ReMe (Remember Me, Refine Me)

Warnings

• gotcha Before v0.3.1.8, attempting to use `chromadb`-backed memory without `chromadb` installed could lead to import errors. While the library now handles this gracefully, users leveraging ChromaDB features must explicitly install `chromadb` (`pip install chromadb`).
  Fix: Install `chromadb` explicitly if using ChromaDB features: `pip install reme-ai[chroma]` or `pip install chromadb`.
• gotcha ReMe relies on external LLM services (e.g., OpenAI, LiteLLM) for key functionalities like summarization and retrieval. Ensure required API keys (e.g., `OPENAI_API_KEY`) are set in environment variables or `*.env` files for `ReMe` to function correctly. Without these, LLM-dependent operations will fail.
  Fix: Set necessary API keys as environment variables (e.g., `export OPENAI_API_KEY='your_key'`) or ensure they are loaded via `python-dotenv`.
• breaking In v0.3.1.4, core text truncation utilities were refactored and replaced with a new marker system. Custom integrations or logic that directly interacted with or relied on previous internal truncation methods might require review and adaptation.
  Fix: Review any custom logic that interacts with ReMe's internal text truncation. Prefer using public APIs for truncation or context management.
• gotcha From v0.3.1.5 onwards, `summarize` and `retrieve` operations are designed to surface failures more explicitly instead of masking them. If your application previously relied on silent failures for these operations, it might now encounter exceptions for operations that fail.
  Fix: Implement proper error handling (try-except blocks) around `summarize_memories` and `retrieve` calls to gracefully manage potential failures.
• gotcha The changelog for v0.3.1.3 states `litellm` was moved to dev dependencies. However, `litellm` is still listed in the `requirements.txt` for v0.3.1.8, implying it remains a core dependency. Users should rely on `pip install reme-ai` to ensure all necessary dependencies are installed.
  Fix: Always install `reme-ai` via `pip install reme-ai` to ensure all required dependencies, including `litellm`, are correctly managed.

Install

• pip install reme-ai Install latest version

Imports

• ReMe

  from reme import ReMe

Quickstart

This quickstart demonstrates how to initialize ReMe, add experiences, and then retrieve and summarize them. Ensure you have an LLM API key (e.g., OPENAI_API_KEY) set as an environment variable, as summarization and retrieval functionalities depend on it. If the key is missing, LLM-dependent operations will fail.

import os
from reme import ReMe

# Ensure your OpenAI API key is set as an environment variable
# e.g., export OPENAI_API_KEY='sk-...'
openai_api_key = os.environ.get('OPENAI_API_KEY', '')
if not openai_api_key:
    print("WARNING: OPENAI_API_KEY is not set. ReMe operations requiring an LLM will fail.")

# Initialize ReMe with a unique agent ID. This creates a local memory store.
re_me = ReMe(agent_id="my_first_agent")

# Add experiences to the agent's memory
re_me.add_experience("I met a new colleague today named Sarah. She works on the frontend team.")
re_me.add_experience("Sarah is quite good at React and uses TypeScript.")
re_me.add_experience("She mentioned she loves to hike and explore national parks.")

# Retrieve relevant memories based on a query
query = "Tell me about Sarah."
retrieved_memories = re_me.retrieve(query)
print(f"\nRetrieved memories for '{query}':")
for mem in retrieved_memories:
    print(f"- {mem}")

# Summarize memories (requires LLM interaction)
try:
    summarized_memory = re_me.summarize_memories(
        query="What do I know about Sarah's work and hobbies?", 
        count=3
    )
    print(f"\nSummarized memory: {summarized_memory}")
except Exception as e:
    print(f"\nFailed to summarize memories (API key likely missing/invalid): {e}")

# To clear the agent's memory store (optional)
# re_me.clear_memory()