-
-
Notifications
You must be signed in to change notification settings - Fork 5
Expand file tree
/
Copy pathvectorizers.yaml
More file actions
42 lines (40 loc) · 1.55 KB
/
vectorizers.yaml
File metadata and controls
42 lines (40 loc) · 1.55 KB
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
vectorizers:
# 3. Appunti Obsidian / Markdown
# Usa strategia "markdown" (taglia sugli header ##)
- name: "personal_notes"
kektor_index: "knowledge_base"
schedule: "30s"
source:
type: "filesystem"
path: "/home/dash/Documenti/nvim comandi"
include_patterns: ["*.md", "*.txt", "*.pdf", "*.docx"]
exclude_patterns: ["draft_*"] # Ignora bozze
index_config:
metric: "cosine"
precision: "float32" # <--- Risparmia 75% RAM!
m: 16 # <--- Migliore qualità grafo
ef_construction: 200 # <--- Costruzione più lenta ma precisa
text_language: "italian" # <--- Stemming italiano per la ricerca ibrida
embedder:
type: "ollama_api"
url: "http://localhost:11434/api/embeddings"
model: "nomic-embed-text-v2-moe:latest"
document_processor:
chunking_strategy: "markdown" # <--- Rispetta la struttura MD
chunk_size: 1000
metadata_template:
source_file: "{{file_path}}"
category: "study_notes"
graph_enabled: true
graph_entity_extraction: true
llm:
base_url: "http://localhost:11434/v1"
model: "gemma3:4b" # O llama3, mistral, gpt-4o-mini...
temperature: 0.0 # Meglio deterministico per estrarre dati
vision_llm:
base_url: "http://localhost:11434/v1"
# IMPORTANTE: Devi aver scaricato questo modello su Ollama!
# Consigliato: llama3.2-vision (piccolo e potente) o llava
model: "gemma3:4b"
temperature: 0.0
max_tokens: 300 # Limitiamo la descrizione per non intasare l'embedding