Feather DB

The Living context Engine

Enter the API key to continue.

esc
Type to search. Use ns:name for namespace, id:123 to open a record, /seed for actions.

Overview

Aggregate state across every namespace on this server.

Ops · last hour
errors:

Top namespaces

View all →
Name Records Dim Modalities
No namespaces yet.

Recent activity

No activity yet.

Namespaces

All `.feather` files on this server.

Name Records Dim Modalities
No namespaces match.

records · modalities:
Showing records (page size 200)
ID Content namespace_id entity_id Importance Recall
No records.
End of list
keys:
ID Score Content namespace_id
Run a search to see results.
Distinct attributes seen across live records. Drives the filter builder.
Walks every live record once to tally attribute keys + value samples.
Scanning…
Records
namespace_id values
none
Sources
none

Attributes

Key Type Count Distinct Samples
No attributes found.
Built from attributes.brand, .channel, .campaign, .adset, .ad, .creative on each record.
Building…
Force-directed graph of typed edges. Drag nodes; click to inspect.
Showing nodes / edges.
context_chain runs vector search for the top k hits, then BFS-expands through typed graph edges for hops steps. Returns a connected sub-graph rooted at the search hits.
ID Score Hop Content
Run a context chain to see expanded nodes.
Edges traversed
Index
Namespaces · secondary index
Auto-compaction
Rebuild the index automatically once deleted/total crosses the threshold. 0 disables.
saving…
On-disk int8 quantization
Persist vectors as int8 — ~3× smaller .feather, dequantized to float32 on load. Applies on next save.
saving…
Compact now
Reclaim space from forgotten/deleted vectors and prune dead edges.
Bulk delete
Delete many records in one call (one save). By record ids and/or a whole entity_id. Use this instead of deleting records one by one.
deleting…
Response

              

Search

Pick a namespace to run a search inside. Or press ⌘K for the quick palette.

Settings

Server info and dashboard configuration.

Server
URL
Version
Namespaces
Total records
Default dim
Uptime
Connection
Base URL
Auth header
X-API-Key: feather-…
cURL

                
              
Python (requests)

                
              
JavaScript (fetch)

                
              
Open OpenAPI docs ↗
Embedding service

When configured, "Ingest text" produces real semantic vectors via the chosen provider. Random vectors otherwise.

Must match your namespace's dim. Vectors are padded/truncated if not.
API key

Stored locally in your browser.

Danger zone
Sign out
Clears the API key from this browser.
Record
Marketing profile

        
Outgoing
no outgoing edges.
Incoming
no incoming edges.
Add edge from this record

Create namespace

A new .feather file will be created on disk.

letters · numbers · underscore · dash only

Add record

Vector is randomly generated for now. Wire an embedding service later for real semantic search.

Ingest text

Server embeds via the configured provider, then stores the record.

Embedding service not configured. Set one in Settings first.

Bulk import

Paste a JSON array of {id, vector?, metadata} records, or upload a .json file. Omit vector to auto-embed metadata.content with the configured provider.

supplied vectors must match this namespace's dim (); records without a vector are embedded server-side

Bulk seed

Generates N random vectors with templated content. Server-side; instant.

Use {i} as a counter token.