omnillm 0.1.5

Production-grade LLM API gateway with multi-key load balancing, per-key rate limiting, circuit breaking, and cost tracking
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

use serde_json::Value;

use super::{
    CacheBreakpoint, CapabilitySet, GenerationConfig, LlmRequest, Message, MessagePart,
    MessageRole, PromptCacheKey, PromptCachePolicy, PromptCacheRetention, RequestItem,
    ToolDefinition, VendorExtensions,
};

/// Helper for constructing cache-friendly prompt layouts.
#[derive(Debug, Clone)]
pub struct PromptLayoutBuilder {
    model: String,
    instructions: Option<String>,
    tools: Vec<ToolDefinition>,
    stable_messages: Vec<Message>,
    dynamic_messages: Vec<Message>,
    generation: GenerationConfig,
    prompt_cache: Option<PromptCachePolicy>,
    generated_key: Option<GeneratedPromptCacheKey>,
}

#[derive(Debug, Clone)]
struct GeneratedPromptCacheKey {
    namespace: String,
    tenant_scope: Option<String>,
    required: bool,
    retention: PromptCacheRetention,
}

impl PromptLayoutBuilder {
    pub fn new(model: impl Into<String>) -> Self {
        Self {
            model: model.into(),
            instructions: None,
            tools: Vec::new(),
            stable_messages: Vec::new(),
            dynamic_messages: Vec::new(),
            generation: GenerationConfig::default(),
            prompt_cache: None,
            generated_key: None,
        }
    }

    pub fn instructions(mut self, instructions: impl Into<String>) -> Self {
        self.instructions = Some(instructions.into());
        self
    }

    pub fn tool(mut self, tool: ToolDefinition) -> Self {
        self.tools.push(tool);
        self
    }

    pub fn stable_message(mut self, message: Message) -> Self {
        self.stable_messages.push(message);
        self
    }

    pub fn dynamic_message(mut self, message: Message) -> Self {
        self.dynamic_messages.push(message);
        self
    }

    pub fn user_input(self, text: impl Into<String>) -> Self {
        self.dynamic_message(Message::text(MessageRole::User, text))
    }

    pub fn dynamic_rag_context(self, value: Value) -> Self {
        self.dynamic_message(Message {
            role: MessageRole::User,
            parts: vec![MessagePart::Json { value }],
            raw_message: None,
            vendor_extensions: VendorExtensions::new(),
        })
    }

    pub fn generation(mut self, generation: GenerationConfig) -> Self {
        self.generation = generation;
        self
    }

    pub fn prompt_cache(mut self, policy: PromptCachePolicy) -> Self {
        self.prompt_cache = Some(policy);
        self.generated_key = None;
        self
    }

    pub fn stable_prefix_cache_key(
        mut self,
        namespace: impl Into<String>,
        tenant_scope: Option<impl Into<String>>,
        retention: PromptCacheRetention,
        required: bool,
    ) -> Self {
        self.generated_key = Some(GeneratedPromptCacheKey {
            namespace: namespace.into(),
            tenant_scope: tenant_scope.map(Into::into),
            required,
            retention,
        });
        self.prompt_cache = None;
        self
    }

    pub fn build(mut self) -> LlmRequest {
        if let Some(generated) = self.generated_key.take() {
            let key = PromptCacheKey::Explicit {
                value: self.generated_cache_key(&generated),
            };
            self.prompt_cache = Some(if generated.required {
                PromptCachePolicy::Required {
                    key: Some(key),
                    retention: generated.retention,
                    breakpoint: CacheBreakpoint::Auto,
                    vendor_extensions: VendorExtensions::new(),
                }
            } else {
                PromptCachePolicy::BestEffort {
                    key: Some(key),
                    retention: generated.retention,
                    breakpoint: CacheBreakpoint::Auto,
                    vendor_extensions: VendorExtensions::new(),
                }
            });
        }

        let mut messages = self.stable_messages;
        messages.extend(self.dynamic_messages);
        let input = messages.iter().cloned().map(RequestItem::from).collect();

        LlmRequest {
            model: self.model,
            instructions: self.instructions,
            input,
            messages,
            capabilities: CapabilitySet {
                tools: self.tools,
                prompt_cache: self.prompt_cache,
                ..Default::default()
            },
            generation: self.generation,
            metadata: VendorExtensions::new(),
            vendor_extensions: VendorExtensions::new(),
        }
    }

    fn generated_cache_key(&self, generated: &GeneratedPromptCacheKey) -> String {
        let fingerprint = serde_json::json!({
            "model": self.model,
            "instructions": self.instructions,
            "tools": self.tools,
            "stable_messages": self.stable_messages,
        });
        let hash = fnv1a64(fingerprint.to_string().as_bytes());
        match &generated.tenant_scope {
            Some(scope) => format!("{}:{}:{hash:016x}", generated.namespace, scope),
            None => format!("{}:{hash:016x}", generated.namespace),
        }
    }
}

fn fnv1a64(bytes: &[u8]) -> u64 {
    const OFFSET: u64 = 0xcbf29ce484222325;
    const PRIME: u64 = 0x100000001b3;
    let mut hash = OFFSET;
    for byte in bytes {
        hash ^= u64::from(*byte);
        hash = hash.wrapping_mul(PRIME);
    }
    hash
}