contextmd 0.1.2

A provider-agnostic middleware that gives LLM API calls persistent, human-readable memory using local Markdown files

ContextMD

A provider-agnostic middleware that gives OpenAI, Anthropic, and LiteLLM API calls persistent, human-readable memory using local Markdown files.

Installation

pip install contextmd

# With provider support
pip install contextmd[openai]      # OpenAI only
pip install contextmd[anthropic]   # Anthropic only
pip install contextmd[litellm]     # LiteLLM (100+ providers)
pip install contextmd[all]         # All providers

Quick Start

OpenAI

from openai import OpenAI
from contextmd import ContextMD

# Wrap your existing client
client = ContextMD(OpenAI(), memory_dir=".contextmd/")

# Use exactly like normal - memory is automatic
response = client.chat.completions.create(
    model="gpt-5.2",
    messages=[{"role": "user", "content": "Hello!"}]
)

Anthropic

from anthropic import Anthropic
from contextmd import ContextMD

client = ContextMD(Anthropic(), memory_dir=".contextmd/")

response = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}]
)

LiteLLM (100+ providers)

import litellm
from contextmd import ContextMD

client = ContextMD(litellm, memory_dir=".contextmd/")

# Works with any LiteLLM-supported model
response = client.completion(
    model="gpt-5.2",
    messages=[{"role": "user", "content": "Hello!"}]
)

# Or use Claude, Gemini, etc.
response = client.completion(
    model="claude-opus-4-6",
    messages=[{"role": "user", "content": "Hello!"}]
)

How It Works

ContextMD intercepts your API calls and:

1. Bootstrap Loading: Injects stored memory into every request
2. Response Processing: Tracks token usage and extracts memorable facts
3. Memory Storage: Saves facts to human-readable Markdown files

Memory Types

• Semantic: Permanent facts (preferences, tech stack, project context)
• Episodic: Time-stamped events (decisions, tasks completed)
• Procedural: Learned workflows ("Always use pnpm")

File Structure

.contextmd/
├── MEMORY.md              # Semantic facts (200 line cap)
├── config.md              # Configuration
├── memory/
│   ├── 2025-03-01.md      # Daily episodic logs
│   └── 2025-03-02.md
└── sessions/
    └── 2025-03-01-auth.md # Session snapshots

API Reference

Manual Memory

# Remember something explicitly
client.remember("User prefers dark mode", memory_type="semantic")
client.remember("Completed auth feature", memory_type="episodic")
client.remember("Always run tests before commit", memory_type="procedural")

Session Management

# Create a named session
with client.new_session("auth-implementation") as session:
    response = client.chat.completions.create(...)
    # Session snapshot saved automatically on exit

# Or manually
session = client.new_session("feature-work")
# ... do work ...
session.end()  # Saves snapshot

Configuration

from contextmd import ContextMD, ContextMDConfig

config = ContextMDConfig(
    memory_line_cap=200,           # Max lines in MEMORY.md
    bootstrap_window_hours=48,     # Hours of episodic memory to load
    compaction_threshold=0.8,      # Token threshold for extraction
    snapshot_message_count=15,     # Messages in session snapshots
    extraction_frequency="session_end",  # When to extract
)

client = ContextMD(openai_client, config=config)

CLI

# Initialize in current directory
contextmd init

# View memory
contextmd show

# View recent activity
contextmd history --hours 24

# List sessions
contextmd sessions

# Add memory manually
contextmd add "User prefers TypeScript" --type semantic

# View statistics
contextmd stats

# Reset all memory
contextmd reset

Architecture

┌─────────────────────────────────────────────────────────────┐
│                     Your Application                         │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                    ContextMD Wrapper                         │
│  ┌─────────────┐  ┌──────────────┐  ┌───────────────────┐  │
│  │   Client    │  │   Memory     │  │    Extraction     │  │
│  │   Wrapper   │──│   Router     │──│    Engine         │  │
│  └─────────────┘  └──────────────┘  └───────────────────┘  │
│         │                │                    │             │
│         │                ▼                    │             │
│         │        ┌──────────────┐             │             │
│         │        │   Storage    │◄────────────┘             │
│         │        │   Layer      │                           │
│         │        └──────────────┘                           │
└─────────│───────────────────────────────────────────────────┘
          │
          ▼
┌─────────────────────────────────────────────────────────────┐
│              Provider Adapters                               │
│  ┌─────────┐    ┌───────────┐    ┌──────────┐              │
│  │ OpenAI  │    │ Anthropic │    │ LiteLLM  │              │
│  └─────────┘    └───────────┘    └──────────┘              │
└─────────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────────┐
│                      LLM Provider                            │
└─────────────────────────────────────────────────────────────┘

Development

# Install with dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Type checking
mypy src/contextmd

# Linting
ruff check src/contextmd

License

MIT