Configuration Reference

Content verified · 2026-04-07

All settings live in src/ormah/config.py and use the ORMAH_ prefix.

Settings are loaded from:

1. ~/.config/ormah/.env
2. local .env

Later files override earlier ones.

Server

Setting	Default
host	127.0.0.1
port	8787
log_format	text
log_level	INFO

Paths

Setting	Default
memory_dir	~/.local/share/ormah/memory

Embeddings

Setting	Default
embedding_provider	local
embedding_model	BAAI/bge-base-en-v1.5
embedding_dim	768
embedding_max_content_chars	512

LLM

Setting	Default
llm_provider	litellm
llm_model	claude-haiku-4-5-20251001
llm_base_url	http://localhost:11434
llm_timeout_seconds	60

Note: setup may persist different values in .env, including none.

Background Intervals

Setting	Default
auto_link_interval_minutes	1440
decay_interval_hours	24
conflict_check_interval_minutes	1440
conflict_check_all_spaces	false
duplicate_check_interval_minutes	1440
auto_cluster_interval_minutes	60
consolidation_interval_minutes	1440
importance_recompute_interval_minutes	120

Hippocampus

Setting	Default
hippocampus_watch_dirs	[]
hippocampus_debounce_seconds	2.0
hippocampus_enabled	true
hippocampus_ignore_patterns	[]

Operational note: default-enabled does not mean active without configured watch dirs.

Session Watcher

Setting	Default
session_watcher_enabled	false
session_watcher_dir	~/.claude/projects
session_watcher_debounce_seconds	60.0
session_watcher_min_turns	5
session_watcher_lookback_hours	72

Search

Setting	Default
fts_weight	0.4
vector_weight	0.6
similarity_threshold	0.4
rrf_k	60
similarity_blend_weight	0.5
fts_only_dampening	0.5
min_result_score	0.1
rrf_min_spread_ratio	0.05

Question-query tuning

Setting	Default
question_fts_weight_scale	0.3
question_vector_weight_scale	1.5
question_similarity_blend_weight	0.85

Post-retrieval tuning

Setting	Default
title_match_boost	2.0
length_penalty_threshold	300
recency_boost	0.05
recency_half_life_days	7.0
access_boost	0.05
tier_boost_core	0.1
tier_boost_working	0.0
tier_boost_archival	-0.1

Important note: current search applies tier as a multiplicative factor after confidence adjustment, while recency and access are additive proportional bonuses.

Spreading Activation

Setting	Default
activation_decay	0.5
activation_seed_count	5
activation_max_per_seed	3

Auto-Link / Merge / Importance

Setting	Default
auto_link_similarity_threshold	0.65
auto_link_cross_space_penalty	0.1
auto_link_max_edges_per_run	500
auto_merge_threshold	0.85
importance_access_weight	0.34
importance_edge_weight	0.33
importance_recency_weight	0.33
importance_access_reference	50
importance_edge_reference	20
importance_recency_half_life_days	14.0
decay_importance_threshold	0.5

Whisper-Out and Nudge

Setting	Default
whisper_out_enabled	true
whisper_out_min_turns	3
whisper_out_interval	10
whisper_nudge_interval	10

Whisper Recall

Setting	Default
whisper_max_nodes	6
whisper_min_relevance_score	0.45
whisper_reranker_enabled	true
whisper_reranker_model	Xenova/ms-marco-MiniLM-L-6-v2
whisper_reranker_min_score	0.40
whisper_reranker_blend_alpha	0.6
whisper_reranker_max_doc_chars	512
whisper_context_buffer_size	5
whisper_session_gap_minutes	10
whisper_intent_threshold	0.65
whisper_topic_shift_enabled	true
whisper_topic_shift_threshold	0.75
whisper_injection_gate	0.50
whisper_exploration_enabled	true

Affinity

Setting	Default
affinity_similarity_threshold	0.70
affinity_half_life_days	30.0
affinity_max_boost	0.15
affinity_implicit_weight	0.8

Space Prioritization

Setting	Default
space_boost_global	1.0
space_boost_other	0.6

FSRS / Tier Limits / Ingestion

Setting	Default
core_memory_cap	50
working_decay_days	14
fsrs_initial_stability	1.0
fsrs_decay_threshold	0.3
fsrs_stability_growth	1.5
fsrs_max_stability	365.0
ingest_max_content_chars	100000

Agent-Backed Maintenance

Setting	Default
claude_maintenance_enabled	false
claude_maintenance_interval_hours	24
claude_maintenance_batch_size	25

Code Anchor

• src/ormah/config.py