanations": [
    {"item_id":"I5678","tokens":["rock","guitar"]

• Use case: Known users with established history; leverages full profile and long-term behavior.

4.4.2 Session-Based Recommendations

Generate recommendations based on the current session’s activity sequence, ideal for anonymous or new users.

Endpoint

Query Parameters

• session (array of item IDs, required)

• top_k (integer, default: 10)

• filters (JSON string, optional)

• explanations (boolean, default false)

Response

{
  "session": ["I1001","I1002","I1003"],
  "recommendations": [
    {"item_id":"I2045","score":0.954},
    {"item_id":"I3098","score":0.942}
  ],
  "explanations": [
    {"item_id":"I2045","tokens":["jazz","live"]

• Use case: Anonymous browsing, new visitors without saved history, or context-driven sessions.

Performance: Both endpoints maintain 50–200 ms (95th percentile) latency with standard quotas (1000 requests/minute).