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

use serde_json::{json, Map, Value};

use crate::types::{
    CacheBreakpoint, LlmRequest, PromptCacheKey, PromptCachePolicy, PromptCacheRetention,
    PromptCacheUsage, TokenUsage, VendorExtensions,
};

use super::super::super::common::{
    extend_with_vendor_extensions, nested_u32, prompt_cache_key_value,
    prompt_cache_vendor_extensions,
};
use super::super::super::ProtocolError;

pub(in crate::protocol::openai) fn emit_openai_prompt_cache_policy(
    map: &mut Map<String, Value>,
    request: &LlmRequest,
) -> Result<(), ProtocolError> {
    let Some(policy) = request.capabilities.effective_prompt_cache() else {
        return Ok(());
    };
    if policy.is_disabled() {
        return Ok(());
    }

    if !policy.breakpoint().is_auto() && policy.is_required() {
        return Err(ProtocolError::UnsupportedFeature(
            "OpenAI prompt cache does not support explicit breakpoints".into(),
        ));
    }

    if let Some(key) = policy.key() {
        map.insert(
            "prompt_cache_key".into(),
            Value::String(prompt_cache_key_value(key, request)),
        );
    }

    match policy.retention() {
        PromptCacheRetention::ProviderDefault => {}
        PromptCacheRetention::Short => {
            map.insert(
                "prompt_cache_retention".into(),
                Value::String("in_memory".into()),
            );
        }
        PromptCacheRetention::Long => {
            map.insert("prompt_cache_retention".into(), Value::String("24h".into()));
        }
    }

    extend_with_vendor_extensions(map, prompt_cache_vendor_extensions(&policy));
    Ok(())
}

pub(in crate::protocol::openai) fn parse_openai_prompt_cache_policy(
    body: &Value,
) -> Option<PromptCachePolicy> {
    let key = body
        .get("prompt_cache_key")
        .and_then(Value::as_str)
        .map(|value| PromptCacheKey::Explicit {
            value: value.to_string(),
        });
    let retention = body
        .get("prompt_cache_retention")
        .and_then(Value::as_str)
        .map(parse_openai_prompt_cache_retention)
        .unwrap_or_default();

    if key.is_none() && retention == PromptCacheRetention::ProviderDefault {
        None
    } else {
        Some(PromptCachePolicy::BestEffort {
            key,
            retention,
            breakpoint: CacheBreakpoint::Auto,
            vendor_extensions: VendorExtensions::new(),
        })
    }
}

pub(in crate::protocol::openai) fn parse_openai_prompt_cache_retention(
    value: &str,
) -> PromptCacheRetention {
    match value {
        "24h" => PromptCacheRetention::Long,
        "in_memory" => PromptCacheRetention::Short,
        _ => PromptCacheRetention::ProviderDefault,
    }
}

pub(in crate::protocol::openai) fn parse_openai_prompt_cache_usage(
    usage: Option<&Value>,
) -> Option<PromptCacheUsage> {
    let usage = usage?;
    let cached_input_tokens = nested_u32(usage, &["input_tokens_details", "cached_tokens"])
        .or_else(|| nested_u32(usage, &["prompt_tokens_details", "cached_tokens"]));

    let prompt_cache = PromptCacheUsage {
        cached_input_tokens,
        ..Default::default()
    };
    (!prompt_cache.is_empty()).then_some(prompt_cache)
}

pub(in crate::protocol::openai) fn openai_responses_usage_json(usage: &TokenUsage) -> Value {
    let mut map = Map::new();
    map.insert("input_tokens".into(), Value::from(usage.prompt_tokens));
    map.insert("output_tokens".into(), Value::from(usage.completion_tokens));
    map.insert("total_tokens".into(), Value::from(usage.total()));
    if let Some(cached_tokens) = usage
        .prompt_cache
        .as_ref()
        .and_then(|prompt_cache| prompt_cache.cached_input_tokens)
    {
        map.insert(
            "input_tokens_details".into(),
            json!({ "cached_tokens": cached_tokens }),
        );
    }
    Value::Object(map)
}

pub(in crate::protocol::openai) fn openai_chat_usage_json(usage: &TokenUsage) -> Value {
    let mut map = Map::new();
    map.insert("prompt_tokens".into(), Value::from(usage.prompt_tokens));
    map.insert(
        "completion_tokens".into(),
        Value::from(usage.completion_tokens),
    );
    map.insert("total_tokens".into(), Value::from(usage.total()));
    if let Some(cached_tokens) = usage
        .prompt_cache
        .as_ref()
        .and_then(|prompt_cache| prompt_cache.cached_input_tokens)
    {
        map.insert(
            "prompt_tokens_details".into(),
            json!({ "cached_tokens": cached_tokens }),
        );
    }
    Value::Object(map)
}