Analytics & Observability

1
[
2
    {
3
        'run_id': UUID('27f124ad-6f70-4641-89ab-f346dc9d1c2f'),
4
        'run_type': 'rag',
5
        'entries': [
6
            {'key': 'search_query', 'value': 'Who is Aristotle?'},
7
            {'key': 'search_latency', 'value': '0.39'},
8
            {'key': 'search_results', 'value': '["{\\"id\\":\\"7ed3a01c-88dc-5a58-a68b-6e5d9f292df2\\",...}"]'},
9
            {'key': 'rag_generation_latency', 'value': '3.79'},
10
            {'key': 'llm_response', 'value': 'Aristotle (Greek: Ἀριστοτέλης Aristotélēs; 384–322 BC) was...'}
11
        ]
12
    },
13
    # More log entries...
14
]

1
app = R2R()
2
3
# Perform some searches / RAG completions
4
# ...
5
6
# Get the latest logs
7
logs = app.logs()
8
print(logs)

1
{
2
    'results': {
3
        'filtered_logs': {
4
            'search_latencies': [
5
                {
6
                    'timestamp': '2024-06-20 21:29:06',
7
                    'log_id': UUID('0f28063c-8b87-4934-90dc-4cd84dda5f5c'),
8
                    'key': 'search_latency',
9
                    'value': '0.66',
10
                    'rn': 3
11
                },
12
                ...
13
            ]
14
        },
15
        'search_latencies': {
16
            'Mean': 0.734,
17
            'Median': 0.523,
18
            'Mode': 0.495,
19
            'Standard Deviation': 0.213,
20
            'Variance': 0.0453
21
        }
22
    }
23
}

1
from r2r import FilterCriteria, AnalysisTypes
2
3
filter_criteria = FilterCriteria(filters={"search_latencies": "search_latency"})
4
analysis_types = AnalysisTypes(analysis_types={"search_latencies": ["basic_statistics", "search_latency"]})
5
6
analytics_results = app.analytics(filter_criteria, analysis_types)
7
print(analytics_results)

1
# Analyze RAG latencies
2
rag_filter = FilterCriteria(filters={"rag_latencies": "rag_generation_latency", "rag_eval": "rag_eval_metric"})
3
rag_analysis = AnalysisTypes(analysis_types={"rag_latencies": ["basic_statistics", "rag_generation_latency"]})
4
rag_analytics = app.analytics(rag_filter, rag_analysis)
5
6
# Track usage patterns by user
7
user_filter = FilterCriteria(filters={"user_patterns": "user_id"})
8
user_analysis = AnalysisTypes(analysis_types={"user_patterns": ["bar_chart", "user_id"]})
9
user_analytics = app.analytics(user_filter, user_analysis)
10
11
# Monitor error rates
12
error_filter = FilterCriteria(filters={"error_rates": "error"})
13
error_analysis = AnalysisTypes(analysis_types={"error_rates": ["basic_statistics", "error"]})
14
error_analytics = app.analytics(error_filter, error_analysis)

1
import random
2
from r2r import R2R, GenerationConfig
3
4
app = R2R()
5
6
# List of sample queries
7
queries = [
8
    "What is artificial intelligence?",
9
    "Explain machine learning.",
10
    "How does natural language processing work?",
11
    "What are neural networks?",
12
    "Describe deep learning.",
13
    # Add more queries as needed
14
]
15
16
# Perform random searches
17
for _ in range(1000):
18
    query = random.choice(queries)
19
    app.rag(query, GenerationConfig(model="openai/gpt-4o-mini"))
20
21
print("Preloading complete. You can now run analytics on this data.")

1
user_id = "your_user_id_here"
2
3
user_filter = FilterCriteria(filters={"user_analytics": "user_id"})
4
user_analysis = AnalysisTypes(analysis_types={
5
    "user_analytics": ["basic_statistics", "user_id"],
6
    "user_search_latencies": ["basic_statistics", "search_latency"]
7
})
8
9
user_analytics = app.analytics(user_filter, user_analysis)
10
print(f"Analytics for user {user_id}:")
11
print(user_analytics)