Coverage for src/local_deep_research/web/warning_checks/hardware.py: 100%

1"""Pure hardware/settings-based warning checks. 

2 

3These functions have zero dependencies on Flask or SQLAlchemy — 

4they take primitive values and return warning dicts (or None). 

5""" 

6 

7from typing import Optional 

8 

9LOCAL_PROVIDERS = frozenset({"ollama", "llamacpp", "lmstudio"}) 

10 

11 

12def check_high_context( 

13 provider: str, local_context: int, dismissed: bool 

14) -> Optional[dict]: 

15 """Return a high_context warning dict if context exceeds 8192 for a local provider.""" 

16 if provider not in LOCAL_PROVIDERS: 

17 return None 

18 if local_context <= 8192: 

19 return None 

20 if dismissed: 

21 return None 

22 

23 return { 

24 "type": "high_context", 

25 "icon": "⚠️", 

26 "title": "High Context Warning", 

27 "message": ( 

28 f"Context size ({local_context:,} tokens) requires sufficient VRAM. " 

29 f"This is recommended for the langgraph-agent strategy. " 

30 f"If you experience slowdowns, reduce context size in settings " 

31 f"and switch to the source-based strategy instead. " 

32 f"Tip: check the metrics page in each research history entry " 

33 f"to monitor actual token usage and VRAM consumption." 

34 ), 

35 "dismissKey": "app.warnings.dismiss_high_context", 

36 "actionUrl": "/metrics/context-overflow", 

37 "actionLabel": "View context metrics", 

38 } 

39 

40 

41def check_model_mismatch( 

42 provider: str, model: str, local_context: int, dismissed: bool 

43) -> Optional[dict]: 

44 """Return a model_mismatch warning dict for large models with high context.""" 

45 if not model: 

46 return None 

47 if provider not in LOCAL_PROVIDERS: 

48 return None 

49 if "70b" not in model.lower(): 

50 return None 

51 if local_context <= 8192: 

52 return None 

53 if dismissed: 

54 return None 

55 

56 return { 

57 "type": "model_mismatch", 

58 "icon": "🧠", 

59 "title": "Model & Context Warning", 

60 "message": ( 

61 f"Large model ({model}) with high context ({local_context:,}) " 

62 f"may exceed VRAM. Consider reducing context size or upgrading " 

63 f"GPU memory." 

64 ), 

65 "dismissKey": "app.warnings.dismiss_model_mismatch", 

66 # Shares the high_context destination on purpose: this warning fires on 

67 # large-model + high-context together, and the actionable lever is 

68 # reducing context size. The "Token Usage & Context Analytics" page is 

69 # where the user sees actual token/context usage to make that call — 

70 # there is no separate GPU/VRAM analytics page. 

71 "actionUrl": "/metrics/context-overflow", 

72 "actionLabel": "View context metrics", 

73 } 

74 

75 

76def check_legacy_server_config(dismissed: bool) -> Optional[dict]: 

77 """Return a warning only if server_config.json has non-default values.""" 

78 from ..server_config import has_legacy_customizations 

79 

80 if dismissed: 

81 return None 

82 if not has_legacy_customizations(): 

83 return None 

84 return { 

85 "type": "legacy_server_config", 

86 "icon": "ℹ️", 

87 "title": "server_config.json Detected", 

88 "message": ( 

89 "A server_config.json file was found with non-default settings. " 

90 "Environment variables are the preferred configuration method. " 

91 "See the documentation for migration details." 

92 ), 

93 "dismissKey": "app.warnings.dismiss_legacy_config", 

94 }