vv-agent 0.6.2

VectorVein agent runtime, SDK, CLI, tools, and workspace backends
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

use super::*;

#[test]
fn vv_llm_client_fails_over_to_next_endpoint_client() {
    let llm = VvLlmClient::new_with_named_endpoint_clients(
        "openai",
        "gpt-4o-alias",
        "gpt-4o-mini",
        vec![
            (
                "primary-endpoint".to_string(),
                "gpt-4o-mini".to_string(),
                Box::new(FailingChatClient) as Box<dyn vv_llm::ChatClient>,
            ),
            (
                "backup-endpoint".to_string(),
                "gpt-4o-mini-backup".to_string(),
                Box::new(UsageMissingChatClient) as Box<dyn vv_llm::ChatClient>,
            ),
        ],
        90.0,
    )
    .with_randomize_endpoints(false)
    .with_retry_policy(1, 0.0);

    let response = llm
        .complete(LlmRequest::new(
            "gpt-4o-alias",
            vec![Message::user("fall over to backup endpoint")],
        ))
        .expect("completion from fallback endpoint");

    assert_eq!(response.content, "estimated usage response");
    assert_eq!(response.raw["used_endpoint_id"], json!("backup-endpoint"));
    assert_eq!(response.raw["used_model_id"], json!("gpt-4o-mini-backup"));
    assert_eq!(response.raw["stream_mode"], json!(true));
}

#[test]
fn vv_llm_client_prefers_last_successful_endpoint() {
    let primary = CountingFailingChatClient::default();
    let primary_probe = primary.clone();
    let llm = VvLlmClient::new_with_named_endpoint_clients(
        "openai",
        "gpt-4o-mini",
        "gpt-4o-mini",
        vec![
            (
                "primary-endpoint".to_string(),
                "gpt-4o-mini".to_string(),
                Box::new(primary) as Box<dyn vv_llm::ChatClient>,
            ),
            (
                "backup-endpoint".to_string(),
                "gpt-4o-mini".to_string(),
                Box::new(UsageMissingChatClient) as Box<dyn vv_llm::ChatClient>,
            ),
        ],
        90.0,
    )
    .with_randomize_endpoints(false)
    .with_retry_policy(1, 0.0);

    let first = llm
        .complete(LlmRequest::new("gpt-4o-mini", vec![Message::user("first")]))
        .expect("first fallback completion");
    let second = llm
        .complete(LlmRequest::new(
            "gpt-4o-mini",
            vec![Message::user("second")],
        ))
        .expect("second preferred completion");

    assert_eq!(first.raw["used_endpoint_id"], json!("backup-endpoint"));
    assert_eq!(second.raw["used_endpoint_id"], json!("backup-endpoint"));
    assert_eq!(primary_probe.calls(), 1);
}

#[test]
fn vv_llm_client_exposes_agent_endpoint_randomization_policy() {
    let llm = VvLlmClient::new_with_named_endpoint_clients(
        "openai",
        "gpt-4o-mini",
        "gpt-4o-mini",
        vec![
            (
                "primary-endpoint".to_string(),
                "gpt-4o-mini".to_string(),
                Box::new(UsageMissingChatClient) as Box<dyn vv_llm::ChatClient>,
            ),
            (
                "backup-endpoint".to_string(),
                "gpt-4o-mini".to_string(),
                Box::new(UsageMissingChatClient) as Box<dyn vv_llm::ChatClient>,
            ),
        ],
        90.0,
    );

    assert!(llm.randomize_endpoints());
    let llm = llm.with_randomize_endpoints(false);
    assert!(!llm.randomize_endpoints());
}

#[test]
fn vv_llm_client_retries_endpoint_before_failover() {
    let flaky = FlakyChatClient::new(1);
    let flaky_probe = flaky.clone();
    let llm = VvLlmClient::new(
        "openai",
        "gpt-4o-mini",
        "gpt-4o-mini",
        Box::new(flaky),
        90.0,
    )
    .with_retry_policy(2, 0.0);

    let response = llm
        .complete(LlmRequest::new(
            "gpt-4o-mini",
            vec![Message::user("retry once")],
        ))
        .expect("retry succeeds");

    assert_eq!(response.content, "flaky success");
    assert_eq!(flaky_probe.calls(), 2);
}

#[test]
fn vv_llm_client_uses_endpoint_model_for_selected_alias() {
    let llm = VvLlmClient::new_with_named_endpoint_clients(
        "openai",
        "gpt-alias",
        "gpt-provider-model",
        vec![(
            "primary-endpoint".to_string(),
            "gpt-provider-model".to_string(),
            Box::new(UsageMissingChatClient) as Box<dyn vv_llm::ChatClient>,
        )],
        90.0,
    );

    let response = llm
        .complete(LlmRequest::new(
            "gpt-alias",
            vec![Message::user("use provider model id")],
        ))
        .expect("completion from primary endpoint");

    assert_eq!(response.raw["used_model_id"], json!("gpt-provider-model"));
}