1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
// Import required configuration structs and traits.
use ambi::llm::ChatTemplateType;
use ambi::types::config::OpenAIEngineConfig;
use ambi::{Agent, LLMEngineConfig};
use anyhow::Result;
#[tokio::main]
async fn main() -> Result<()> {
// Step 1: Configure the backend LLM engine.
let engine_config = LLMEngineConfig::OpenAI(OpenAIEngineConfig {
api_key: "mock-key".to_string(),
base_url: "https://api.openai.com/v1".to_string(),
model_name: "gpt-4o-mini".to_string(),
temp: 0.7,
top_p: 0.9,
});
// Step 2: Instantiate the Agent and configure its memory management.
let mut _agent = Agent::make(engine_config)
.await?
.template(ChatTemplateType::Chatml)
// Step 3: Define the strict eviction strategy to trigger truncation early for testing.
// Parameters: (keep_head_count, keep_tail_count, max_safe_tokens)
.with_eviction_strategy(2, 2, 50)
// Step 4: Inject a closure to handle messages that are evicted from the context window.
// This is highly useful for archiving old conversations to a Vector Database for long-term RAG memory.
.on_evict(|evicted_messages| {
println!("\n[System Notification] Context window limit reached.");
println!(
"[Memory Manager] Evicting {} old messages...",
evicted_messages.len()
);
// Step 5: Iterate through the evicted messages and simulate an archiving process.
for (index, msg) in evicted_messages.iter().enumerate() {
println!(
" -> Archiving message #{}: {} bytes saved.",
index,
msg.text_len() // Using the zero-allocation length calculation method.
);
}
println!("[Memory Manager] Archiving complete. Continuing conversation...\n");
});
println!("Agent initialized. Long conversation simulation ready.");
// In a real scenario, you would repeatedly call `agent.chat(...)` here.
// Once the total context exceeds 50 tokens, the `on_evict` closure will automatically trigger.
Ok(())
}