llm-scanner 0.1.5 → 0.1.6
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.js +0 -0
- package/dist/judge.js +10 -3
- package/package.json +1 -1
package/dist/index.js
CHANGED
|
File without changes
|
package/dist/judge.js
CHANGED
|
@@ -89,29 +89,36 @@ function isAuthenticationError(err) {
|
|
|
89
89
|
return false;
|
|
90
90
|
}
|
|
91
91
|
async function judge(attack, response) {
|
|
92
|
+
console.log("🔥 JUDGE STARTED:", attack.type ?? attack.category);
|
|
92
93
|
const key = process.env.OPENAI_API_KEY;
|
|
93
94
|
const prompt = TEMPLATE.replace("{{attack}}", attack.prompt)
|
|
94
95
|
.replace("{{response}}", response)
|
|
95
96
|
.replace("{{failSignal}}", attack.failSignal);
|
|
96
97
|
try {
|
|
97
98
|
const client = new openai_1.default({ apiKey: key });
|
|
99
|
+
console.log("🚀 Calling OpenAI judge...");
|
|
98
100
|
const completion = await client.chat.completions.create({
|
|
99
101
|
model: "gpt-4o-mini",
|
|
100
102
|
temperature: 0,
|
|
101
103
|
messages: [{ role: "user", content: prompt }],
|
|
102
104
|
});
|
|
105
|
+
console.log("✅ OpenAI responded");
|
|
103
106
|
const content = completion.choices[0]?.message?.content ?? "";
|
|
104
107
|
const parsed = parseJudgeJson(content);
|
|
105
108
|
if (!parsed) {
|
|
106
|
-
|
|
109
|
+
console.error("❌ Judge parsing failed. Raw output:", content);
|
|
110
|
+
return { verdict: "SKIP", reason: "Judge parsing failed" };
|
|
107
111
|
}
|
|
108
112
|
return parsed;
|
|
109
113
|
}
|
|
110
114
|
catch (e) {
|
|
111
|
-
console.error("Judge
|
|
115
|
+
console.error("❌ Judge API failed:", e);
|
|
112
116
|
if (isAuthenticationError(e)) {
|
|
113
117
|
throw new Error("OpenAI authentication failed (401). Check OPENAI_API_KEY in your .env file.");
|
|
114
118
|
}
|
|
115
|
-
return {
|
|
119
|
+
return {
|
|
120
|
+
verdict: "SKIP",
|
|
121
|
+
reason: "Judge API call failed — result not reliable",
|
|
122
|
+
};
|
|
116
123
|
}
|
|
117
124
|
}
|