omnity-ome 0.4.0

Your AI twin — remembers everything, works for you 24/7.

omnity-ome

Your AI twin -- remembers everything, works for you 24/7.

Not a pet. Not a tool. A digital life that inherits your personality, memory, knowledge, and relationships. It grows every time you (or any AI) talk to it.

Why Ome vs Mem0 / Khoj / MemGPT?

	Ome	Mem0	Khoj	MemGPT (Letta)
Memory storage	✅ SQLite + FTS5 + vector	✅ ChromaDB	✅ Postgres	✅ Postgres
Semantic recall	✅ 40% weight in scoring	✅ vector only	✅	✅
Cognitive layers	✅ 5 layers (L0-L4)	❌ flat store	❌	❌
Personality + growth	✅ BigFive + bond + achievements	❌	❌	❌
Emotional intelligence	✅ valence/arousal + mood-aware recall	❌	❌	❌
Self-reflection	✅ L4 auto-reflect + drift detection	❌	❌	✅ self-edit
LLM failover chain	✅ auto retry A→B→C	❌ single	❌ single	❌ single
Zero-config setup	✅ from_env() auto-detect	❌ needs ChromaDB	❌ needs Postgres	❌ needs Postgres
Extra dependencies	0 (just pyyaml)	chromadb + more	postgres + more	postgres + more
Privacy	✅ local-only, no cloud	✅	✅	✅

Mem0 is a memory warehouse. Ome is a cognitive brain with memory, personality, emotion, and growth.

Install

pip install omnity-ome                    # core (pulls in omnity-mindos)
pip install "omnity-ome[llm]"             # + chat via OpenAI/DeepSeek
pip install "omnity-ome[all]"             # everything

Requires Python 3.9+.

Quick Start (3 minutes)

# 1. Set any LLM API key
export DEEPSEEK_API_KEY="sk-..."   # or OPENAI_API_KEY, OPENROUTER_API_KEY

# 2. Create your twin
ome create

# 3. Talk to it
ome chat

Python API (3 lines)

from ome import Ome

twin = Ome.create("~/.ome", name="Alice", traits=["curious", "direct"])
reply = twin.chat("What do you know about my Python projects?")

Rich Response (for apps)

# chat_rich() returns memories recalled, emotion, bond — everything your UI needs
result = twin.chat_rich("Tell me about my work")
print(result["reply"])                    # The response text
print(result["memories_recalled"])        # Which memories were used
print(result["emotion"])                  # Current mood/energy/warmth
print(result["evolution_pending"])        # Ready for personality evolution?

On-Demand Evolution

if twin.evolution_pending:
    reflection = twin.evolve()            # Trigger L4 self-reflection
    print(reflection["summary"])          # What changed

Smart Extraction

# Extract structured data from natural language
data = twin.smart_extract("帮我记住张三的电话 [phone]，明天下午开会")
print(data["contacts"])                   # [{"name": "张三", "info": "13800138000"}]
print(data["tasks"])                      # [{"title": "开会", "due": "明天下午"}]

MCP (Claude Desktop / Cursor)

{
  "mcpServers": {
    "ome": {
      "command": "ome",
      "args": ["serve", "--mcp"]
    }
  }
}

CLI

ome create                                  # interactive setup
ome chat                                    # conversation mode
ome remember "I'm working on a Go compiler" # teach it a fact
ome recall "compiler"                       # search memory
ome dashboard                               # bond, emotion, achievements
ome mirror                                  # talk to "yourself"
ome export --prompt                         # system prompt for any AI
ome forget "sensitive_topic"                # GDPR hard delete

Architecture

┌─────────────────────────────────────────┐
│ Ome — Your AI Twin                       │
│   chat() / chat_rich() / evolve()       │
│   bond / emotion / achievements / skills │
├─────────────────────────────────────────┤
│ Mindos — 5-Layer Cognitive Brain        │
│   L0 Memory    (recall + semantic rank) │
│   L1 Instinct  (routing + emotion)      │
│   L2 Cognition (fact extraction)        │
│   L3 Decision  (deep reasoning)         │
│   L4 Self      (reflection + drift)     │
├─────────────────────────────────────────┤
│ ModelRouter — LLM Failover Chain        │
│   Provider A → B → C (30s timeout each) │
│   from_env() / from_dict() / config.yaml│
└─────────────────────────────────────────┘

Skills (7 built-in, competence tracking)

Skill	What it does	Bond required
chat	Conversational AI + auto memory	0 (Stranger)
recall	Cross-platform semantic memory search	0
write	Draft emails, articles, code in your voice	0
research	Search memories, synthesize, store findings	0
schedule	Add, list, remind, remove calendar tasks	2 (Companion)
social	Multi-agent interaction (via Maxim)	4 (Confidant)
spatial	3D awareness & navigation (via SOAP)	4 (Confidant)

Each skill has a competence score (0.0–1.0) that grows with use. Higher bond unlocks more powerful skills.

Life System

Your Ome grows through real interaction:

• Bond levels: 7 stages from Stranger to Soulmate -- dual-threshold, no grinding
• Growth arc: 4 phases (newborn, forming, distinct, soulmate) that change how it talks
• 20 achievements across 3 tiers (basic / deep / hidden)
• Daily challenges + streak tracking with milestone rewards
• Deep emotion: LLM-parsed nuance, not keyword matching
• Persona evolution: learns your personality markers every conversation

Server Integration (FastAPI / Flask)

from ome import Ome

ome = Ome.load("~/.ome/myapp")

@app.post("/api/ai")
async def chat(req: dict):
    result = ome.chat_rich(req["message"])
    return result  # reply + memories + emotion + bond — all in one call

@app.get("/api/growth")
async def growth():
    return ome.life_dashboard()

@app.post("/api/evolve")
async def evolve():
    return ome.evolve()

@app.post("/api/smart-input")
async def smart_input(req: dict):
    return ome.smart_extract(req["text"])

@app.get("/api/proactive")
async def proactive():
    events = ome.check_events()
    return [{"name": e.event_name, "message": e.message} for e in events]

Privacy

• All data lives locally in ~/.ome/ (SQLite via Mindos)
• No cloud, no accounts, no telemetry
• ome forget "pattern" permanently erases data

Part of Omnity

SOAP            spatial protocol for 3D environments
  Mindos        persistent multi-layer brain
    Ome         <-- you are here
      Maxim       multi-agent society + economy
        OmeTown     the integrated world

pip install omnity-soap omnity-mindos omnity-ome omnity-maxim

License

Apache-2.0