neomemx 0.1.2

A high-performance memory library for AI agents with semantic search
Documentation 
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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74

//! OpenAI LLM implementation

use async_trait::async_trait;
use tracing::info;

use super::base::{LlmBase, LlmResponse, Message, ResponseFormat, Tool};
use super::openai_compat::{OpenAICompatClient, OpenAICompatConfig};
use crate::config::OpenAILlmConfig;
use crate::error::{NeomemxError, Result};

/// OpenAI LLM client.
pub struct OpenAILlm {
    client: OpenAICompatClient,
}

impl OpenAILlm {
    /// Creates a new OpenAI client with the given configuration.
    pub fn new(config: OpenAILlmConfig) -> Result<Self> {
        let api_key = config.get_api_key().ok_or_else(|| {
            NeomemxError::LlmError(
                "OpenAI API key not found. Set OPENAI_API_KEY or provide in config.".to_string(),
            )
        })?;

        info!("Creating OpenAI LLM with model: {}", config.model);

        let is_reasoning = Self::is_reasoning_model(&config.model);

        Ok(Self {
            client: OpenAICompatClient::new(OpenAICompatConfig {
                base_url: config.base_url,
                api_key,
                model: config.model,
                temperature: config.temperature,
                max_tokens: config.max_tokens,
                top_p: config.top_p,
                provider_name: "OpenAI",
            })
            .with_skip_sampling_params(is_reasoning),
        })
    }

    fn is_reasoning_model(model: &str) -> bool {
        let model = model.to_lowercase();
        ["o1", "o1-preview", "o3-mini", "o3", "gpt-5"]
            .iter()
            .any(|m| model.contains(m))
    }
}

#[async_trait]
impl LlmBase for OpenAILlm {
    async fn generate_response(
        &self,
        messages: Vec<Message>,
        response_format: Option<ResponseFormat>,
        tools: Option<Vec<Tool>>,
        tool_choice: Option<String>,
    ) -> Result<LlmResponse> {
        self.client.chat_completion(messages, response_format, tools, tool_choice).await
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_is_reasoning_model() {
        assert!(OpenAILlm::is_reasoning_model("o1-preview"));
        assert!(OpenAILlm::is_reasoning_model("o1"));
        assert!(!OpenAILlm::is_reasoning_model("gpt-4"));
    }
}