Advanced Search

Personalized Learned Fusion

Uses Thompson Sampling to learn optimal feature weights from user interactions. Learns which features (text, image, etc.) matter most for each user/context.

1 stage

Read Docs

Integrate Now

No credit card to start · Build from $25/mo

Pipeline Stages

learned_fusion_search

filter

feature_search

View parameters

{
  "searches": [
    {
      "feature_uri": "mixpeek://text_extractor@v1/multilingual_e5_large_instruct_v1",
      "query": {
        "input_mode": "text",
        "value": "{{INPUT.query}}"
      },
      "top_k": 100
    },
    {
      "feature_uri": "mixpeek://multimodal_extractor@v1/vertex_multimodal_embedding",
      "query": {
        "input_mode": "text",
        "value": "{{INPUT.query}}"
      },
      "top_k": 100
    }
  ],
  "final_top_k": 25,
  "fusion": "learned",
  "learning_config": {
    "context_features": [
      "INPUT.user_id"
    ],
    "demographic_features": [
      "INPUT.user_segment"
    ],
    "fallback_strategy": "hierarchical",
    "min_interactions": 5,
    "reward_signal": "click"
  }
}

Input Schema

query

text

required

Search query

user_id

string

User ID for personalization

user_segment

string

User segment for cold start fallback

Use Cases

Personalized search experiences
Learning feature importance over time
Adaptive multimodal search
A/B testing feature weights

Requirements

Collection must have multiple feature types indexed
Interactions API must be called for feedback

Ready to use this template?

Clone it into your namespace and start querying in minutes.