Skip to content

README.md

Latest commit

 

History

History
303 lines (204 loc) · 6.45 KB

README.md

File metadata and controls

303 lines (204 loc) · 6.45 KB

AI Integration Guide

CASA is designed to work with any AI assistant. This guide shows you how to integrate with popular tools.

Quick Comparison

Tool Setup Time Persistence Best For
Claude Code 2 min Automatic Developers, terminal users
Cursor 2 min Automatic Coding + learning
ChatGPT Custom GPT 10 min Permanent Non-technical users
Claude.ai Projects 5 min Per-project Research, long-form
Any AI (manual) 1 min Per-chat Quick use

Claude Code (Recommended for Devs)

Claude Code automatically reads CLAUDE.md in your project root.

Setup

  1. Clone CASA repo (or copy to your project)
  2. Fill out config.yaml with your profile
  3. That's it—Claude Code reads both files automatically

Usage

# In terminal with Claude Code
claude

# Then just ask
> explain kubernetes pods
> /capture kubernetes pods
> /card kubernetes networking

What Happens

Claude Code reads:

  • CLAUDE.md → Instructions on how to help you learn
  • config.yaml → Your personal profile for analogies

Every explanation automatically uses your interests and connects to your projects.

Cursor

Cursor reads .cursorrules in your project root.

Setup

  1. Copy .cursorrules to your project
  2. Fill out config.yaml
  3. Open project in Cursor

Usage

Just chat with Cursor's AI:

explain docker networking in CASA style

Or use commands:

/learn docker volumes
/capture what I just learned about kubernetes

ChatGPT Custom GPT

Create a permanent learning companion in ChatGPT.

Setup

  1. Go to ChatGPT → Explore GPTs → Create
  2. Name: "CASA Learning Companion"
  3. Description: "Personalized learning using the CASA system"
  4. Instructions: Copy from ai/SYSTEM_PROMPT.md (the full prompt section)
  5. Fill in YOUR profile in the instructions
  6. Save and publish (private)

Usage

Open your Custom GPT and chat normally. It remembers your profile across conversations.

Advanced: With Actions

Add ability to save captures:

  1. Create a simple webhook (n8n, Zapier, or custom)
  2. Add as an Action in GPT settings
  3. GPT can now save captures to your notes automatically

Claude.ai Projects

Claude Projects maintain context across conversations.

Setup

  1. Create a new Project in Claude.ai
  2. Upload CLAUDE.md and config.yaml to project files
  3. Add project instructions:
You are my CASA learning companion. Read the uploaded CLAUDE.md for instructions and config.yaml for my profile. Always personalize explanations using my interests and connect to my projects.

Usage

Every conversation in that project uses your learning profile.

Gemini

Google Gemini doesn't have persistent instructions, but you can use Gems.

Setup (Gemini Advanced with Gems)

  1. Create a new Gem
  2. Paste the system prompt from ai/SYSTEM_PROMPT.md
  3. Fill in your profile
  4. Save

Setup (Basic)

Paste the system prompt at the start of each conversation.

Local LLMs (Ollama, LM Studio)

For privacy-focused learning with local models.

Ollama

Create a Modelfile:

FROM llama3

SYSTEM """
You are my personalized learning companion using the CASA system.

MY PROFILE:
- Interests: [your interests]
- Projects: [your projects]
- Career goal: [your goal]

[rest of system prompt from SYSTEM_PROMPT.md]
"""

Then:

ollama create casa-learner -f Modelfile
ollama run casa-learner

LM Studio

  1. Go to Settings → System Prompt
  2. Paste the CASA system prompt
  3. Save as preset

API Integration

For building your own tools using AI APIs.

Python + Claude API

import anthropic
import yaml

# Load config
with open('config.yaml') as f:
    config = yaml.safe_load(f)

# Build system prompt from config
system_prompt = f"""
You are a CASA learning companion.

User's interests (for analogies): {config['profile']['interests']}
User's projects: {[p['name'] for p in config['profile']['projects']]}
User's career goal: {config['profile']['career']['target']}

Always use their interests for analogies, connect to their projects,
and frame relevance to their career. Follow CASA format for captures.
"""

client = anthropic.Anthropic()

def learn(topic):
    response = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=1024,
        system=system_prompt,
        messages=[
            {"role": "user", "content": f"Explain {topic} using my personal anchors"}
        ]
    )
    return response.content[0].text

# Usage
print(learn("TCP three-way handshake"))

Python + OpenAI API

from openai import OpenAI
import yaml

with open('config.yaml') as f:
    config = yaml.safe_load(f)

system_prompt = f"""
You are a CASA learning companion.
User's interests: {config['profile']['interests']}
User's projects: {[p['name'] for p in config['profile']['projects']]}
Career goal: {config['profile']['career']['target']}
Always personalize using their interests and projects.
"""

client = OpenAI()

def learn(topic):
    response = client.chat.completions.create(
        model="gpt-4",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": f"Explain {topic}"}
        ]
    )
    return response.choices[0].message.content

print(learn("Kubernetes pods"))

n8n AI Workflow

Automate learning with n8n + AI.

Capture Enhancement Flow

[Telegram Message]
    → [Parse Capture]
    → [Claude API: Enhance with config.yaml profile]
    → [Save to Notion/Obsidian]
    → [Reply with enhanced capture]

See workflows/n8n-automation/ for full workflow JSON.

Testing Your Setup

After configuring any AI, test with this prompt:

"Explain how DNS works"

Check for:

  • ✅ Uses analogy from YOUR interests (not generic)
  • ✅ Mentions YOUR specific project
  • ✅ Connects to YOUR career goal
  • ✅ Includes practical action you can try

If generic, remind it:

"Remember to use my specific interests for analogies and connect to my projects"

Files Reference

File Purpose Used By
config.yaml Your personal profile All tools
CLAUDE.md Claude Code instructions Claude Code
.cursorrules Cursor AI rules Cursor
ai/SYSTEM_PROMPT.md Universal prompt ChatGPT, Claude.ai, Gemini, etc.

The AI is a learning multiplier. The better your profile, the better your learning.