reasonkit-core 0.1.8

The Reasoning Engine — Auditable Reasoning for Production AI | Rust-Native | Turn Prompts into Protocols
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

# ReasonKit Core Configuration
# Default settings for the knowledge base

[general]
# Data directory (can be overridden via CLI or environment)
data_dir = "./data"
# Log level: trace, debug, info, warn, error
log_level = "info"

[storage]
# Storage backend: "qdrant", "local"
backend = "qdrant"

[storage.qdrant]
# Qdrant connection settings
host = "localhost"
port = 6334
grpc_port = 6333
# Use embedded mode (no external server needed)
embedded = true
# Collection name for documents
collection = "reasonkit_docs"
# Vector dimensions (depends on embedding model)
vector_size = 1024
# Distance metric: "Cosine", "Euclid", "Dot"
distance = "Cosine"
# Quantization for compression (reduces memory by 4x)
quantization = true

[storage.local]
# Local JSON storage settings
documents_path = "./data/documents"
index_path = "./data/indexes"

[embedding]
# Embedding backend: "api", "local"
backend = "api"

[embedding.api]
# API provider: "openai", "anthropic", "cohere", "voyage"
provider = "openai"
# Model for dense embeddings
model = "text-embedding-3-small"
# Dimensions (some models allow customization)
dimensions = 1536
# Batch size for embedding requests
batch_size = 100

[embedding.local]
# Local embedding using ONNX
model_path = "./models/bge-m3-onnx"
device = "cpu"  # or "cuda"

[indexing]
# BM25 index settings
bm25_enabled = true
# HNSW parameters
hnsw_m = 16
hnsw_ef_construction = 200

[processing]
# Chunk settings
chunk_size = 512  # tokens
chunk_overlap = 50  # tokens
# Minimum chunk size to avoid tiny fragments
min_chunk_size = 100

[retrieval]
# Default number of results
top_k = 10
# Hybrid search alpha (0 = BM25 only, 1 = vector only)
alpha = 0.7
# Minimum score threshold
min_score = 0.0
# Enable reranking (requires ColBERT or cross-encoder)
rerank = false

[raptor]
# RAPTOR tree settings
enabled = false
# Maximum tree depth
max_depth = 3
# Cluster size for summarization
cluster_size = 10
# LLM for summarization
summarizer = "claude-3-haiku"

[server]
# API server settings
host = "127.0.0.1"
port = 8080
# Enable CORS
cors = true
# Request timeout (seconds)
timeout = 30