<!-- generated by epythet -->

# chromadol

Data Object Layer for ChromaDB

To install:	`pip install chromadol`

[Documentation](https://i2mint.github.io/chromadol/)

# Example usage

To make a `ChromaClient` DOL, you can specify a `chromadb` `Client`, `PersistentClient` (etc.)
instance, or specify a string (which will be interpreted as a path to a directory to
save the data to in a `PersistentClient` instance).

```none
>>> from chromadol import ChromaClient
>>> import tempfile, os 
>>> with tempfile.TemporaryDirectory() as temp_dir:
...     tempdir = os.path.join(temp_dir, "chromadol_test")
...     os.makedirs(tempdir)
>>> client = ChromaClient(tempdir)
```

Removing all contents of client to be able to run a test on a clean slate

```none
>>> for k in client:
...     del client[k]
```

There’s nothing yet:

```none
>>> list(client)
[]
```

Now let’s “get” a collection.

```none
>>> collection = client['chromadol_test']
```

Note that just accessing the collection creates it (by default)

```none
>>> list(client)
['chromadol_test']
```

Here’s nothing in the collection yet:

```none
>>> list(collection)
[]
```

So let’s write something.
Note that `chromadb` is designed to operate on multiple documents at once,
so the “chromadb-natural” way of specifying it’s keys and contents (and any extras)
would be like this:

```none
>>> collection[['piece', 'of']] = {
...     'documents': ['contents for piece', 'contents for of'],
...     'metadatas': [{'author': 'me'}, {'author': 'you'}],
... }
```

Now we have two documents in the collection:

```none
>>> len(collection)
2
```

Note, though, that the order of the documents is not guaranteed.

```none
>>> sorted(collection)
['of', 'piece']

>>> assert collection['piece'] == {
...     'ids': ['piece'],
...     'embeddings': None,
...     'metadatas': [{'author': 'me'}],
...     'documents': ['contents for piece'],
...     'uris': None,
...     'data': None
... }

>>> assert collection['of'] == {
...     'ids': ['of'],
...     'embeddings': None,
...     'metadatas': [{'author': 'you'}],
...     'documents': ['contents for of'],
...     'uris': None,
...     'data': None
... }
```

You can also read multiple documents at once.
But note that the order of the documents is not guaranteed.

```none
>>> collection[['piece', 'of']] == collection[['of', 'piece']]
True
```

You can read or write one document at a time too.

```none
>>> collection['cake'] = {
...     "documents": "contents for cake",
... }
>>> sorted(collection)  # sorting because order is not guaranteed
['cake', 'of', 'piece']
>>> assert collection['cake'] == {
...     'ids': ['cake'],
...     'embeddings': None,
...     'metadatas': [None],
...     'documents': ['contents for cake'],
...     'uris': None,
...     'data': None,
... }
```

In fact, see that if you only want to specify the “documents” part of the information,
you can just write a string instead of a dictionary:

```none
>>> collection['cake'] = 'a different cake'
>>> assert collection['cake'] == {
...     'ids': ['cake'],
...     'embeddings': None,
...     'metadatas': [None],
...     'documents': ['a different cake'],
...     'uris': None,
...     'data': None,
... }
```

<p class="epythet-aggregates">This documentation as a single file: <a href="chromadol.md">chromadol.md</a> (Markdown, for agents).</p>