Explorer

What does the Explorer view do?

The Explorer view provides a tabbed interface for browsing, searching, and visualizing grains stored in the Areev context database. It is accessible at /memories/:memoryId/grains from the sidebar under the Explore group.

The Explorer has four tabs: Grains (a searchable, filterable table of all grains), Graph (an interactive knowledge graph visualization of entity relationships), Timeline (a chronological view of grain creation), and Import (drag-and-drop document upload — see the Ingest page for details).

The Grains tab displays a paginated table of AI memory grains showing subject, grain type, creation timestamp, and a content preview. The search bar at the top executes a Tantivy BM25 full-text query across grain content and metadata fields. Filter controls let you narrow results by grain type (belief, event, state, workflow, action, observation, goal, reasoning, consensus, consent). Results update as you type with debounced queries to the /api/memories/{memory_id}/recall endpoint. When no search query is active, grains are listed via the /api/memories/{memory_id}/grains endpoint with pagination and sorting.

Clicking any row opens a detail drawer on the right side of the screen. The drawer shows the grain’s full content, all metadata key-value pairs, the provenance chain, the supersession history, any contradictions, feedback history, access log, and audit entries. The autonomous memory engine tracks every supersession as a versioned event, so you can inspect the full lineage of any grain from creation through supersessions.

How do I search for specific grains?

Type a query into the search bar to execute a BM25 full-text search across all grain content and metadata, or use the grain type filter to narrow results.

The Explorer delegates search to the Areev server’s /api/memories/{memory_id}/recall endpoint, which runs the query through the Tantivy full-text index. The AI agent memory engine ranks results by BM25 relevance combined with vector similarity (hybrid search), and the Explorer displays them in ranked order.

The grain type dropdown offers the ten OMS 1.2 types. Combining a text query with a type filter creates an intersection — only grains matching both the text query and the selected type appear in results.

# Equivalent API call for searching grains (replace {memory_id} with your memory ID)
curl -s -X POST http://localhost:4009/api/memories/{memory_id}/recall \
  -H "Content-Type: application/json" \
  -d '{"query": "user preferences", "grain_type": "belief"}'
# Python SDK equivalent
import requests
memory_id = "my-memory"
resp = requests.post(f"http://localhost:4009/api/memories/{memory_id}/recall", json={
    "query": "user preferences",
    "grain_type": "belief"
})
grains = resp.json()["results"]

How do I inspect a grain’s details?

Click any row in the Explorer table to open the detail drawer, which displays the grain’s content, metadata, provenance, supersession history, and more.

The detail drawer loads comprehensive grain information via the /api/memories/{memory_id}/grains/{hash} endpoint. This returns the grain’s full fields, provenance chain, supersession chain, contradictions, feedback history, access log, consolidation path, and audit entries — all in a single response.

The drawer supports inline supersession: you can edit a grain’s value fields directly and submit a corrected version. Identity fields (like subject, relation, tool name) are shown read-only, while value fields (like object, content, conclusion) are editable. Certain grain types display contextual banners — for example, consent grains warn that supersession affects data processing legality.

The supersession chain shows the full lineage of the grain from creation through its current version, and contradiction detection highlights conflicting grains on the same subject.