One table for every site, sliced well.

teleon.dev/docs/partitions stores rows from every site in one partitioned table with faceted tags, so cross-site queries stay simple and each site's slice stays fast.

teleon.dev/docs/partitionsPartitioned rowshttpsjsonmcp

Read the docs Copy the agent prompt

One table, sliced by site

files_site

skill

template

dataset

media_site

scene

motion

cause_site

brief

open file

Each site's rows live in their own partition of the same table, tagged with facets.

What it does

One table, many sites

Rows carry a site facet in place of separate tables per site.

Faceted tags

Tags and LSH bands sit in an inverted index for quick filters.

Seams covered

Overlapping band sizes keep near-boundary items findable.

The contract

Rows are upserted with facets; filters read the facet index first.

Tableteleon_canonical_rows

Indexteleon_canonical_facets

BandsSmall, medium and large, with shifted buckets

AccessRING-2

Identifiers

teleon_canonical_partitioned/v1
teleon_multiscale_lsh_projection/v2

Execution contracts

Multi-scale and seam-shifted LSH with the partitioned core

Request and responseCopy

POST /v1/partitions/filter
{
  "facets": {"site": "files_site", "kind": "skill"},
  "limit": 2
}

# Response
{
  "rows": [
    {"id": "row_88", "site": "files_site", "kind": "skill"},
    {"id": "row_91", "site": "files_site", "kind": "skill"}
  ]
}

Endpoints

Method	Path	Access	What it does	Status
`GET`	`/v1/partitions`	RING-2	List partitions and sizes	Listed
`POST`	`/v1/partitions/filter`	RING-2	Filter rows by facets
`PUT`	`/v1/partitions/rows`	RING-2	Upsert a row with facets

Listed paths appear in the service’s published files today. In the one-host plan every path here is served under api.teleon.dev/v1/partitions.

How it connects

Reads from

Database and graph coreteleon.dev/docs/db

Feeds into

Searchteleon.dev/docs/search Recordsteleon.dev/docs/records

Queuesaq70 aq71

StoresS18 Unified Partitioned Core Rows and LSH Bands

For agents

Give your agent the read-only guide first. It reports back before any write.

Agent prompt

Read https://teleon.dev/docs/partitions.md and filter rows for one site. Read only.

Copy

llms.txt Agent card MCP card

Words on this page

Full glossary

Partition: One slice of a large table, chosen by a key such as site or queue, so reads and writes stay fast.
LSH: Locality-sensitive hashing: near-duplicate items land in the same bucket, so similar things are found fast.
Alpha-cut: A threshold on a continuous score. Items above the cut join a bitmap, so filters stay fast while scores stay continuous.