> ## Documentation Index
> Fetch the complete documentation index at: https://docs.memloom.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Graph & embeddings

> index, reembed, auto-index, schema: building the graph and keeping vectors consistent

## index

```bash theme={null}
memloom index [--rebuild]
```

Extract entities from unindexed memories and context chunks into the graph (one LLM call per
item; needs an API key). Prints one line per item with the entities found. With auto-index on
(the default in cloud mode) new items are indexed in the background, so this command usually
reports nothing pending.

`--rebuild` wipes extracted entities and their edges (mentions and relationships found in
your content), then re-extracts from scratch. It does not touch memories, conflicts, or the
`replaces`/`distinct` edges created by resolving conflicts.

## reembed

```bash theme={null}
memloom reembed [--force]
```

Recompute every stored embedding with the embedding provider currently configured in
`~/.memloom/config.env`, then stamp the store with the new fingerprint. Run this after
switching providers or models, for example after adding `OPENROUTER_API_KEY` to leave
offline mode: without it the daemon refuses to start, because old and new vectors live in
incompatible spaces.

<Warning>
  The daemon must be stopped first (`memloom stop`); reembed opens the store directly.
</Warning>

Interrupting is safe: memories and files are never touched, only their vectors, and running
the command again resumes where it stopped. Costs one embedding API call per 64 items.

`--force` re-embeds even when the store already matches the configured provider and nothing
is missing.

A typical offline-to-cloud migration:

```bash theme={null}
memloom stop
# set OPENROUTER_API_KEY in ~/.memloom/config.env
memloom reembed
memloom serve      # or any command; then:
memloom index      # build the graph (offline mode couldn't)
```

## auto-index

```bash theme={null}
memloom auto-index [on|off]
```

Show or set background entity extraction. When on, new memories and files are indexed a few
seconds after they land (debounced into batched runs, visible in the Console). The setting
persists across daemon restarts; `MEMLOOM_AUTO_INDEX` in config.env is only the default
before the first use of this switch. Needs an API key; offline mode cannot enable it.

```bash theme={null}
memloom auto-index        # show the current state
memloom auto-index off    # index only via 'memloom index' / the Console
```

## schema

```bash theme={null}
memloom schema [list|disable|enable|delete]
```

Manage the extraction vocabulary: the entity types and predicates the LLM is allowed to use
when indexing. With no arguments, prints the vocabulary with usage counts and status.

| Subcommand                                       | Effect                                                                                                                                            |
| ------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------- |
| `schema disable <entity_type\|predicate> <name>` | Stop using an entry for future extraction. Entities already extracted under it stay in the graph. Works on built-ins and user-tier entries alike. |
| `schema enable <entity_type\|predicate> <name>`  | Re-enable a disabled entry.                                                                                                                       |
| `schema delete <entity_type\|predicate> <name>`  | Permanently remove a DISABLED user-tier entry. Built-ins can only be disabled, and an active entry must be disabled first.                        |

The vocabulary itself, and how the LLM proposes new entries for your approval, is covered in
[The graph](/concepts/graph).