Add Engram-AI — experience-driven memory layer for AI agents

[Okabe-Shota]

Add Engram-AI to LLM Apps with Memory Tutorials

What it does:

Engram-AI (engram-forge on PyPI) is an experience-driven memory layer for AI agents. Most memory systems store facts; Engram-AI stores causality — what the agent did, what happened, and whether it worked:

from engram_ai import Forge

forge = Forge()
forge.record(
    action="Used Optional[str] in API response model",
    context="REST endpoint design",
    outcome="User rejected — 'no null values please'",
    valence=-0.8,  # learning signal: -1.0 to +1.0
)

# Query past experiences, partitioned by outcome
result = forge.query("REST API design")
# result["best"] — what worked
# result["avoid"] — what to avoid

# Crystallize repeated patterns into reusable skills
skills = forge.crystallize(min_experiences=3)

# Write learned skills back to agent config
forge.evolve("./CLAUDE.md")

Why it's relevant to this list:

• Directly addresses the problem of AI agents repeating the same mistakes across sessions
• Works with Claude Code (automatic recording via hooks), plus standalone Python API
• No API key required for core features

Links:

• GitHub: https://github.com/kajaha06251020/Engram-AI
• PyPI: https://pypi.org/project/engram-forge/

License: Apache 2.0 | Python 3.10+

[Copilot]

Pull request overview

Adds Engram-AI (engram-forge) as an additional entry under the “LLM Apps with Memory Tutorials” section of the awesome list, introducing an experience-driven memory layer concept for AI agents.

Changes:

• Add a new README list item linking to the Engram-AI GitHub repo.
• Describe Engram-AI’s experience schema (action/context/outcome/valence) and related features in the list entry.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

[awesomekoder]

Thanks for the interest! This PR adds README links but doesn't include runnable code in its own folder. For this repo, we need self-contained tutorials with the actual implementation included.