Memory

Mudabbir uses a two-tier memory system: file-based sessions for conversation history and optional Mem0 for semantic long-term memory.

File Store

JSON-based session storage with indexing and search.

Mem0 Integration

Semantic memory with auto-learning and vector search.

Sessions

Session management, grouping, and search.

Context Building

How memory is assembled into the agent’s context window.

Memory Isolation

Per-user memory scoping for multi-channel deployments.

Memory Architecture

┌─────────────────────────────────────┐
│         Context Builder             │
│  (Assembles context for the agent)  │
├──────────┬──────────────────────────┤
│          │                          │
│  File    │    Mem0 Semantic         │
│  Store   │    Memory (optional)     │
│          │                          │
│ Sessions │    Auto-learn            │
│ Facts    │    Semantic search       │
│ Index    │    Vector store          │
└──────────┴──────────────────────────┘

Quick Setup

File Store (Default)

No configuration needed. Sessions are automatically saved to ~/.mudabbir/memory/.

Mem0 (Optional)

For semantic memory with auto-learning:

Terminal window
curl -fsSL https://mudabbir.fly.dev/install.sh | sh


# Or add the memory extra manually
pip install mudabbir[memory]


export MUDABBIR_MEM0_AUTO_LEARN=true
export MUDABBIR_MEM0_LLM_PROVIDER="ollama"
export MUDABBIR_MEM0_LLM_MODEL="llama3.2"
export MUDABBIR_MEM0_EMBEDDER_PROVIDER="ollama"
export MUDABBIR_MEM0_EMBEDDER_MODEL="nomic-embed-text"
Last updated: February 21, 2026
1 min read
Edit this page