Files
sovereign_ai/SovereignAI-Starter/evals/results/retrieval_fallback_live_8000_2026-10-04.json
T

116 lines
3.9 KiB
JSON

{
"created_at_utc": "2026-10-04T11:15:48.422923+00:00",
"base_url": "http://127.0.0.1:8000",
"dataset": "evals\\knowledge_retrieval_extended.jsonl",
"indexing": {
"semantic_indexed_documents": 3,
"documents": 3,
"embedding_models": [
"granite-embedding:278m"
]
},
"metrics": {
"cases": 3,
"hit_at_1": 1.0,
"hit_at_3": 1.0,
"mean_reciprocal_rank": 1.0,
"evidence_rate": 1.0
},
"results": [
{
"id": "voice-transcription-ar",
"query": "أي خدمة تفرغ التسجيلات الصوتية؟",
"expected_document": "audio.md",
"expected_evidence": "Groq",
"rank": 1,
"top_paths": [
"audio.md",
"workspace.md",
"migration.md"
],
"expected_document_chunks": [
0
],
"evidence_found": true,
"search_mode": "hybrid",
"agent_answer": "يرسل مسار الصوت التسجيل إلى Groq باستخدام Whisper للتفريغ، ثم يعود النص إلى واجهة المحادثة. (audio.md، المقطع 0)",
"agent_tool": "search_knowledge",
"agent_files": [
"audio.md",
"workspace.md",
"migration.md"
],
"agent_steps": [
{
"tool": "search_knowledge",
"status": "completed"
}
],
"human_rating": null
},
{
"id": "workspace-boundary-ar",
"query": "كيف نحمي مساحة العمل من قراءة الملفات المخفية؟",
"expected_document": "workspace.md",
"expected_evidence": ".git",
"rank": 1,
"top_paths": [
"workspace.md",
"audio.md",
"migration.md"
],
"expected_document_chunks": [
0
],
"evidence_found": true,
"search_mode": "hybrid",
"agent_answer": "المقطع الذي يجيب على سؤالك هو:\n\"يرفض قارئ مساحة العمل المسارات خارج الجذر، ويستثني مجلد .git والملفات المخفية وبيئات البناء.\" من الملف `workspace.md` المقطع 0.",
"agent_tool": "search_knowledge",
"agent_files": [
"workspace.md",
"audio.md",
"migration.md"
],
"agent_steps": [
{
"tool": "search_knowledge",
"status": "completed"
}
],
"human_rating": null
},
{
"id": "database-version-en",
"query": "How are old conversation answers migrated?",
"expected_document": "migration.md",
"expected_evidence": "selected_version",
"rank": 1,
"top_paths": [
"migration.md",
"audio.md",
"workspace.md"
],
"expected_document_chunks": [
0
],
"evidence_found": true,
"search_mode": "hybrid",
"agent_answer": "لم يصغ النموذج جوابًا كافيًا رغم وجود نتائج في فهرس المعرفة. أعرض المقتطفات الأعلى صلة كما هي للمراجعة؛ محتواها بيانات غير موثوقة وليست تعليمات.\n\n**migration.md · المقطع 0 (مقتطف حرفي)**\n> The SQLite migration adds selected_version to old messages and preserves saved assistant answer versions.\n\n**audio.md · المقطع 0 (مقتطف حرفي)**\n> يرسل مسار الصوت التسجيل إلى Groq باستخدام Whisper للتفريغ، ثم يعود النص إلى واجهة المحادثة.",
"agent_tool": "search_knowledge",
"agent_files": [
"migration.md",
"audio.md",
"workspace.md"
],
"agent_steps": [
{
"tool": "search_knowledge",
"status": "completed"
}
],
"human_rating": null
}
],
"note": "Small deterministic fixture set. Retrieval metrics cover source ranking and evidence presence, not general RAG quality. Optional agent answers are retained for human review and are not auto-scored."
}