Cognitive Enhancement Engine (认知力增强引擎)#

Lightweight AI Agent cognitive engine with working memory, TF-IDF vector memory, planning, reasoning, reflection, and metacognitive monitoring. Zero external dependencies — pure Python standard library.

Quick Start#

bash

# Run built-in demo
python skills/cognitive-enhancement-engine/engine.py

# Or one-click setup
bash skills/cognitive-enhancement-engine/scripts/setup.sh     # Linux/macOS/WSL
skills\cognitive-enhancement-engine\scripts\setup.bat         # Windows

Core Usage#

python

from engine import CognitiveEnhancer

# Create engine
brain = CognitiveEnhancer(long_term_capacity=1000)

# Learn
brain.memorize("Paris is the capital of France.", importance=0.9)
brain.perceive("User asked about French capital")

# Retrieve
results = brain.recall("capital of France", top_k=3)

# Plan
plan = brain.plan("Build a web application")

# Reason
answer = brain.reason("What is the capital of France?")

# Reflect
suggestions = brain.reflect()

# Full task execution
result = brain.execute_task("Calculate 15% tip on $200 bill")
print(result)

# Status
status = brain.get_status()

API Overview#

MethodDescriptionperceive(observation)Store perception into working memoryrecall(query, top_k)Search long-term memorymemorize(content, importance)Store into long-term memoryplan(goal)Decompose goal into actionable stepsreason(problem)Memory-based reasoningreflect()Discover failure patterns, suggest improvementsexecute_task(goal, executor)End-to-end task executionget_status()Return engine runtime status

Configuration#

ParameterDefaultDescriptionlong_term_capacity1000Max long-term memoriesworking_memory_size10Working memory FIFO sizesimilarity_threshold0.15Recall similarity threshold

Features#

• TF-IDF Vector Memory — Inverted-index fast similarity search
• Working Memory — FIFO short-term context cache
• Planner — Goal decomposition + automatic task type detection (calculate/search/summarize/translate/write)
• Reasoner — Memory-retrieval based Q&A
• Reflector — Failure pattern tracking and root cause mining
• Metacognitive Monitor — Task duration & error rate tracking, dynamic adjustment

Installation#

MethodCommandOne-click (Linux/macOS)bash scripts/setup.shOne-click (Windows)scripts\setup.batCopy-onlyCopy engine.py to your projectClawHubclawhub install cognitive-enhancement-engine

File Structure#

text

cognitive-enhancement-engine/
├── SKILL.md
├── engine.py              # Core engine (~17KB)
├── index.js               # Node.js bridge
├── package.json
├── assets/
│   └── icon.svg
├── references/
│   ├── API_SPEC.md
│   └── USE_GUIDE.md
└── scripts/
    ├── setup.sh
    ├── setup.bat
    ├── test-basic.py
    └── test-client.js

License#

MIT