feat: Tool-Use functions call

CHANGELOG.md

@@ -1,3 +1,61 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+## [0.8.0-rc.2](https://github.com/inference-gateway/rust-sdk/compare/0.8.0-rc.1...0.8.0-rc.2) (2025-02-10)
+
+### ♻️ Improvements
+
+* Implement Default trait for Message and MessageRole structs ([cb43ca8](https://github.com/inference-gateway/rust-sdk/commit/cb43ca8034cec51bb4037e21dd1be9a8ce7312f3))
+* Initialize the client with tools optionally ([9c56170](https://github.com/inference-gateway/rust-sdk/commit/9c56170a17939b7ecf8a71ae04425313cf97727c))
+* Use in the tests two different conventions to see what's possible ([eef16e8](https://github.com/inference-gateway/rust-sdk/commit/eef16e8afedd2b28a2a54aa25f2364adfe11973b))
+
+### 📚 Documentation
+
+* Enhance InferenceGatewayAPI docblocks with additional error handling and return details ([b666c33](https://github.com/inference-gateway/rust-sdk/commit/b666c3338d81e193396d43d927a5b09569f70752))
+* Improve the docs for Tool-Use, update it ([9e5f5da](https://github.com/inference-gateway/rust-sdk/commit/9e5f5da14120b553ef045502dd704be5dcccd67f))
+* Improve the example by mentioning the workflow of messages exchange ([ef474b4](https://github.com/inference-gateway/rust-sdk/commit/ef474b427fd28527b589d7ff966acda4ea1d3cec))
+* Remove redundant whitespace in InferenceGatewayAPI documentation ([422e857](https://github.com/inference-gateway/rust-sdk/commit/422e8571c862547bf78aa2415528bf0814084319))
+* Resort functions examples, healthcheck is less important so put it at the bottom ([e6eb55d](https://github.com/inference-gateway/rust-sdk/commit/e6eb55d89a32a39603496303ed2b4d05b8bbf885))
+* Update the example in the README.md ([0fda0dc](https://github.com/inference-gateway/rust-sdk/commit/0fda0dc4dbed741dbfe8b62197ed6aea43e6b9e5))
+
+### ✅ Miscellaneous
+
+* Add assertion for tool_calls length in response test ([4edcd4b](https://github.com/inference-gateway/rust-sdk/commit/4edcd4b43e64bca6a1ac1856e8e57c4eabd0e0a8))
+
+## [0.8.0-rc.1](https://github.com/inference-gateway/rust-sdk/compare/0.7.1-rc.3...0.8.0-rc.1) (2025-02-09)
+
+### ✨ Features
+
+* Add optional tool_call_id to Message struct and update serialization tests ([d51e911](https://github.com/inference-gateway/rust-sdk/commit/d51e9115bbcd4fb5f098e99dfda9f11c00febf07))
+
+## [0.7.1-rc.3](https://github.com/inference-gateway/rust-sdk/compare/0.7.1-rc.2...0.7.1-rc.3) (2025-02-09)
+
+### ♻️ Improvements
+
+* Remove ToolParameterType and ToolParameter structs to simplify code ([0e4d1b4](https://github.com/inference-gateway/rust-sdk/commit/0e4d1b4a0519adb1c23dc5f25f841af72e7c6527))
+* Rename parameters to arguments in ToolFunctionResponse and update ToolCallResponse structure ([3f36c5f](https://github.com/inference-gateway/rust-sdk/commit/3f36c5f969a7b46417613c9fd11f1760d841ff76))
+
+## [0.7.1-rc.2](https://github.com/inference-gateway/rust-sdk/compare/0.7.1-rc.1...0.7.1-rc.2) (2025-02-09)
+
+### 🐛 Bug Fixes
+
+* Enhance Tool and ToolFunction structures with description and parameters updates, use JSON raw value for now ([cdf6eca](https://github.com/inference-gateway/rust-sdk/commit/cdf6eca712c07003841cea55d57ae62cfa3bca2b))
+
+## [0.7.1-rc.1](https://github.com/inference-gateway/rust-sdk/compare/0.7.0...0.7.1-rc.1) (2025-02-08)
+
+### ♻️ Improvements
+
+* Remove tool_calls from Message struct and update related tests ([265bdec](https://github.com/inference-gateway/rust-sdk/commit/265bdec49b9a4394cb78e0bf286f0c267a9a3b06))
+
+### 👷 CI
+
+* Improve the Release Process Allowing to have Release Candidates ([#3](https://github.com/inference-gateway/rust-sdk/issues/3)) ([b97e5da](https://github.com/inference-gateway/rust-sdk/commit/b97e5da228ccfa8719f0211a0a28fa344bf4954c))
+
+### 📚 Documentation
+
+* Remove unnecessary code block from README.md ([dcb7c26](https://github.com/inference-gateway/rust-sdk/commit/dcb7c26493be82790615b7434acbf3ddb13b4fc1))
+
 ## [0.7.0](https://github.com/inference-gateway/rust-sdk/compare/0.6.3...0.7.0) (2025-02-07)
 
 ### ✨ Features

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "inference-gateway-sdk"
-version = "0.7.0"
+version = "0.8.0-rc.2"
 edition = "2021"
 description = "Rust SDK for interacting with various language models through the Inference Gateway"
 license = "MIT"

README.md

@@ -10,8 +10,8 @@ An SDK written in Rust for the [Inference Gateway](https://github.com/inference-
     - [Listing Models from a specific provider](#listing-models-from-a-specific-provider)
     - [Generating Content](#generating-content)
     - [Streaming Content](#streaming-content)
-    - [Health Check](#health-check)
     - [Tool-Use](#tool-use)
+    - [Health Check](#health-check)
   - [Contributing](#contributing)
   - [License](#license)
 
@@ -193,10 +193,12 @@ use futures_util::{StreamExt, pin_mut};
         Message {
             role: MessageRole::System,
             content: system_message,
+            tool_call_id: None
         },
         Message {
             role: MessageRole::User,
             content: "Write a poem".to_string(),
+            tool_call_id: None
         },
     ];
     let client = InferenceGatewayClient::new("http://localhost:8080");
@@ -224,19 +226,6 @@ use futures_util::{StreamExt, pin_mut};
 // ...rest of the main function
 ```
 
-### Health Check
-
-To check if the Inference Gateway is running, use the `health_check` method:
-
-```rust
-// ...rest of the imports
-use log::info;
-
-// ...main function
-let is_healthy = client.health_check().await?;
-info!("API is healthy: {}", is_healthy);
-```
-
 ### Tool-Use
 
 You can pass to the generate_content function also tools, which will be available for the LLM to use:
@@ -250,33 +239,67 @@ use inference_gateway_sdk::{
     Provider,
     MessageRole,
     Tool,
-    ToolType,
-    ToolParameters,
-    ToolParameterType,
+    ToolFunction,
+    ToolType
 };
 
-let resp = client.generate_content(Provider::Groq, "deepseek-r1-distill-llama-70b", vec![
-Message{
-    role: MessageRole::System,
-    content: "You are an helpful assistent.".to_string()
-},
-Message{
-    role: MessageRole::User,
-    content: "What is the current weather in Berlin?".to_string()
-}
-], vec![
+let tools = vec![
     Tool {
         r#type: ToolType::Function,
-        name: "get_current_weather".to_string(),
-        description: "Get the weather for a location".to_string(),
-        parameters: ToolParameters {
-            name: "location".to_string(),
-            r#type: ToolParameterType::String,
-            default: None,
-            description: "The city name".to_string(),
+        function: ToolFunction {
+            name: "get_current_weather".to_string(),
+            description: "Get the weather for a location".to_string(),
+            parameters: json!({
+                "type": "object",
+                "properties": {
+                    "location": {
+                        "type": "string",
+                        "description": "The city name"
+                    }
+                },
+                "required": ["location"]
+            }),
         },
     },
+];
+let resp = client.with_tools(Some(tools)).generate_content(Provider::Groq, "deepseek-r1-distill-llama-70b", vec![
+Message {
+    role: MessageRole::System,
+    content: "You are an helpful assistent.".to_string(),
+    ..Default::default()
+},
+Message {
+    role: MessageRole::User,
+    content: "What is the current weather in Berlin?".to_string(),
+    ..Default::default()
+}
 ]).await?;
+
+for tool_call in resp.response.tool_calls {
+    log::info!("Tool Call Requested by the LLM: {:?}", tool_call);
+    // Make the function call with the parameters requested by the LLM
+
+    let message = Message {
+        role: MessageRole::Tool,
+        content: "The content from the tool".to_string(),
+        tool_call_id: Some(tool_call.id) // the tool call id so the LLM can reference it
+    };
+
+    // Append this message to the next request
+}
+```
+
+### Health Check
+
+To check if the Inference Gateway is running, use the `health_check` method:
+
+```rust
+// ...rest of the imports
+use log::info;
+
+// ...main function
+let is_healthy = client.health_check().await?;
+info!("API is healthy: {}", is_healthy);
 ```
 
 ## Contributing