streaming_chat/
streaming_chat.rs

1use helios_engine::config::LLMConfig;
2/// Example: Using streaming responses with Helios Engine
3///
4/// This example demonstrates how to use the streaming API to get
5/// real-time responses from the LLM, including detection of thinking tags.
6use helios_engine::{ChatMessage, ChatSession, LLMClient};
7use std::io::{self, Write};
8
9#[tokio::main]
10async fn main() -> helios_engine::Result<()> {
11    println!("🚀 Helios Engine - Streaming Example");
12    println!("=====================================\n");
13
14    // Setup LLM configuration
15    let llm_config = LLMConfig {
16        model_name: "gpt-3.5-turbo".to_string(),
17        base_url: "https://api.openai.com/v1".to_string(),
18        api_key: std::env::var("OPENAI_API_KEY")
19            .unwrap_or_else(|_| "your-api-key-here".to_string()),
20        temperature: 0.7,
21        max_tokens: 2048,
22    };
23
24    let client = LLMClient::new(helios_engine::llm::LLMProviderType::Remote(llm_config)).await?;
25
26    println!("Example 1: Simple Streaming Response");
27    println!("======================================\n");
28
29    let messages = vec![
30        ChatMessage::system("You are a helpful assistant."),
31        ChatMessage::user("Write a short poem about coding."),
32    ];
33
34    print!("Assistant: ");
35    io::stdout().flush()?;
36
37    let response = client
38        .chat_stream(messages, None, |chunk| {
39            print!("{}", chunk);
40            io::stdout().flush().unwrap();
41        })
42        .await?;
43
44    println!("\n\n");
45
46    println!("Example 2: Interactive Streaming Chat");
47    println!("======================================\n");
48
49    let mut session = ChatSession::new().with_system_prompt("You are a helpful coding assistant.");
50
51    let questions = vec![
52        "What is Rust?",
53        "What are its main benefits?",
54        "Show me a simple example.",
55    ];
56
57    for question in questions {
58        println!("User: {}", question);
59        session.add_user_message(question);
60
61        print!("Assistant: ");
62        io::stdout().flush()?;
63
64        let response = client
65            .chat_stream(session.get_messages(), None, |chunk| {
66                print!("{}", chunk);
67                io::stdout().flush().unwrap();
68            })
69            .await?;
70
71        session.add_assistant_message(&response.content);
72        println!("\n");
73    }
74
75    println!("\nExample 3: Streaming with Thinking Tags");
76    println!("=========================================\n");
77    println!("When using models that support thinking tags (like o1),");
78    println!("you can detect and display them during streaming.\n");
79
80    struct ThinkingTracker {
81        in_thinking: bool,
82        thinking_buffer: String,
83    }
84
85    impl ThinkingTracker {
86        fn new() -> Self {
87            Self {
88                in_thinking: false,
89                thinking_buffer: String::new(),
90            }
91        }
92
93        fn process_chunk(&mut self, chunk: &str) -> String {
94            let mut output = String::new();
95            let mut chars = chunk.chars().peekable();
96
97            while let Some(c) = chars.next() {
98                if c == '<' {
99                    let remaining: String = chars.clone().collect();
100                    if remaining.starts_with("thinking>") {
101                        self.in_thinking = true;
102                        self.thinking_buffer.clear();
103                        output.push_str("\n💭 [Thinking");
104                        for _ in 0..9 {
105                            chars.next();
106                        }
107                        continue;
108                    } else if remaining.starts_with("/thinking>") {
109                        self.in_thinking = false;
110                        output.push_str("]\n");
111                        for _ in 0..10 {
112                            chars.next();
113                        }
114                        continue;
115                    }
116                }
117
118                if self.in_thinking {
119                    self.thinking_buffer.push(c);
120                    if self.thinking_buffer.len() % 3 == 0 {
121                        output.push('.');
122                    }
123                } else {
124                    output.push(c);
125                }
126            }
127
128            output
129        }
130    }
131
132    let messages = vec![ChatMessage::user(
133        "Solve this problem: What is 15 * 234 + 89?",
134    )];
135
136    let mut tracker = ThinkingTracker::new();
137    print!("Assistant: ");
138    io::stdout().flush()?;
139
140    let _response = client
141        .chat_stream(messages, None, |chunk| {
142            let output = tracker.process_chunk(chunk);
143            print!("{}", output);
144            io::stdout().flush().unwrap();
145        })
146        .await?;
147
148    println!("\n\n✅ Streaming examples completed!");
149    println!("\nKey benefits of streaming:");
150    println!("  • Real-time response display");
151    println!("  • Better user experience for long responses");
152    println!("  • Ability to show thinking/reasoning process");
153    println!("  • Early cancellation possible (future feature)");
154
155    Ok(())
156}
streaming_chat/streaming_chat.rs

streaming_chat/
streaming_chat.rs
