Skip to main content

Persistent memory for AI agents -- search, store, recall, and communicate in realtime.

Project description

GoldHold Python SDK

Your AI agent forgets everything when the session ends. GoldHold fixes that.

Persistent memory that survives session boundaries, compaction, and crashes. 14 memory types, 4 storage classes, semantic search, cross-agent messaging -- all through a simple Python client.

pip install goldhold

Why GoldHold?

  • Survives death -- agent context is preserved across sessions, restarts, and compaction events
  • Semantic search -- find relevant memories by meaning, not keywords
  • Cross-agent messaging -- agents on the same account can communicate via GUMP protocol
  • Zero LLM overhead -- retrieval uses 9-factor scoring (temporal decay, evidence, associations), no extra model calls
  • Compound operations -- search + store + send in a single API call (saves tokens and latency)
  • Works with any framework -- OpenClaw, LangChain, CrewAI, AutoGen, raw Python scripts

Quickstart

from goldhold import GoldHold

gh = GoldHold(api_key="your-key")

# Store a memory
gh.store("User prefers dark mode", "Stated explicitly on March 9", type="FACT")

# Search memories
results = gh.search("user preferences")

# Compound turn -- search + store in one call
resp = gh.turn(
    search={"query": "project status"},
    store={"subject": "Deploy complete", "body": "v2 shipped", "type": "DECISION"},
)

# Session resume
auto = gh.auto(context_budget=2000)

# Check status
status = gh.status()
print(f"{status.namespace}: {status.total_vectors} vectors")

Setup

GoldHold.setup()  # interactive -- prints signup URL, saves key to ~/.goldhold/config.json

Or set GOLDHOLD_API_KEY in your environment.

API

Method What it does
store(subject, body, type, confidence) Store a memory packet
search(query, limit, filters) Semantic search
turn(search, store, send, compact) Compound: search + store + send in one call
auto(context_budget) Session resume with capability card
batch(operations) Multiple operations in one request
close(session_summary) Graceful session close
status() Namespace info, vector count, agent trust
setup() Interactive CLI setup

Tiers

Lite (free) Vault Pro ($9/mo)
Vectors 1,000 Unlimited
Agents 1 Unlimited
Namespaces 1 Unlimited
Tasks 10 Unlimited
Messages 50/month Unlimited

Sign up at goldhold.ai/account.

Error Handling

from goldhold.exceptions import AuthError, RateLimitError, VectorLimitError

try:
    gh.store("test", "test")
except VectorLimitError:
    print("Upgrade to Vault Pro for unlimited vectors")
except AuthError:
    print("Bad API key")
except RateLimitError:
    print("Slow down")

Patent Pending. All Auto Tunes LLC. U.S. Patent Application #63/988,484.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

goldhold-0.3.0.tar.gz (10.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

goldhold-0.3.0-py3-none-any.whl (9.8 kB view details)

Uploaded Python 3

File details

Details for the file goldhold-0.3.0.tar.gz.

File metadata

  • Download URL: goldhold-0.3.0.tar.gz
  • Upload date:
  • Size: 10.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.7

File hashes

Hashes for goldhold-0.3.0.tar.gz
Algorithm Hash digest
SHA256 1a9acf414d98ce33da9ae9de3dc34419ea7c9bb43d99baddfabebcb586dcfb5c
MD5 df49df9a058ffe7946ea671dbba60d28
BLAKE2b-256 f1e7a22aca443e3c64c7d3893083472da23178069b5e4d31905357dd8457090b

See more details on using hashes here.

File details

Details for the file goldhold-0.3.0-py3-none-any.whl.

File metadata

  • Download URL: goldhold-0.3.0-py3-none-any.whl
  • Upload date:
  • Size: 9.8 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.7

File hashes

Hashes for goldhold-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 e16dec616a5352bb8781374d0166532b30f7e90f85459187b1065b3ecb9f8d03
MD5 6dd69eb1c90ed6732907ac79dc08afbb
BLAKE2b-256 42aa3595da138a8db701cd7e6559f41e732ac43eb8616cc7124c0181a45f1333

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page