llmshim 0.7.0

Blazing fast LLM API translation layer in pure Rust
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
172
173
174
175
176
177

//! Live checks for the current provider models and Fable conversation binding.
//! cargo test --features proxy --test integration_current_models -- --ignored --nocapture
use futures::StreamExt;
use serde_json::{json, Value};

#[tokio::test]
#[ignore = "requires provider API keys and incurs usage charges"]
async fn current_models_completion_and_streaming() {
    llmshim::env::load_all();
    let router = llmshim::router::Router::from_env();
    for model in [
        "gemini/gemini-3.8-flash",
        "anthropic/claude-opus-5",
        "anthropic/claude-fable-5",
        "anthropic/claude-fable-5-1",
        "xai/grok-4.6",
    ] {
        let input = json!({"model": model, "messages": [{"role": "user", "content": "Reply with only pong."}],
            "max_tokens": 512, "reasoning_effort": "none"});
        let result = llmshim::completion(&router, &input).await.unwrap();
        assert_eq!(result["choices"][0]["finish_reason"], "stop", "{model}");
        assert!(result["choices"][0]["message"]["content"]
            .as_str()
            .unwrap()
            .to_lowercase()
            .contains("pong"));
        println!("PASS {model} completion (reasoning_effort:none)");
        let mut stream = llmshim::stream(&router, &input).await.unwrap();
        let mut text = String::new();
        let mut finished = false;
        while let Some(chunk) = stream.next().await {
            let chunk: Value = serde_json::from_str(&chunk.unwrap()).unwrap();
            if let Some(delta) = chunk["choices"][0]["delta"]["content"].as_str() {
                text.push_str(delta);
            }
            finished |= chunk["choices"][0]["finish_reason"] == "stop";
        }
        assert!(text.to_lowercase().contains("pong"), "{model}: {text}");
        assert!(finished, "{model} lacks a completion event");
        println!("PASS {model} streaming");
    }
}

#[tokio::test]
#[ignore = "requires provider API keys and incurs usage charges"]
async fn fable_tool_roundtrips_and_history_binding() {
    llmshim::env::load_all();
    let router = llmshim::router::Router::from_env();
    for model in ["anthropic/claude-fable-5", "anthropic/claude-fable-5-1"] {
        let mut input = json!({"model": model, "max_tokens": 2048,
            "messages": [{"role": "system", "content": "Use the weather tool when asked about weather. Reply briefly."},
                {"role": "user", "content": "Call lookup_weather for Paris. After receiving the result, repeat the condition."}],
            "tools": [{"type": "function", "function": {"name": "lookup_weather", "description": "Look up weather for a city.",
                "parameters": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"], "additionalProperties": false}}}],
            "tool_choice": if model.ends_with("5-1") { "auto" } else { "required" },
            "reasoning_effort": "low"
        });
        if model.ends_with("5-1") {
            input["messages"][1]["content"] = json!("Work out whether 1739 times 2843 is divisible by 7, then call lookup_weather for Paris. After receiving the tool result, reply with the weather condition only.");
            input["x-anthropic"] = json!({
                "thinking": {"type": "adaptive", "display": "summarized", "block_binding": {"prefix_mismatch_behavior": "error"}},
                "output_config": {"effort": "high"}, "extra_betas": ["thinking-binding-controls-2026-08-01"]
            });
        }
        let result = llmshim::completion(&router, &input).await.unwrap();
        let assistant = result["choices"][0]["message"].clone();
        let call = assistant["tool_calls"][0].clone();
        assert_eq!(call["function"]["name"], "lookup_weather");
        assert_eq!(
            serde_json::from_str::<Value>(call["function"]["arguments"].as_str().unwrap()).unwrap()
                ["city"],
            "Paris"
        );
        if model.ends_with("5-1") {
            assert!(
                assistant["reasoning"][0]["signature"]
                    .as_str()
                    .is_some_and(|s| !s.is_empty())
                    || assistant["reasoning"][0]["data"]
                        .as_str()
                        .is_some_and(|s| !s.is_empty()),
                "missing thinking block on {model}; response fields: {:?}",
                assistant.as_object().unwrap().keys().collect::<Vec<_>>()
            );
        }
        input["messages"].as_array_mut().unwrap().push(assistant);
        input["messages"]
            .as_array_mut()
            .unwrap()
            .push(json!({"role": "tool", "tool_call_id": call["id"], "content": "Sunny"}));
        input["tool_choice"] = json!("none");
        if model.ends_with("5-1") {
            input["messages"].as_array_mut().unwrap().push(json!({"role": "system", "content": "For this turn, reply with only the weather condition."}));
        }
        let reply = llmshim::completion(&router, &input).await.unwrap();
        assert!(reply["choices"][0]["message"]["content"]
            .as_str()
            .unwrap()
            .to_lowercase()
            .contains("sunny"));
        println!("PASS {model} tool-result round trip");
        if model.ends_with("5-1") {
            input["messages"][0]["content"] =
                json!("Changed initial instructions, invalidating signed thinking.");
            let error = llmshim::completion(&router, &input).await.unwrap_err();
            assert!(matches!(
                error,
                llmshim::error::ShimError::ProviderError { status: 400, .. }
            ));
            assert!(error.to_string().contains("signature") || error.to_string().contains("bound"));
            println!("PASS Fable 5.1 binding enforcement rejects an edited prefix");
        }
    }
}

#[cfg(feature = "proxy")]
#[tokio::test]
#[ignore = "requires provider API keys and incurs usage charges"]
async fn current_models_proxy_routes() {
    llmshim::env::load_all();
    let router = llmshim::router::Router::from_env();
    let listener = tokio::net::TcpListener::bind("127.0.0.1:0").await.unwrap();
    let base = format!("http://{}", listener.local_addr().unwrap());
    let server = tokio::spawn(async move {
        axum::serve(listener, llmshim::proxy::app(router, None))
            .await
            .unwrap();
    });
    // Abort the task even if an assertion fails.
    struct ServerGuard(tokio::task::AbortHandle);
    impl Drop for ServerGuard {
        fn drop(&mut self) {
            self.0.abort();
        }
    }
    let _guard = ServerGuard(server.abort_handle());
    let client = reqwest::Client::builder()
        .timeout(std::time::Duration::from_secs(120))
        .build()
        .unwrap();
    let catalog: Value = client
        .get(format!("{base}/v1/models"))
        .send()
        .await
        .unwrap()
        .json()
        .await
        .unwrap();
    for model in [
        "gemini/gemini-3.8-flash",
        "anthropic/claude-opus-5",
        "anthropic/claude-fable-5",
        "anthropic/claude-fable-5-1",
        "xai/grok-4.6",
    ] {
        let advertised = catalog["models"]
            .as_array()
            .unwrap()
            .iter()
            .any(|entry| entry["id"] == model);
        // Fable 5 remains callable by explicit ID but is superseded in discovery.
        assert_eq!(advertised, model != "anthropic/claude-fable-5");
        let response = client.post(format!("{base}/v1/chat")).json(&json!({"model": model,
            "messages": [{"role": "user", "content": "Reply with only pong."}],
            "config": {"reasoning_effort": "none", "max_tokens": 512, "temperature": 0.5, "top_p": 0.8, "top_k": 10}}))
            .send().await.unwrap();
        assert_eq!(response.status(), 200, "{model}");
        let response: Value = response.json().await.unwrap();
        assert_eq!(response["provider"], model.split_once('/').unwrap().0);
        assert!(response["message"]["content"]
            .as_str()
            .unwrap()
            .to_lowercase()
            .contains("pong"));
        println!("PASS {model} real HTTP proxy request (advertised: {advertised})");
    }
}