project-rag 0.1.0

RAG-based codebase indexing and semantic search - dual purpose library and MCP server
Documentation 
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

# Project-RAG Configuration File
#
# Copy this file to one of these locations:
#   - Linux:   ~/.config/project-rag/config.toml
#   - macOS:   ~/Library/Application Support/project-rag/config.toml
#   - Windows: %APPDATA%\project-rag\config.toml
#
# Configuration priority: CLI args > Environment variables > Config file > Defaults

[vector_db]
# Vector database backend: "lancedb" (default, embedded) or "qdrant" (external server)
backend = "lancedb"

# LanceDB database directory path (only used if backend = "lancedb")
# Default: Platform-specific data directory + "/project-rag/lancedb"
# lancedb_path = "/custom/path/to/lancedb"

# Qdrant server URL (only used if backend = "qdrant")
# Default: "http://localhost:6334"
# qdrant_url = "http://localhost:6334"

# Collection name for storing embeddings
# Default: "code_embeddings"
collection_name = "code_embeddings"

[embedding]
# Embedding model to use. Supported models:
#   - "all-MiniLM-L6-v2" (default, 384 dims, fast)
#   - "all-MiniLM-L12-v2" (384 dims, more accurate)
#   - "BAAI/bge-base-en-v1.5" (768 dims, high quality)
#   - "BAAI/bge-small-en-v1.5" (384 dims, balanced)
model_name = "all-MiniLM-L6-v2"

# Number of texts to process in each embedding batch
# Larger batches are faster but use more memory
# Default: 32
batch_size = 32

# Timeout in seconds for embedding generation
# Default: 30
timeout_secs = 30

[indexing]
# Chunk size for FixedLines chunking strategy (lines per chunk)
# Default: 50
chunk_size = 50

# Maximum file size to index in bytes (1 MB = 1048576)
# Files larger than this will be skipped
# Default: 1048576 (1 MB)
max_file_size = 1048576

# File patterns to include (empty = include all)
# Example: include_patterns = ["**/*.rs", "**/*.py"]
include_patterns = []

# File patterns to exclude (always applied)
# Default: ["target", "node_modules", ".git", "dist", "build"]
exclude_patterns = ["target", "node_modules", ".git", "dist", "build"]

[search]
# Minimum similarity score for search results (0.0 to 1.0)
# Higher values return fewer, more relevant results
# Default: 0.7
min_score = 0.7

# Maximum number of search results to return
# Default: 10
limit = 10

# Enable hybrid search (vector similarity + BM25 keyword matching)
# Combining both methods generally improves search quality
# Default: true
hybrid = true

[cache]
# Path to hash cache file for incremental indexing
# Default: Platform-specific cache directory + "/project-rag/hash_cache.json"
# hash_cache_path = "/custom/path/to/hash_cache.json"

# Path to git cache file for git history indexing
# Default: Platform-specific cache directory + "/project-rag/git_cache.json"
# git_cache_path = "/custom/path/to/git_cache.json"

# Environment Variable Overrides
# ==============================
# You can override any configuration value using environment variables:
#
#   PROJECT_RAG_DB_BACKEND         - Vector database backend
#   PROJECT_RAG_LANCEDB_PATH       - LanceDB path
#   PROJECT_RAG_QDRANT_URL         - Qdrant server URL
#   PROJECT_RAG_MODEL              - Embedding model name
#   PROJECT_RAG_BATCH_SIZE         - Embedding batch size
#   PROJECT_RAG_MIN_SCORE          - Minimum search score
#
# Example:
#   export PROJECT_RAG_MODEL="BAAI/bge-base-en-v1.5"
#   export PROJECT_RAG_MIN_SCORE="0.8"