## Stack Context
Part 2 of a 3-PR stack moving memory deduplication into the SDKs. See `sdk-dedup/tools-ts` (parent) for the full context and the TypeScript implementation this mirrors.
## What?
Port the normalized, priority-ordered (`static > dynamic > search`) profile deduplication into the Python SDKs.
- Each request injects one **owned memory block that replaces** the prior block rather than accumulating.
- Dedup is **request-local** (no shared state), so it stays correct under concurrency.
Covers OpenAI, Agent Framework (middleware + context provider), Cartesia, and Pipecat.
## Why?
Keeps the Python SDKs at behavioral parity with the TypeScript SDK so all integrations deduplicate memory the same way.
## Testing
- OpenAI: 31 passed, 11 skipped (live)
- Agent Framework: 59 passed
- Cartesia: 8 passed
- Pipecat: 8 passed
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- CURSOR_SUMMARY -->
---
> [!NOTE]
> **Medium Risk**
> Changes memory formatting and system-prompt injection across multiple SDK integrations; incorrect dedup or replacement could alter LLM context, but there is no auth or data-store risk.
>
> **Overview**
> Ports **normalized cross-source memory deduplication** and **replace-not-append injection** into the Python OpenAI, Agent Framework, Cartesia, and Pipecat packages so they match the TypeScript SDK behavior.
>
> **Deduplication** uses request-local keys: strip optional `[YYYY-MM-DD]` prefixes, normalize whitespace, and compare with `casefold`, with priority **static → dynamic → search**. In **`query` mode**, profile static/dynamic are excluded from dedup input so facts that only appear in search (or overlap profile) are not dropped before formatting.
>
> **Injection** no longer appends memory text every turn. OpenAI and Agent Framework middleware **strip prior owned `<supermemory context="user-memories" readonly>` blocks** and **replace** them once per request while keeping the caller’s system instructions; extra system messages lose stale blocks only. New helpers (`strip`/`replace`/`wrap`) live in each package’s utils.
>
> Tests cover normalized fact variants, query-mode search retention, and stale block replacement.
>
> <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit 42f308b224. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup>
<!-- /CURSOR_SUMMARY -->
## Summary
This PR introduces comprehensive Supermemory integration for the Microsoft Agent Framework, providing three complementary approaches to add persistent memory capabilities to agents: middleware for automatic memory injection, context providers for session-based memory management, and tools for explicit memory operations.
## Key Changes
- **SupermemoryChatMiddleware**: Automatic memory injection middleware that fetches relevant memories from Supermemory before LLM calls and optionally saves conversations. Supports three modes:
- `"profile"`: Injects all static and dynamic profile memories
- `"query"`: Searches for memories relevant to the current user message
- `"full"`: Combines both profile and query modes
- **SupermemoryContextProvider**: Idiomatic context provider following the Agent Framework pattern (similar to built-in Mem0 integration). Integrates with the session pipeline via `before_run()` and `after_run()` hooks for automatic memory retrieval and storage.
- **SupermemoryTools**: FunctionTool-compatible tools that agents can use for explicit memory operations:
- `search_memories()`: Search for specific memories
- `add_memory()`: Add new memories
- `get_profile()`: Retrieve user profile
- **Utility Functions**: Helper functions for:
- Memory deduplication across static, dynamic, and search result sources
- Profile-to-markdown conversion for LLM consumption
- Message extraction and conversation formatting
- Logging with configurable verbosity
- **Exception Hierarchy**: Custom exceptions for better error handling:
- `SupermemoryConfigurationError`: Missing/invalid configuration
- `SupermemoryAPIError`: API request failures
- `SupermemoryNetworkError`: Network connectivity issues
- `SupermemoryMemoryOperationError`: Memory operation failures
- **Comprehensive Documentation**: README with quick start examples, configuration options, and API reference for all three integration approaches.
- **Test Suite**: Unit tests covering middleware, context provider, tools, and utility functions with proper mocking and error scenarios.
## Implementation Details
- Supports both async (aiohttp) and sync (requests) HTTP clients with automatic fallback
- Handles multiple message formats (dict, objects with attributes, content arrays)
- Configurable memory storage with optional conversation grouping via `conversation_id`
- Environment variable fallback for API key configuration (`SUPERMEMORY_API_KEY`)
- Background task management for non-blocking memory operations in middleware
- Proper async/sync compatibility for the Supermemory SDK
https://claude.ai/code/session_012idB5y6UGK3zmeFULgTc4z
