ChatClientEnsembleMember<T>
Models & Types · AiDotNet.Agentic.Pipeline
One model in a weighted ensemble: the client, how often it is picked, and its own default settings.
For Beginners
This describes one model in a group of models. The weight decides how often it gets picked: a member with weight 3 is chosen about three times as often as one with weight 1. You can also give it its own temperature or token limit, which is used whenever that member answers.
How It Works
Mixing models during a search is a real strategy rather than a convenience: a strong, expensive model produces the occasional large improvement while a cheap one produces most of the volume, and the weight is the dial between them. Attaching settings to the member rather than to the call is what lets one member run hot and another run cold within a single run.
Name is what appears in selection statistics and in the response’s model id, so two members that
wrap the same underlying model with different settings remain distinguishable in the record — which they are
not upstream, where a member is identified only by its model string.
Constructors
| Constructor | Summary |
|---|---|
ChatClientEnsembleMember(IChatClient<T> client, double weight, ChatOptions chatOptions, string name) |
Initializes an ensemble member. |
Properties
| Property | Summary |
|---|---|
ChatOptions ChatOptions { get; } |
Gets this member’s default settings, or null when it has none. |
IChatClient<T> Client { get; } |
Gets the chat client this member calls. |
double Weight { get; } |
Gets the relative selection weight. |
string Name { get; } |
Gets the stable name used in statistics and in the reported response model id. |
Methods
| Method | Summary |
|---|---|
string ToString() |
Returns the member’s name and weight. |