mem0-analytics 0.1.0

Real-time analytics and monitoring infrastructure for Mem0 — tracks all LLMs, vector stores, and embedders automatically.

🧠 Mem0 Analytics

Real-time analytics and monitoring infrastructure for the Mem0 ecosystem. Plug it in once — and it automatically tracks every memory, model, vector store, and embedder you use.

---

🧩 Overview

Mem0 Analytics is the official data, analytics, and monitoring layer for Mem0. It automatically traces every memory interaction, measures latency and efficiency across the entire stack, and presents insights through a rich in-terminal dashboard or PostHog cloud visualization.

No setup, no configuration — just:

pip install mem0 mem0-analytics

and Mem0 Analytics automatically activates. Every add, search, update, reset, and query operation is tracked — across all supported LLMs, embedders, and vector stores.

---

⚙️ What It Does

• 🧠 Autoinstruments Mem0 — wraps every memory call transparently
• ⚡ Tracks performance — latency, tail (P95), TTFR, and system load
• 💾 Monitors all layers — LLM, embedder, and vector database
• 🔁 Aggregates KPIs every 60 s locally (SQLite store)
• 📊 Visualizes metrics in a live, auto-updating terminal dashboard
• ☁️ Optionally syncs to PostHog for team dashboards

---

🖥 Dashboard

Real-time monitoring of:

• ⚡ Latency (avg & P95) by operation
• 🧩 Embedding & Vector performance
• 💾 Cache effectiveness
• 🧠 TTFR (Time-to-First-Response)
• 🧮 Success, error, and resource metrics
• ✅ System stability indicator

Runs completely local — powered by rich. No servers, no dependencies beyond SQLite.

---

☁️ Cloud Analytics (Optional)

For org-wide tracking, enable PostHog sync:

export POSTHOG_API_KEY=<your_key>
export POSTHOG_HOST=https://app.posthog.com

Analytics are automatically batched and sent every minute.

---

📊 Metrics Tracked

Category	Metrics	Description
Performance	avg_latency_ms, latency_p95, ttfr_ms	End-to-end and tail latency
Embedder / Vector	avg_embed_latency, avg_vector_latency	Stage-wise breakdown
Efficiency	cache_effectiveness, usage_count	Cache reuse and throughput
System Health	cpu_percent, mem_used_mb	Runtime system stats
Reliability	success_rate, error_rate	Stability and health signals

---

🧱 Architecture

Mem0 (any model, vector, embedder)
   ↓
mem0-analytics → captures metrics automatically
   ↓
SQLite (~/.mem0_metrics.db) → local store
   ↓
Live CLI Dashboard  ←  Aggregator updates every 60 s
   ↓
(Optional) PostHog sync for cloud dashboards

Local-first, privacy-safe, fully offline by default.

---

🚀 Quick Start

pip install mem0 mem0-analytics

That’s it — analytics auto-activates with Mem0.

View the live dashboard

python -m mem0_analytics.dashboard

Data is stored locally at:

~/.mem0_metrics.db

Updated automatically every minute.

---

🧠 Ecosystem Coverage

Mem0 Analytics supports all major backends out of the box:

Layer	Supported
LLMs	OpenAI (gpt-4o, gpt-5-nano), Ollama (smollm2, smollm2:135m), Claude, Gemini, Groq, Llama, DeepSeek, etc.
Vector Stores	Qdrant, ChromaDB, FAISS, Weaviate, Pinecone, Milvus, Redis, LanceDB
Embedders	OpenAI, Ollama, Hugging Face, Sentence-Transformers, InstructorXL, BGE, etc.

If it works with Mem0 — it’s already tracked by Mem0 Analytics.

---

🔬 Engineering Highlights

• 🪶 Lightweight (no external DB required)
• 🧱 Built on SQLite + rich for local telemetry
• 🔁 Background aggregator with rolling KPIs
• ☁️ Optional PostHog sync for teams
• 🧩 Pluggable architecture (add any provider)
• 💡 Minimal overhead — <1 ms per operation

---

🧭 Roadmap

• Local SQLite metrics layer
• Terminal dashboard
• PostHog publishing
• Cost & token usage metrics
• Prometheus exporter
• Alerting / anomaly detection
• Multi-agent comparison mode

---

🤝 Contributing

Contributions are open — help extend analytics across new backends, metrics, or visualizations.

📜 License

Released under the MIT License. See LICENSE for details.