nexil 0.6.2

Provider-agnostic LLM toolkit — streaming, tool calls, tape storage, OAuth
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

//! Provider policy decisions shared across request paths.

use serde::{Deserialize, Serialize};

/// Provider-specific behavioural toggles.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct ProviderPolicy {
    pub enable_responses_without_capability: bool,
    pub include_usage_in_completion_stream: bool,
    pub completion_max_tokens_arg: String,
    pub responses_tools_blocked_model_prefixes: Vec<String>,
}

impl Default for ProviderPolicy {
    fn default() -> Self {
        Self {
            enable_responses_without_capability: false,
            include_usage_in_completion_stream: false,
            completion_max_tokens_arg: "max_tokens".to_owned(),
            responses_tools_blocked_model_prefixes: Vec::new(),
        }
    }
}

fn normalize_provider_name(provider_name: &str) -> String {
    provider_name.trim().to_lowercase()
}

/// Look up the policy for a given provider name.
///
/// Unknown providers fall back to the default policy.
pub fn provider_policy(provider_name: &str) -> ProviderPolicy {
    let normalized = normalize_provider_name(provider_name);
    match normalized.as_str() {
        "github-copilot" => ProviderPolicy {
            include_usage_in_completion_stream: true,
            completion_max_tokens_arg: "max_tokens".to_owned(),
            ..Default::default()
        },
        "openai" => ProviderPolicy {
            enable_responses_without_capability: true,
            include_usage_in_completion_stream: true,
            completion_max_tokens_arg: "max_completion_tokens".to_owned(),
            ..Default::default()
        },
        "openrouter" => ProviderPolicy {
            enable_responses_without_capability: true,
            include_usage_in_completion_stream: true,
            responses_tools_blocked_model_prefixes: vec!["anthropic/".to_owned()],
            ..Default::default()
        },
        _ => ProviderPolicy::default(),
    }
}

fn responses_tools_blocked_for_model(provider_name: &str, model_id: &str) -> bool {
    let policy = provider_policy(provider_name);
    let lowered_model = model_id.trim().to_lowercase();
    policy
        .responses_tools_blocked_model_prefixes
        .iter()
        .any(|prefix| lowered_model.starts_with(prefix))
}

/// Return the reason why the responses format should be rejected, or `None`
/// if it is acceptable.
pub fn responses_rejection_reason(
    provider_name: &str,
    model_id: &str,
    has_tools: bool,
    supports_responses: bool,
) -> Option<String> {
    if has_tools && responses_tools_blocked_for_model(provider_name, model_id) {
        return Some(
            "responses format is not supported for this model when tools are enabled".to_owned(),
        );
    }
    if supports_responses {
        return None;
    }
    if provider_policy(provider_name).enable_responses_without_capability {
        return None;
    }
    Some("responses format is not supported by this provider".to_owned())
}

/// Whether the provider+model combination supports the messages format.
pub fn supports_messages_format(provider_name: &str, model_id: &str) -> bool {
    let normalized_provider = normalize_provider_name(provider_name);
    let normalized_model = model_id.trim().to_lowercase();
    normalized_provider == "anthropic" || normalized_model.starts_with("anthropic/")
}

/// Whether usage data should be included in the completion stream.
pub fn should_include_completion_stream_usage(provider_name: &str) -> bool {
    provider_policy(provider_name).include_usage_in_completion_stream
}

/// The argument name used for max tokens in completion requests.
pub fn completion_max_tokens_arg(provider_name: &str) -> String {
    provider_policy(provider_name).completion_max_tokens_arg
}