gistfs 0.3.0

Use GitHub Gists as a persistent key-value filesystem — ideal for AI agent memory

gistfs

Use GitHub Gists as a persistent key-value filesystem — ideal for AI agent memory.

Install

pip install gistfs

With optional integrations:

pip install gistfs[llamaindex]   # LlamaIndex KVStore
pip install gistfs[langgraph]    # LangGraph BaseStore
pip install gistfs[all]          # everything

Quick start

Creating a new gist

No need to create a gist manually — bootstrap one from code:

from gistfs import GistFS

gfs = GistFS.create(description="my agent memory")
print(gfs.gist_id)  # save this for later

# Or with GistMemory:
from gistfs import GistMemory
mem = GistMemory.create(description="my agent memory")

As a filesystem (context manager)

from gistfs import GistFS

with GistFS(gist_id="your_gist_id") as gfs:
    gfs.write("config.json", {"model": "gpt-4", "temperature": 0.7})
    config = gfs.read("config.json")
    print(gfs.list_files())
    gfs.delete("config.json")

File-like interface

with GistFS(gist_id="your_gist_id") as gfs:
    with gfs.open("notes.txt", "w") as f:
        f.write("hello world")

    with gfs.open("notes.txt", "r") as f:
        content = f.read()

    with gfs.open("notes.txt", "a") as f:
        f.write("\nappended line")

As AI agent memory

from gistfs import GistMemory

with GistMemory(gist_id="your_gist_id") as mem:
    mem.put("conversation_1", {"messages": [{"role": "user", "content": "hi"}]})
    history = mem.get("conversation_1")
    all_data = mem.get_all()
    mem.delete("conversation_1")

LlamaIndex integration

from gistfs.integrations.llamaindex import GistKVStore

store = GistKVStore(gist_id="your_gist_id")
store.put("doc1", {"text": "hello world"}, collection="docstore")
doc = store.get("doc1", collection="docstore")

LangGraph integration

from gistfs.integrations.langgraph import GistStore

store = GistStore(gist_id="your_gist_id")
store.put(("user", "prefs"), "theme", {"value": "dark"})
item = store.get(("user", "prefs"), "theme")

Encryption

Install with encryption support:

pip install gistfs[encryption]

File content is encrypted (Fernet) and base64-encoded before being sent to GitHub, and decrypted transparently on read. Pass an encryption_key to any class:

from gistfs import GistFS, derive_key
from gistfs.integrations.langgraph import GistStore

# Derive a key from a password (store the salt for later reuse)
key, salt = derive_key("mypassword", salt=b"gistfs-demo-salt")

# GistFS — files are encrypted at rest in the gist
with GistFS(gist_id="your_gist_id", encryption_key=key) as gfs:
    gfs.write("secrets.json", {"api_key": "sk-123"})
    assert gfs.read("secrets.json") == {"api_key": "sk-123"}

# LangGraph store — same key, same transparent encryption
store = GistStore(gist_id="your_gist_id", encryption_key=key)
store.put(("user", "prefs"), "theme", {"value": "dark"})

Authentication

Set the GITHUB_TOKEN environment variable with a GitHub personal access token that has gist scope. Read-only operations on public gists work without a token.

export GITHUB_TOKEN=ghp_your_token_here

Or pass it directly:

gfs = GistFS(gist_id="abc123", token="ghp_...")