harn-vm 0.10.41

Async bytecode virtual machine for the Harn programming language
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

//! Prompt-cache breakpoint placement.

use super::fixtures::{base_request_payload, cache_control_count};
use crate::llm::providers::openai_compat::OpenAiCompatibleProvider;
use serde_json::json;

#[test]
fn openrouter_anthropic_cache_uses_top_level_breakpoint() {
    let mut payload = base_request_payload();
    payload.model = "anthropic/claude-sonnet-4-6".to_string();
    payload.cache = true;

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert_eq!(body["cache_control"], json!({"type": "ephemeral"}));
    assert_eq!(cache_control_count(&body), 1);
}

#[test]
fn openrouter_qwen_explicit_cache_uses_last_content_block() {
    let mut payload = base_request_payload();
    payload.model = "qwen/qwen3.6-plus".to_string();
    payload.cache = true;
    payload.messages = vec![json!({
        "role": "user",
        "content": [
            {"type": "text", "text": "stable reference"},
            {"type": "text", "text": "question"}
        ],
    })];

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert!(body.get("cache_control").is_none());
    assert_eq!(
        body["messages"][0]["content"][1]["cache_control"],
        json!({"type": "ephemeral"})
    );
    assert_eq!(cache_control_count(&body), 1);
}

#[test]
fn openrouter_gemini_explicit_cache_uses_last_content_block() {
    let mut payload = base_request_payload();
    payload.model = "google/gemini-2.5-flash".to_string();
    payload.cache = true;
    payload.messages = vec![json!({
        "role": "user",
        "content": [
            {"type": "text", "text": "stable reference"},
            {"type": "text", "text": "question"}
        ],
    })];

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert_eq!(
        body["messages"][0]["content"][1]["cache_control"],
        json!({"type": "ephemeral"})
    );
    assert_eq!(cache_control_count(&body), 1);
}

#[test]
fn openrouter_automatic_cache_route_does_not_emit_cache_control() {
    let mut payload = base_request_payload();
    payload.model = "deepseek/deepseek-v3".to_string();
    payload.cache = true;

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert_eq!(cache_control_count(&body), 0);
}

#[test]
fn openrouter_qwen_open_weight_route_does_not_emit_cache_control() {
    let mut payload = base_request_payload();
    payload.model = "qwen/qwen3.6-35b-a3b".to_string();
    payload.cache = true;

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert_eq!(cache_control_count(&body), 0);
}

#[test]
fn openrouter_explicit_cache_preserves_existing_message_breakpoint() {
    let mut payload = base_request_payload();
    payload.model = "qwen/qwen3-coder-plus".to_string();
    payload.cache = true;
    payload.messages = vec![json!({
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "stable reference",
                "cache_control": {"type": "ephemeral"}
            },
            {"type": "text", "text": "question"}
        ],
    })];

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert_eq!(
        body["messages"][0]["content"][0]["cache_control"],
        json!({"type": "ephemeral"})
    );
    assert!(body["messages"][0]["content"][1]
        .get("cache_control")
        .is_none());
    assert_eq!(cache_control_count(&body), 1);
}

#[test]
fn openrouter_cache_preserves_existing_tool_breakpoint() {
    let mut payload = base_request_payload();
    payload.model = "anthropic/claude-sonnet-4-6".to_string();
    payload.cache = true;
    payload.native_tools = Some(vec![json!({
        "type": "function",
        "cache_control": {"type": "ephemeral"},
        "function": {
            "name": "lookup",
            "description": "Lookup stable context",
            "parameters": {"type": "object"}
        }
    })]);

    let body = OpenAiCompatibleProvider::build_request_body(&payload, false);

    assert!(body.get("cache_control").is_none());
    assert_eq!(
        body["tools"][0]["cache_control"],
        json!({"type": "ephemeral"})
    );
    assert_eq!(cache_control_count(&body), 1);
}