llmtrace/config.deploy.yaml at main · techlab-innov/llmtrace · GitHub

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
# LLMTrace Proxy -- Basilica deployment config (4-model ensemble)
# Storage and upstream are overridden by LLMTRACE_* env vars at runtime.

listen_addr: "0.0.0.0:8080"
upstream_url: "https://api.openai.com"

storage:
  profile: "memory"
  auto_migrate: true

logging:
  level: "info"
  format: "json"

timeout_ms: 60000
connection_timeout_ms: 5000
max_connections: 1000

enable_tls: false
enable_security_analysis: true
enable_trace_storage: true
enable_streaming: true
max_request_size_bytes: 52428800
security_analysis_timeout_ms: 10000
trace_storage_timeout_ms: 10000

rate_limiting:
  enabled: true
  requests_per_second: 200
  burst_size: 400
  window_seconds: 60

circuit_breaker:
  enabled: true
  failure_threshold: 10
  recovery_timeout_ms: 30000
  half_open_max_calls: 3

security_analysis:
  ml_enabled: true
  ml_model: "protectai/deberta-v3-base-prompt-injection-v2"
  ml_threshold: 0.8
  ml_cache_dir: "/home/llmtrace/.cache/llmtrace/models"
  ml_preload: true
  ml_download_timeout_seconds: 600
  jailbreak_enabled: true
  jailbreak_threshold: 0.7
  operating_point: "balanced"
  over_defence: false
  injecguard_enabled: true
  injecguard_model: "leolee99/InjecGuard"
  injecguard_threshold: 0.85
  piguard_enabled: true
  piguard_model: "leolee99/PIGuard"
  piguard_threshold: 0.85

streaming_analysis:
  enabled: true
  token_interval: 50
  output_enabled: true

health_check:
  enabled: true
  path: "/health"
  interval_seconds: 10
  timeout_ms: 5000
  retries: 3

shutdown:
  timeout_seconds: 30