feat: update examples

rig-core/examples/anthropic_streaming.rs

@@ -19,5 +19,13 @@ async fn main() -> Result<(), anyhow::Error> {
 
     stream_to_stdout(agent, &mut stream).await?;
 
+
+    if let Some(response) = stream.response {
+        println!("Usage: {:?} tokens", response.usage.output_tokens);
+    };
+
+    println!("Message: {:?}", stream.message);
+    
+    
     Ok(())
 }

rig-core/examples/anthropic_streaming_with_tools.rs

@@ -107,5 +107,12 @@ async fn main() -> Result<(), anyhow::Error> {
     println!("Calculate 2 - 5");
     let mut stream = calculator_agent.stream_prompt("Calculate 2 - 5").await?;
     stream_to_stdout(calculator_agent, &mut stream).await?;
+
+    if let Some(response) = stream.response {
+        println!("Usage: {:?} tokens", response.usage.output_tokens);
+    };
+
+    println!("Message: {:?}", stream.message);
+
     Ok(())
 }

rig-core/examples/gemini_streaming.rs

@@ -19,5 +19,10 @@ async fn main() -> Result<(), anyhow::Error> {
 
     stream_to_stdout(agent, &mut stream).await?;
 
+    if let Some(response) = stream.response {
+        println!("Usage: {:?} tokens", response.usage_metadata.total_token_count);
+    };
+
+    println!("Message: {:?}", stream.message);
     Ok(())
 }

rig-core/examples/gemini_streaming_with_tools.rs

@@ -107,5 +107,12 @@ async fn main() -> Result<(), anyhow::Error> {
     println!("Calculate 2 - 5");
     let mut stream = calculator_agent.stream_prompt("Calculate 2 - 5").await?;
     stream_to_stdout(calculator_agent, &mut stream).await?;
+
+    if let Some(response) = stream.response {
+        println!("Usage: {:?} tokens", response.usage_metadata.total_token_count);
+    };
+
+    println!("Message: {:?}", stream.message);
+    
     Ok(())
 }

rig-core/examples/ollama_streaming.rs

@@ -17,5 +17,10 @@ async fn main() -> Result<(), anyhow::Error> {
 
     stream_to_stdout(agent, &mut stream).await?;
 
+    if let Some(response) = stream.response {
+        println!("Usage: {:?} tokens", response.eval_count);
+    };
+
+    println!("Message: {:?}", stream.message);
     Ok(())
 }

rig-core/examples/ollama_streaming_with_tools.rs

@@ -107,5 +107,12 @@ async fn main() -> Result<(), anyhow::Error> {
     println!("Calculate 2 - 5");
     let mut stream = calculator_agent.stream_prompt("Calculate 2 - 5").await?;
     stream_to_stdout(calculator_agent, &mut stream).await?;
+
+    if let Some(response) = stream.response {
+        println!("Usage: {:?} tokens", response.eval_count);
+    };
+
+    println!("Message: {:?}", stream.message);
+    
     Ok(())
 }

rig-core/examples/openai_streaming.rs

@@ -21,5 +21,7 @@ async fn main() -> Result<(), anyhow::Error> {
         println!("Usage: {:?}", response.usage)
     };
     
+    println!("Message: {:?}", stream.message);
+
     Ok(())
 }

rig-core/examples/openai_streaming_with_tools.rs

@@ -107,5 +107,12 @@ async fn main() -> Result<(), anyhow::Error> {
     println!("Calculate 2 - 5");
     let mut stream = calculator_agent.stream_prompt("Calculate 2 - 5").await?;
     stream_to_stdout(calculator_agent, &mut stream).await?;
+    
+    if let Some(response) = stream.response {
+        println!("Usage: {:?}", response.usage)
+    };
+
+    println!("Message: {:?}", stream.message);
+    
     Ok(())
 }

rig-core/src/providers/gemini/completion.rs

@@ -609,7 +609,7 @@ pub mod gemini_api_types {
         HarmCategoryCivicIntegrity,
     }
 
-    #[derive(Debug, Deserialize, Clone)]
+    #[derive(Debug, Deserialize, Clone, Default)]
     #[serde(rename_all = "camelCase")]
     pub struct UsageMetadata {
         pub prompt_token_count: i32,

rig-core/src/providers/gemini/streaming.rs

@@ -9,18 +9,24 @@ use crate::{
     streaming::{self, StreamingCompletionModel},
 };
 
+#[derive(Debug, Deserialize, Default, Clone)]
+#[serde(rename_all = "camelCase")]
+pub struct PartialUsage {
+    pub total_token_count: i32,
+}
+
 #[derive(Debug, Deserialize)]
 #[serde(rename_all = "camelCase")]
 pub struct StreamGenerateContentResponse {
     /// Candidate responses from the model.
     pub candidates: Vec<ContentCandidate>,
     pub model_version: Option<String>,
-    pub usage_metadata: UsageMetadata,
+    pub usage_metadata: Option<PartialUsage>,
 }
 
 #[derive(Clone)]
 pub struct StreamingCompletionResponse {
-    pub usage_metadata: UsageMetadata,
+    pub usage_metadata: PartialUsage,
 }
 
 impl StreamingCompletionModel for CompletionModel {
@@ -90,7 +96,9 @@ impl StreamingCompletionModel for CompletionModel {
 
                     if choice.finish_reason.is_some() {
                         yield Ok(streaming::RawStreamingChoice::FinalResponse(StreamingCompletionResponse {
-                            usage_metadata: data.usage_metadata,
+                            usage_metadata: PartialUsage {
+                                total_token_count: data.usage_metadata.unwrap().total_token_count,
+                            }
                         }))
                     }
                 }

rig-core/src/providers/ollama.rs

@@ -486,6 +486,18 @@ impl StreamingCompletionModel for CompletionModel {
                             continue;
                         }
                     }
+                    
+                    if response.done {
+                        yield Ok(RawStreamingChoice::FinalResponse(StreamingCompletionResponse {
+                            total_duration: response.total_duration,
+                            load_duration: response.load_duration,
+                            prompt_eval_count: response.prompt_eval_count,
+                            prompt_eval_duration: response.prompt_eval_duration,
+                            eval_count: response.eval_count,
+                            eval_duration: response.eval_duration,
+                            done_reason: response.done_reason,
+                        }));
+                    }
                 }
             }
         });

rig-core/src/providers/openai/streaming.rs

@@ -46,12 +46,11 @@ struct StreamingChoice {
 struct StreamingCompletionChunk {
     choices: Vec<StreamingChoice>,
     usage: Option<Usage>,
-    finish_reason: Option<String>,
 }
 
 #[derive(Clone)]
 pub struct StreamingCompletionResponse {
-    pub usage: Option<Usage>,
+    pub usage: Usage,
 }
 
 impl StreamingCompletionModel for CompletionModel {
@@ -62,7 +61,10 @@ impl StreamingCompletionModel for CompletionModel {
     ) -> Result<streaming::StreamingCompletionResponse<Self::StreamingResponse>, CompletionError>
     {
         let mut request = self.create_completion_request(completion_request)?;
-        request = merge(request, json!({"stream": true}));
+        request = merge(
+            request,
+            json!({"stream": true, "stream_options": {"include_usage": true}}),
+        );
 
         let builder = self.client.post("/chat/completions").json(&request);
         send_compatible_streaming_request(builder).await
@@ -86,6 +88,11 @@ pub async fn send_compatible_streaming_request(
     let inner = Box::pin(stream! {
         let mut stream = response.bytes_stream();
 
+        let mut final_usage = Usage {
+            prompt_tokens: 0,
+            total_tokens: 0
+        };
+        
         let mut partial_data = None;
         let mut calls: HashMap<usize, (String, String)> = HashMap::new();
 
@@ -110,8 +117,6 @@ pub async fn send_compatible_streaming_request(
             for line in text.lines() {
                 let mut line = line.to_string();
 
-
-
                 // If there was a remaining part, concat with current line
                 if partial_data.is_some() {
                     line = format!("{}{}", partial_data.unwrap(), line);
@@ -137,7 +142,8 @@ pub async fn send_compatible_streaming_request(
                     continue;
                 };
 
-                let choice = data.choices.first().expect("Should have at least one choice");
+
+                if let Some(choice) = data.choices.first() {
 
                     let delta = &choice.delta;
 
@@ -177,16 +183,12 @@ pub async fn send_compatible_streaming_request(
                         }
                     }
 
-                if let Some(content) = &choice.delta.content {
-                    yield Ok(streaming::RawStreamingChoice::Message(content.clone()))
+                 
                 }
 
-                if data.finish_reason.is_some() {
-                    yield Ok(RawStreamingChoice::FinalResponse(StreamingCompletionResponse {
-                        usage: data.usage
-                    }))
+                if let Some(usage) = data.usage {
+                    final_usage = usage.clone();
                 }
-
             }
         }
 
@@ -195,8 +197,12 @@ pub async fn send_compatible_streaming_request(
                 continue;
             };
 
-            yield Ok(streaming::RawStreamingChoice::ToolCall(name, "".to_string(), arguments))
+            yield Ok(RawStreamingChoice::ToolCall(name, "".to_string(), arguments))
         }
+
+        yield Ok(RawStreamingChoice::FinalResponse(StreamingCompletionResponse {
+            usage: final_usage.clone()
+        }))
     });
 
     Ok(streaming::StreamingCompletionResponse::new(inner))

rig-core/src/streaming.rs

@@ -91,21 +91,25 @@ impl<R: Clone + Unpin> Stream for StreamingCompletionResponse<R> {
 
         match stream.inner.as_mut().poll_next(cx) {
             Poll::Pending => Poll::Pending,
-
             Poll::Ready(None) => {
-                let content = vec![AssistantContent::text(stream.text.clone())];
+
+                let mut content = vec![];
 
                 stream.tool_calls.iter().for_each(|(n, d, a)| {
-                    AssistantContent::tool_call(n, d, a.clone());
+                    content.push(AssistantContent::tool_call(n, d, a.clone()));
                 });
 
+                if content.len() == 0 || stream.text.len() > 0 {
+                    content.insert(0, AssistantContent::text(stream.text.clone()));
+                }
+
                 stream.message = Message::Assistant {
                     content: OneOrMany::many(content)
                         .expect("There should be at least one assistant message"),
                 };
                 
                 Poll::Ready(None)
-            }
+            },
             Poll::Ready(Some(Err(err))) => Poll::Ready(Some(Err(err))),
             Poll::Ready(Some(Ok(choice))) => match choice {
                 RawStreamingChoice::Message(text) => {
@@ -120,7 +124,8 @@ impl<R: Clone + Unpin> Stream for StreamingCompletionResponse<R> {
                 }
                 RawStreamingChoice::FinalResponse(response) => {
                     stream.response = Some(response);
-                    Poll::Pending
+
+                    stream.poll_next_unpin(cx)
                 }
             },
         }