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

# Changelogs

## v0.0.18

Currently under beta testing:

* 0.0.18-beta1

### Model Chaining

Model chaining allows you to chain models together so that the output of one model can be the input to another model. For all the models in the chain, all the request and entities in the model input are the same.

ModelChainComponent

#### ChainedModelInput

This is the input for any models within a model chain. It's a child class of the
[ModelInput](/model_service#modelinput) and has two additional properties:

* `upstream_model_output`: the output of the upstream model
* `upstream_model_name`: the name of the upstream model

### Support dict as model output

Starting from v0.0.18, Wyvern will support **serializable** dictionary as model output.

If the model output is a dictionary, wyvern will generate the same number of
model events as the number of keys in the dictionary. For each
model event ([ModelEventData](/sdk_ref#modeleventdata-objects)), a new
optional property `model_key` is added to reprent the key of in the
dictionary. The `ModelEventData.model_output` will be the value of the key in the
model output.

### Model output cache

Now ModelComponent initialization can take `cache_output` as an optional
input. If `cache_output` is set to `True`, the model output will be cached.
The default value is True.

WyvernRequest has a new global map `model_output_map` which is used to
cache model output. The key of the map is the model name. The value is another
map (the inner map). They key of the inner map is the entity identifier and
the value is the model output value.

#### Component.get\_model\_output

Within any component, now you can use `self.get_model_output` to get the
model cached model output.