vault-search-mcp/config.example.yaml at main · everton-dgn/vault-search-mcp · 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
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
# Canonical vault-search-mcp configuration.
#
# Copy this file to config.yaml and change only the required fields. Relative
# paths are resolved from the configuration file's directory. The effective
# configuration is captured on first import; restart the MCP process and daemon
# after changing it. config.yaml and config.yml are ignored by Git by default.

paths:
  vault_path: "vaults/obsidian_vault"
  data_dir: "data"
  lancedb_table: "vault_chunks"

search:
  candidates: 50
  candidates_max: 500
  candidates_multiplier: 2
  top_k: 10
  top_k_min: 1
  top_k_max: 100
  score_precision: 4
  list_notes_default_limit: 500
  list_notes_max_limit: 5000

indexing:
  batch_size: 500
  workers: null
  max_chunks_per_note: 1000
  extensions:
    - ".md"
    - ".mdx"
    - ".txt"
    - ".pdf"
    - ".canvas"
  ignored_folders:
    - ".obsidian"
    - ".smart-env"
    - ".trash"
    - ".git"

fts:
  # The language-neutral default works with multilingual vaults. Set a language
  # such as "English" only when language-specific stemming is intentional.
  language: null

prewarm:
  enabled: true
  max_ram_percent: 0.25
  min_available_ram: 2147483648
  bytes_per_chunk: 5120

embedding:
  model: "BAAI/bge-m3"
  reranker_model: "cross-encoder/ms-marco-MiniLM-L-6-v2"
  # null plus auto select compatible precision and backend settings at runtime.
  use_fp16: null
  device: "auto"
  batch_size: 32
  query_max_length: 512
  corpus_max_length: 1024
  dimension: 1024
  reranker_normalize: true
  idle_timeout: 1800

chunking:
  size: 2000
  overlap: 200
  header_levels: 4
  separators:
    - "\n\n"
    - "\n"
    - ". "
    - " "

security:
  max_query_length: 10000
  max_content_size: 10485760
  max_path_length: 500
  max_frontmatter_keys: 100

watcher:
  debounce: 2.0
  poll_factor: 2
  thread_join_timeout: 5

pdf:
  ocr_enabled: true
  ocr_languages: "eng"
  ocr_dpi: 150

vector_index:
  min_chunks: 5000
  auto_create: true
  index_type: "IVF_PQ"
  num_sub_vectors: 128
  distance_type: "cosine"

navigation:
  folder_tree_max_depth: 10
  folder_tree_max_depth_limit: 50

daemon:
  # Only loopback hosts are accepted. The internal HTTP protocol is not public.
  host: "127.0.0.1"
  port: 9847
  timeout: 120.0
  auto_use: true
  max_request_bytes: 2097152
  max_texts: 512
  max_text_length: 100000

frontmatter:
  enabled: false
  mode: "lenient"
  allow_extra_fields: true
  ai:
    enabled: false
    allow_external_processing: false
    provider: null
    allow_defer_required_on_create: true
    # Only {model} is accepted. Note content is sent through stdin.
    command: []
    # Set provider-specific model identifiers only when enabling this feature.
    primary_model: null
    fallback_model: null
    timeout_seconds: 8.0
    max_attempts: 2
    max_note_chars: 12000
  schema:
    id:
      type: "uuid"
      on_missing: "auto"
    created_at:
      type: "datetime"
      on_missing: "auto"
      aliases: ["created", "date"]