Recommendations

Coming soon (Phase 7)

Personalized content discovery, similar items, and trend detection. Not yet callable — returns a gated error today.

Not yet callable

Recommendations capabilities are planned for Phase 7 of the platform roadmap, alongside moderation. They depend on the RAG subsystem (Phase 5) being stable first.

These capabilities are unpublished today — calling them returns 400 (only published capabilities can execute). Do not depend on these endpoints in production code yet.

Intended Capabilities (Phase 7)

SlugWhat it will do
data-enrichment:contextual-rerankRerank search results by contextual relevance to a user or session
data-enrichment:cooking-video-intelligenceExtract structured metadata from cooking video content

Additional personalized feed and trend detection capabilities will be scoped during Phase 7 planning.

Intended API Shape

When live, recommendation capabilities will follow the standard capability execution pattern.

JavaScript SDK (future)

// NOT callable yet — returns 403 today
const run = await client.capabilities.run('data-enrichment:contextual-rerank', {
  query: 'gluten-free pasta dinner',
  candidates: [
    { id: 'item_1', text: 'Classic carbonara recipe' },
    { id: 'item_2', text: 'Zucchini noodle bolognese' },
  ],
})

cURL (future)

# NOT callable yet — returns 403 today
curl -X POST https://api.whizur.ai/v1/capabilities/data-enrichment:contextual-rerank/execute \
  -H "X-API-Key: cw_dev_..." \
  -H "Content-Type: application/json" \
  -d '{
    "input": {
      "query": "gluten-free pasta dinner",
      "candidates": [
        { "id": "item_1", "text": "Classic carbonara recipe" },
        { "id": "item_2", "text": "Zucchini noodle bolognese" }
      ]
    }
  }'

Intended input shape — data-enrichment:contextual-rerank

{
  "query": string,           // user query or preference signal
  "candidates": [
    { "id": string, "text": string }
  ],
  "topK"?: number            // default: return all, ranked
}

Intended output artifact

{
  "ranked": [
    { "id": string, "score": number }
  ]
}

Why Phase 7?

Recommendation quality depends on vector embeddings from the RAG subsystem (Phase 5). Phase 7 builds personalization on top of that stable foundation. See context/architecture/launch-scope.md for the full phase plan.

What is Live Now