Releases: awaescher/OllamaSharp
Release 5.2.9
Release 5.2.8
Improved the source code generator for Ollama tools:
- supports line breaks in summaries (#251)
- supports tools without namespace (#262)
- won't cause warnings for missing XML documentation of the generator tool class (#263)
Recap: Providing tools is as simple as implementing a method and decorating it with the OllamaTool attribute.
/// <summary>
/// Get the current weather for a city
/// </summary>
/// <param name="city">Name of the city</param>
[OllamaTool]
public static string GetWeather(string city) => $"It's cold at only 6°C in {city}.";Make sure to add
<GenerateDocumentationFile>true</GenerateDocumentationFile>to your project file to include the XML documentation when building.
See tool support for more information.
Release 5.2.7
Improvements to the Chat class
- Adds support for multiple tool calls
- Adds an event
OnThinkthat gets fired for thinking tokens ifThinkis set totrue - Adds an event
OnToolCallthat gets fired once the AI model wants to call a tool - Adds an event
OnToolResultthat gets fired once a result is available for a tool call that the AI model requested
Demo Console:
- Prints tool calls as they happen in between messages from the AI model
Release 5.2.6
The Chat class can now yield thoughts over OnThink(string) when thinking is allowed (Think=true) #273.
You can easily experiment with this in the console demos by using the command /togglethink that switches the think mode from null → false → true → null ...
Release 5.2.5
Release 5.2.4
Release 5.2.3
Release 5.2.2
→ learn more about think in release 5.2.0
Release 5.2.1
- Support
thinkwith Chat #255
→ learn more about think in release 5.2.0
Release 5.2.0
Important notes
The Think parameter controls whether the models are allowed to reason and how the reasoning tokens are handled.
| Reasoning model | Marked "thinking" in Ollama library | Think=false |
Think=true |
|---|---|---|---|
| Yes | Yes | Disables reasoning | Enables reasoning + Separates reasoning tokens in Message.Thinking |
| Yes | No | Reasoning stays enabled + Reasoning tags will appear as <think></think> in the models response |
❌ Error: Model does not support thinking → Download the model again |
| No | No | No effect | ❌ Error: Model does not support thinking |
