llmkit 0.1.3

Production-grade LLM client - 100+ providers, 11,000+ models. Pure Rust.
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
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212

//! Nvidia NIM provider for AI inference microservices.
//!
//! Nvidia NIM (NVIDIA Inference Microservice) provides optimized
//! inference for AI models with an OpenAI-compatible API.
//!
//! # Features
//! - OpenAI-compatible API
//! - Nvidia-optimized models (Llama, Mistral, Nemotron, etc.)
//! - Streaming responses
//! - Function calling support
//! - Vision model support

use serde::{Deserialize, Serialize};

use crate::error::{Error, Result};

/// Nvidia NIM provider for optimized AI inference
pub struct NvidiaNIMProvider {
    #[allow(dead_code)]
    api_key: String,
    #[allow(dead_code)]
    base_url: String,
}

impl NvidiaNIMProvider {
    /// Create a new Nvidia NIM provider with API key
    pub fn new(api_key: &str) -> Self {
        Self {
            api_key: api_key.to_string(),
            base_url: "https://integrate.api.nvidia.com/v1".to_string(),
        }
    }

    /// Create with custom base URL (for self-hosted NIM)
    pub fn with_base_url(api_key: &str, base_url: &str) -> Self {
        Self {
            api_key: api_key.to_string(),
            base_url: base_url.to_string(),
        }
    }

    /// Create from environment variable `NVIDIA_NIM_API_KEY`
    pub fn from_env() -> Result<Self> {
        let api_key = std::env::var("NVIDIA_NIM_API_KEY").map_err(|_| {
            Error::Configuration("NVIDIA_NIM_API_KEY environment variable not set".to_string())
        })?;
        Ok(Self::new(&api_key))
    }

    /// Get list of available models
    pub async fn list_models(&self) -> Result<Vec<String>> {
        Ok(vec![
            // Nvidia Nemotron models
            "nvidia/llama-3.1-nemotron-70b-instruct".to_string(),
            "nvidia/nemotron-4-340b-instruct".to_string(),
            "nvidia/nemotron-mini-4b-instruct".to_string(),
            // Llama models (Nvidia optimized)
            "meta/llama-3.3-70b-instruct".to_string(),
            "meta/llama-3.2-3b-instruct".to_string(),
            "meta/llama-3.2-1b-instruct".to_string(),
            "meta/llama-3.1-405b-instruct".to_string(),
            "meta/llama-3.1-70b-instruct".to_string(),
            "meta/llama-3.1-8b-instruct".to_string(),
            // Vision models
            "meta/llama-3.2-90b-vision-instruct".to_string(),
            "meta/llama-3.2-11b-vision-instruct".to_string(),
            // Mistral models
            "mistralai/mistral-large-2-instruct".to_string(),
            "mistralai/mixtral-8x22b-instruct-v0.1".to_string(),
            "mistralai/mixtral-8x7b-instruct-v0.1".to_string(),
            "mistralai/mistral-7b-instruct-v0.3".to_string(),
            // Google Gemma
            "google/gemma-2-27b-it".to_string(),
            "google/gemma-2-9b-it".to_string(),
            "google/gemma-2-2b-it".to_string(),
            // Microsoft Phi
            "microsoft/phi-3.5-mini-instruct".to_string(),
            "microsoft/phi-3-medium-128k-instruct".to_string(),
            // Qwen
            "qwen/qwen2.5-72b-instruct".to_string(),
            "qwen/qwen2.5-7b-instruct".to_string(),
            // DeepSeek
            "deepseek-ai/deepseek-r1".to_string(),
        ])
    }

    /// Get model details
    pub fn get_model_info(model: &str) -> Option<NvidiaNIMModelInfo> {
        match model {
            "nvidia/llama-3.1-nemotron-70b-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 131072,
                supports_vision: false,
                supports_tools: true,
                max_output_tokens: 16384,
            }),
            "nvidia/nemotron-4-340b-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 4096,
                supports_vision: false,
                supports_tools: true,
                max_output_tokens: 4096,
            }),
            "meta/llama-3.3-70b-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 131072,
                supports_vision: false,
                supports_tools: true,
                max_output_tokens: 16384,
            }),
            "meta/llama-3.1-405b-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 131072,
                supports_vision: false,
                supports_tools: true,
                max_output_tokens: 16384,
            }),
            "meta/llama-3.2-90b-vision-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 131072,
                supports_vision: true,
                supports_tools: true,
                max_output_tokens: 16384,
            }),
            "meta/llama-3.2-11b-vision-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 131072,
                supports_vision: true,
                supports_tools: true,
                max_output_tokens: 16384,
            }),
            "mistralai/mistral-large-2-instruct" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 131072,
                supports_vision: false,
                supports_tools: true,
                max_output_tokens: 16384,
            }),
            "deepseek-ai/deepseek-r1" => Some(NvidiaNIMModelInfo {
                name: model.to_string(),
                context_window: 65536,
                supports_vision: false,
                supports_tools: false,
                max_output_tokens: 32768,
            }),
            _ => None,
        }
    }
}

/// Nvidia NIM model information
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct NvidiaNIMModelInfo {
    /// Model name
    pub name: String,
    /// Context window size
    pub context_window: u32,
    /// Whether this model supports vision
    pub supports_vision: bool,
    /// Whether this model supports function calling
    pub supports_tools: bool,
    /// Maximum output tokens
    pub max_output_tokens: u32,
}

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

    #[test]
    fn test_nvidia_nim_provider_creation() {
        let provider = NvidiaNIMProvider::new("test-key");
        assert_eq!(provider.api_key, "test-key");
        assert_eq!(provider.base_url, "https://integrate.api.nvidia.com/v1");
    }

    #[test]
    fn test_nvidia_nim_custom_base_url() {
        let provider = NvidiaNIMProvider::with_base_url("test-key", "https://custom.api/v1");
        assert_eq!(provider.base_url, "https://custom.api/v1");
    }

    #[tokio::test]
    async fn test_list_models() {
        let provider = NvidiaNIMProvider::new("test-key");
        let models = provider.list_models().await.unwrap();
        assert!(!models.is_empty());
        assert!(models.iter().any(|m| m.contains("nemotron")));
    }

    #[test]
    fn test_get_model_info() {
        let info =
            NvidiaNIMProvider::get_model_info("nvidia/llama-3.1-nemotron-70b-instruct").unwrap();
        assert!(info.supports_tools);
        assert!(!info.supports_vision);
        assert_eq!(info.context_window, 131072);
    }

    #[test]
    fn test_get_model_info_vision() {
        let info = NvidiaNIMProvider::get_model_info("meta/llama-3.2-90b-vision-instruct").unwrap();
        assert!(info.supports_vision);
        assert!(info.supports_tools);
    }

    #[test]
    fn test_model_info_invalid() {
        let info = NvidiaNIMProvider::get_model_info("invalid-model");
        assert!(info.is_none());
    }
}