llmshim 0.1.23

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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300

/// Tests for multi-model conversation support (Cursor-style provider switching).
/// Verifies that conversation history from one provider can be sent to another
/// without breaking, and that provider-specific fields are sanitized.
use llmshim::provider::Provider;
use llmshim::providers::anthropic::Anthropic;
use llmshim::providers::openai::OpenAi;
use serde_json::json;

fn openai() -> OpenAi {
    OpenAi::new("test-key".into())
}

fn anthropic() -> Anthropic {
    Anthropic::new("test-key".into())
}

// ============================================================
// Basic multi-model conversation
// ============================================================

#[test]
fn openai_response_sent_to_anthropic() {
    let a = anthropic();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "What is Rust?"},
            {"role": "assistant", "content": "Rust is a systems programming language."},
            {"role": "user", "content": "How does it compare to Go?"},
        ],
        "max_tokens": 100,
    });
    let result = a.transform_request("claude-sonnet-4-6", &req).unwrap();
    let messages = result.body["messages"].as_array().unwrap();
    assert_eq!(messages.len(), 3);
    assert_eq!(messages[1]["role"], "assistant");
    assert_eq!(
        messages[1]["content"],
        "Rust is a systems programming language."
    );
}

#[test]
fn anthropic_response_sent_to_openai() {
    let o = openai();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "What is Rust?"},
            {"role": "assistant", "content": "Rust is a systems programming language."},
            {"role": "user", "content": "How does it compare to Go?"},
        ],
    });
    let result = o.transform_request("gpt-5.4", &req).unwrap();
    // Responses API uses "input" not "messages"
    let input = result.body["input"].as_array().unwrap();
    assert_eq!(input.len(), 3);
    assert_eq!(
        input[1]["content"],
        "Rust is a systems programming language."
    );
}

// ============================================================
// reasoning_content sanitization
// ============================================================

#[test]
fn reasoning_content_stripped_when_sent_to_openai() {
    let o = openai();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "Solve this math problem."},
            {
                "role": "assistant",
                "content": "The answer is 42.",
                "reasoning_content": "Let me think step by step..."
            },
            {"role": "user", "content": "Are you sure?"},
        ],
    });
    let result = o.transform_request("gpt-5.4", &req).unwrap();
    let input = result.body["input"].as_array().unwrap();
    assert!(input[1].get("reasoning_content").is_none());
    assert_eq!(input[1]["content"], "The answer is 42.");
}

#[test]
fn reasoning_content_stripped_when_sent_to_anthropic() {
    let a = anthropic();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "Solve this."},
            {
                "role": "assistant",
                "content": "The answer is 42.",
                "reasoning_content": "Step by step reasoning..."
            },
            {"role": "user", "content": "Explain more."},
        ],
        "max_tokens": 100,
    });
    let result = a.transform_request("claude-sonnet-4-6", &req).unwrap();
    let messages = result.body["messages"].as_array().unwrap();
    assert!(messages[1].get("reasoning_content").is_none());
    assert_eq!(messages[1]["content"], "The answer is 42.");
}

// ============================================================
// Tool calls across providers
// ============================================================

#[test]
fn openai_tool_call_history_sent_to_anthropic() {
    let a = anthropic();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "What's the weather in Paris?"},
            {
                "role": "assistant",
                "content": null,
                "tool_calls": [{
                    "id": "call_abc",
                    "type": "function",
                    "function": {"name": "get_weather", "arguments": "{\"city\":\"Paris\"}"}
                }]
            },
            {"role": "tool", "tool_call_id": "call_abc", "content": "Sunny, 22C"},
            {"role": "user", "content": "Thanks!"},
        ],
        "max_tokens": 100,
    });
    let result = a.transform_request("claude-sonnet-4-6", &req).unwrap();
    let messages = result.body["messages"].as_array().unwrap();
    // tool_calls → tool_use content blocks
    let content = messages[1]["content"].as_array().unwrap();
    assert_eq!(content[0]["type"], "tool_use");
    assert_eq!(content[0]["name"], "get_weather");
    // tool result → user with tool_result
    assert_eq!(messages[2]["role"], "user");
    assert_eq!(messages[2]["content"][0]["type"], "tool_result");
}

#[test]
fn anthropic_tool_response_format_works_for_openai() {
    let a = anthropic();
    let resp = json!({
        "id": "msg_123",
        "content": [{"type": "tool_use", "id": "tu_456", "name": "search", "input": {"query": "rust"}}],
        "stop_reason": "tool_use",
        "usage": {"input_tokens": 10, "output_tokens": 5}
    });
    let normalized = a.transform_response("claude-sonnet-4-6", resp).unwrap();

    let o = openai();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "search for rust"},
            normalized["choices"][0]["message"],
            {"role": "tool", "tool_call_id": "tu_456", "content": "Results: ..."},
            {"role": "user", "content": "summarize"},
        ],
    });
    let result = o.transform_request("gpt-5.4", &req).unwrap();
    let input = result.body["input"].as_array().unwrap();
    // Assistant tool_calls are translated to Responses API function_call items.
    assert_eq!(input[1]["type"], "function_call");
    assert_eq!(input[1]["name"], "search");
    assert_eq!(input[1]["call_id"], "tu_456");
    // Tool result is translated to function_call_output.
    assert_eq!(input[2]["type"], "function_call_output");
    assert_eq!(input[2]["call_id"], "tu_456");
}

// ============================================================
// System/developer role across providers
// ============================================================

#[test]
fn developer_role_from_openai_handled_by_anthropic() {
    let a = anthropic();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "developer", "content": "You are a helpful assistant."},
            {"role": "user", "content": "Hi"},
            {"role": "assistant", "content": "Hello!"},
            {"role": "user", "content": "Follow up"},
        ],
        "max_tokens": 100,
    });
    let result = a.transform_request("claude-sonnet-4-6", &req).unwrap();
    assert_eq!(result.body["system"], "You are a helpful assistant.");
    let messages = result.body["messages"].as_array().unwrap();
    assert_eq!(messages.len(), 3);
}

#[test]
fn system_role_becomes_instructions_for_openai() {
    let o = openai();
    let req = json!({
        "model": "x",
        "messages": [
            {"role": "system", "content": "Be concise."},
            {"role": "user", "content": "What is 2+2?"},
            {"role": "assistant", "content": "4"},
            {"role": "user", "content": "And 3+3?"},
        ],
    });
    let result = o.transform_request("gpt-5.4", &req).unwrap();
    // System message → instructions in Responses API
    assert_eq!(result.body["instructions"], "Be concise.");
    let input = result.body["input"].as_array().unwrap();
    assert_eq!(input.len(), 3); // system extracted out
    assert_eq!(input[0]["role"], "user");
}

// ============================================================
// Multi-hop: A → B → A
// ============================================================

#[test]
fn round_trip_openai_anthropic_openai() {
    let o = openai();
    let a = anthropic();

    // Simulate Anthropic response with thinking
    let anthropic_resp = json!({
        "id": "msg_789",
        "content": [
            {"type": "thinking", "thinking": "Let me think of a joke...", "signature": "sig"},
            {"type": "text", "text": "Why did the crab never share? Because he's shellfish!"}
        ],
        "stop_reason": "end_turn",
        "usage": {"input_tokens": 30, "output_tokens": 20}
    });
    let normalized_resp = a
        .transform_response("claude-sonnet-4-6", anthropic_resp)
        .unwrap();
    assert_eq!(
        normalized_resp["choices"][0]["message"]["reasoning_content"],
        "Let me think of a joke..."
    );

    // Send to OpenAI with reasoning_content in history — should be stripped
    let req3 = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "Tell me a joke"},
            normalized_resp["choices"][0]["message"],
            {"role": "user", "content": "That was funny!"},
        ],
    });
    let result3 = o.transform_request("gpt-5.4", &req3).unwrap();
    let input = result3.body["input"].as_array().unwrap();
    assert!(input[1].get("reasoning_content").is_none());
    assert_eq!(
        input[1]["content"],
        "Why did the crab never share? Because he's shellfish!"
    );
}

#[test]
fn round_trip_anthropic_openai_anthropic() {
    let o = openai();
    let a = anthropic();

    // Turn 2: Send to OpenAI
    let req2 = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "Hi"},
            {"role": "assistant", "content": "Hello from Claude!"},
            {"role": "user", "content": "Now in Spanish"},
        ],
    });
    let result2 = o.transform_request("gpt-5.4", &req2).unwrap();
    assert_eq!(result2.body["input"].as_array().unwrap().len(), 3);

    // Turn 3: Back to Anthropic
    let req3 = json!({
        "model": "x",
        "messages": [
            {"role": "user", "content": "Hi"},
            {"role": "assistant", "content": "Hello from Claude!"},
            {"role": "user", "content": "Now in Spanish"},
            {"role": "assistant", "content": "Hola desde Claude!"},
            {"role": "user", "content": "And in French?"},
        ],
        "max_tokens": 100,
    });
    let result3 = a.transform_request("claude-sonnet-4-6", &req3).unwrap();
    let messages = result3.body["messages"].as_array().unwrap();
    assert_eq!(messages.len(), 5);
    assert_eq!(messages[3]["content"], "Hola desde Claude!");
}