@mrtrinhvn/ag-kit 1.0.10 → 1.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/LICENSE +21 -0
- package/README.md +1 -0
- package/bin/cli.js +62 -0
- package/package.json +7 -1
- package/template/.agent/rules/GEMINI.md +1 -1
- package/template/.agent/skills/regent-orchestrator/SKILL.md +31 -0
- package/template/.agent/skills/telegram-bridge/SKILL.md +30 -0
- package/.agent/.shared/ui-ux-pro-max/data/charts.csv +0 -26
- package/.agent/.shared/ui-ux-pro-max/data/colors.csv +0 -97
- package/.agent/.shared/ui-ux-pro-max/data/icons.csv +0 -101
- package/.agent/.shared/ui-ux-pro-max/data/landing.csv +0 -31
- package/.agent/.shared/ui-ux-pro-max/data/products.csv +0 -97
- package/.agent/.shared/ui-ux-pro-max/data/prompts.csv +0 -24
- package/.agent/.shared/ui-ux-pro-max/data/react-performance.csv +0 -45
- package/.agent/.shared/ui-ux-pro-max/data/stacks/flutter.csv +0 -53
- package/.agent/.shared/ui-ux-pro-max/data/stacks/html-tailwind.csv +0 -56
- package/.agent/.shared/ui-ux-pro-max/data/stacks/jetpack-compose.csv +0 -53
- package/.agent/.shared/ui-ux-pro-max/data/stacks/nextjs.csv +0 -53
- package/.agent/.shared/ui-ux-pro-max/data/stacks/nuxt-ui.csv +0 -51
- package/.agent/.shared/ui-ux-pro-max/data/stacks/nuxtjs.csv +0 -59
- package/.agent/.shared/ui-ux-pro-max/data/stacks/react-native.csv +0 -52
- package/.agent/.shared/ui-ux-pro-max/data/stacks/react.csv +0 -54
- package/.agent/.shared/ui-ux-pro-max/data/stacks/shadcn.csv +0 -61
- package/.agent/.shared/ui-ux-pro-max/data/stacks/svelte.csv +0 -54
- package/.agent/.shared/ui-ux-pro-max/data/stacks/swiftui.csv +0 -51
- package/.agent/.shared/ui-ux-pro-max/data/stacks/vue.csv +0 -50
- package/.agent/.shared/ui-ux-pro-max/data/styles.csv +0 -59
- package/.agent/.shared/ui-ux-pro-max/data/typography.csv +0 -58
- package/.agent/.shared/ui-ux-pro-max/data/ui-reasoning.csv +0 -101
- package/.agent/.shared/ui-ux-pro-max/data/ux-guidelines.csv +0 -100
- package/.agent/.shared/ui-ux-pro-max/data/web-interface.csv +0 -31
- package/.agent/.shared/ui-ux-pro-max/scripts/__pycache__/core.cpython-313.pyc +0 -0
- package/.agent/.shared/ui-ux-pro-max/scripts/__pycache__/design_system.cpython-313.pyc +0 -0
- package/.agent/.shared/ui-ux-pro-max/scripts/core.py +0 -258
- package/.agent/.shared/ui-ux-pro-max/scripts/design_system.py +0 -1067
- package/.agent/.shared/ui-ux-pro-max/scripts/search.py +0 -106
- package/.agent/ARCHITECTURE.md +0 -288
- package/.agent/agents/backend-specialist.md +0 -263
- package/.agent/agents/code-archaeologist.md +0 -106
- package/.agent/agents/database-architect.md +0 -226
- package/.agent/agents/debugger.md +0 -225
- package/.agent/agents/devops-engineer.md +0 -242
- package/.agent/agents/documentation-writer.md +0 -104
- package/.agent/agents/explorer-agent.md +0 -73
- package/.agent/agents/frontend-specialist.md +0 -556
- package/.agent/agents/game-developer.md +0 -162
- package/.agent/agents/mobile-developer.md +0 -377
- package/.agent/agents/orchestrator.md +0 -416
- package/.agent/agents/penetration-tester.md +0 -188
- package/.agent/agents/performance-optimizer.md +0 -187
- package/.agent/agents/product-manager.md +0 -112
- package/.agent/agents/product-owner.md +0 -95
- package/.agent/agents/project-planner.md +0 -406
- package/.agent/agents/qa-automation-engineer.md +0 -103
- package/.agent/agents/quant-architect.md +0 -31
- package/.agent/agents/security-auditor.md +0 -170
- package/.agent/agents/seo-specialist.md +0 -111
- package/.agent/agents/test-engineer.md +0 -158
- package/.agent/mcp_config.json +0 -24
- package/.agent/rules/GEMINI.md +0 -280
- package/.agent/scripts/auto_preview.py +0 -148
- package/.agent/scripts/checklist.py +0 -217
- package/.agent/scripts/session_manager.py +0 -120
- package/.agent/scripts/verify_all.py +0 -327
- package/.agent/skills/api-patterns/SKILL.md +0 -81
- package/.agent/skills/api-patterns/api-style.md +0 -42
- package/.agent/skills/api-patterns/auth.md +0 -24
- package/.agent/skills/api-patterns/documentation.md +0 -26
- package/.agent/skills/api-patterns/graphql.md +0 -41
- package/.agent/skills/api-patterns/rate-limiting.md +0 -31
- package/.agent/skills/api-patterns/response.md +0 -37
- package/.agent/skills/api-patterns/rest.md +0 -40
- package/.agent/skills/api-patterns/scripts/api_validator.py +0 -211
- package/.agent/skills/api-patterns/security-testing.md +0 -122
- package/.agent/skills/api-patterns/trpc.md +0 -41
- package/.agent/skills/api-patterns/versioning.md +0 -22
- package/.agent/skills/app-builder/SKILL.md +0 -75
- package/.agent/skills/app-builder/agent-coordination.md +0 -71
- package/.agent/skills/app-builder/feature-building.md +0 -53
- package/.agent/skills/app-builder/project-detection.md +0 -34
- package/.agent/skills/app-builder/scaffolding.md +0 -118
- package/.agent/skills/app-builder/tech-stack.md +0 -40
- package/.agent/skills/app-builder/templates/SKILL.md +0 -39
- package/.agent/skills/app-builder/templates/astro-static/TEMPLATE.md +0 -76
- package/.agent/skills/app-builder/templates/chrome-extension/TEMPLATE.md +0 -92
- package/.agent/skills/app-builder/templates/cli-tool/TEMPLATE.md +0 -88
- package/.agent/skills/app-builder/templates/electron-desktop/TEMPLATE.md +0 -88
- package/.agent/skills/app-builder/templates/express-api/TEMPLATE.md +0 -83
- package/.agent/skills/app-builder/templates/flutter-app/TEMPLATE.md +0 -90
- package/.agent/skills/app-builder/templates/monorepo-turborepo/TEMPLATE.md +0 -90
- package/.agent/skills/app-builder/templates/nextjs-fullstack/TEMPLATE.md +0 -82
- package/.agent/skills/app-builder/templates/nextjs-saas/TEMPLATE.md +0 -100
- package/.agent/skills/app-builder/templates/nextjs-static/TEMPLATE.md +0 -106
- package/.agent/skills/app-builder/templates/nuxt-app/TEMPLATE.md +0 -101
- package/.agent/skills/app-builder/templates/python-fastapi/TEMPLATE.md +0 -83
- package/.agent/skills/app-builder/templates/react-native-app/TEMPLATE.md +0 -93
- package/.agent/skills/architecture/SKILL.md +0 -55
- package/.agent/skills/architecture/context-discovery.md +0 -43
- package/.agent/skills/architecture/examples.md +0 -94
- package/.agent/skills/architecture/pattern-selection.md +0 -68
- package/.agent/skills/architecture/patterns-reference.md +0 -50
- package/.agent/skills/architecture/trade-off-analysis.md +0 -77
- package/.agent/skills/bash-linux/SKILL.md +0 -199
- package/.agent/skills/behavioral-modes/SKILL.md +0 -242
- package/.agent/skills/brainstorming/SKILL.md +0 -168
- package/.agent/skills/brainstorming/dynamic-questioning.md +0 -350
- package/.agent/skills/business-ops/SKILL.md +0 -26
- package/.agent/skills/clean-code/SKILL.md +0 -202
- package/.agent/skills/cli-generator/SKILL.md +0 -48
- package/.agent/skills/code-review-checklist/SKILL.md +0 -109
- package/.agent/skills/cognitive-session/SKILL.md +0 -28
- package/.agent/skills/data-science/SKILL.md +0 -28
- package/.agent/skills/database-design/SKILL.md +0 -52
- package/.agent/skills/database-design/database-selection.md +0 -43
- package/.agent/skills/database-design/indexing.md +0 -39
- package/.agent/skills/database-design/migrations.md +0 -48
- package/.agent/skills/database-design/optimization.md +0 -36
- package/.agent/skills/database-design/orm-selection.md +0 -30
- package/.agent/skills/database-design/schema-design.md +0 -56
- package/.agent/skills/database-design/scripts/schema_validator.py +0 -172
- package/.agent/skills/deployment-procedures/SKILL.md +0 -241
- package/.agent/skills/doc.md +0 -177
- package/.agent/skills/documentation-templates/SKILL.md +0 -194
- package/.agent/skills/frontend-design/SKILL.md +0 -418
- package/.agent/skills/frontend-design/animation-guide.md +0 -331
- package/.agent/skills/frontend-design/color-system.md +0 -311
- package/.agent/skills/frontend-design/decision-trees.md +0 -418
- package/.agent/skills/frontend-design/motion-graphics.md +0 -306
- package/.agent/skills/frontend-design/scripts/accessibility_checker.py +0 -183
- package/.agent/skills/frontend-design/scripts/ux_audit.py +0 -722
- package/.agent/skills/frontend-design/typography-system.md +0 -345
- package/.agent/skills/frontend-design/ux-psychology.md +0 -541
- package/.agent/skills/frontend-design/visual-effects.md +0 -383
- package/.agent/skills/game-development/2d-games/SKILL.md +0 -119
- package/.agent/skills/game-development/3d-games/SKILL.md +0 -135
- package/.agent/skills/game-development/SKILL.md +0 -167
- package/.agent/skills/game-development/game-art/SKILL.md +0 -185
- package/.agent/skills/game-development/game-audio/SKILL.md +0 -190
- package/.agent/skills/game-development/game-design/SKILL.md +0 -129
- package/.agent/skills/game-development/mobile-games/SKILL.md +0 -108
- package/.agent/skills/game-development/multiplayer/SKILL.md +0 -132
- package/.agent/skills/game-development/pc-games/SKILL.md +0 -144
- package/.agent/skills/game-development/vr-ar/SKILL.md +0 -123
- package/.agent/skills/game-development/web-games/SKILL.md +0 -150
- package/.agent/skills/geo-fundamentals/SKILL.md +0 -156
- package/.agent/skills/geo-fundamentals/scripts/geo_checker.py +0 -289
- package/.agent/skills/i18n-localization/SKILL.md +0 -154
- package/.agent/skills/i18n-localization/scripts/i18n_checker.py +0 -241
- package/.agent/skills/intelligent-routing/SKILL.md +0 -335
- package/.agent/skills/knowledge-management/SKILL.md +0 -66
- package/.agent/skills/lint-and-validate/SKILL.md +0 -45
- package/.agent/skills/lint-and-validate/scripts/lint_runner.py +0 -172
- package/.agent/skills/lint-and-validate/scripts/type_coverage.py +0 -173
- package/.agent/skills/llm-routing-quirks/SKILL.md +0 -41
- package/.agent/skills/mcp-builder/SKILL.md +0 -176
- package/.agent/skills/memory-architecture/SKILL.md +0 -107
- package/.agent/skills/mini-antigravity-injection/SKILL.md +0 -61
- package/.agent/skills/mobile-design/SKILL.md +0 -394
- package/.agent/skills/mobile-design/decision-trees.md +0 -516
- package/.agent/skills/mobile-design/mobile-backend.md +0 -491
- package/.agent/skills/mobile-design/mobile-color-system.md +0 -420
- package/.agent/skills/mobile-design/mobile-debugging.md +0 -122
- package/.agent/skills/mobile-design/mobile-design-thinking.md +0 -357
- package/.agent/skills/mobile-design/mobile-navigation.md +0 -458
- package/.agent/skills/mobile-design/mobile-performance.md +0 -767
- package/.agent/skills/mobile-design/mobile-testing.md +0 -356
- package/.agent/skills/mobile-design/mobile-typography.md +0 -433
- package/.agent/skills/mobile-design/platform-android.md +0 -666
- package/.agent/skills/mobile-design/platform-ios.md +0 -561
- package/.agent/skills/mobile-design/scripts/mobile_audit.py +0 -670
- package/.agent/skills/mobile-design/touch-psychology.md +0 -537
- package/.agent/skills/nextjs-react-expert/1-async-eliminating-waterfalls.md +0 -312
- package/.agent/skills/nextjs-react-expert/2-bundle-bundle-size-optimization.md +0 -240
- package/.agent/skills/nextjs-react-expert/3-server-server-side-performance.md +0 -490
- package/.agent/skills/nextjs-react-expert/4-client-client-side-data-fetching.md +0 -264
- package/.agent/skills/nextjs-react-expert/5-rerender-re-render-optimization.md +0 -581
- package/.agent/skills/nextjs-react-expert/6-rendering-rendering-performance.md +0 -432
- package/.agent/skills/nextjs-react-expert/7-js-javascript-performance.md +0 -684
- package/.agent/skills/nextjs-react-expert/8-advanced-advanced-patterns.md +0 -150
- package/.agent/skills/nextjs-react-expert/9-cache-components.md +0 -103
- package/.agent/skills/nextjs-react-expert/SKILL.md +0 -267
- package/.agent/skills/nextjs-react-expert/scripts/convert_rules.py +0 -222
- package/.agent/skills/nextjs-react-expert/scripts/react_performance_checker.py +0 -252
- package/.agent/skills/nodejs-best-practices/SKILL.md +0 -333
- package/.agent/skills/parallel-agents/SKILL.md +0 -175
- package/.agent/skills/performance-profiling/SKILL.md +0 -143
- package/.agent/skills/performance-profiling/scripts/lighthouse_audit.py +0 -76
- package/.agent/skills/plan-writing/SKILL.md +0 -153
- package/.agent/skills/powershell-windows/SKILL.md +0 -167
- package/.agent/skills/product-management/SKILL.md +0 -30
- package/.agent/skills/python-patterns/SKILL.md +0 -441
- package/.agent/skills/red-team-tactics/SKILL.md +0 -199
- package/.agent/skills/rust-pro/SKILL.md +0 -176
- package/.agent/skills/seo-fundamentals/SKILL.md +0 -129
- package/.agent/skills/seo-fundamentals/scripts/seo_checker.py +0 -219
- package/.agent/skills/server-management/SKILL.md +0 -161
- package/.agent/skills/systematic-debugging/SKILL.md +0 -120
- package/.agent/skills/tailwind-patterns/SKILL.md +0 -269
- package/.agent/skills/tdd-workflow/SKILL.md +0 -148
- package/.agent/skills/testing-patterns/SKILL.md +0 -178
- package/.agent/skills/testing-patterns/scripts/test_runner.py +0 -219
- package/.agent/skills/vulnerability-scanner/SKILL.md +0 -276
- package/.agent/skills/vulnerability-scanner/checklists.md +0 -121
- package/.agent/skills/vulnerability-scanner/scripts/security_scan.py +0 -458
- package/.agent/skills/web-design-guidelines/SKILL.md +0 -57
- package/.agent/skills/webapp-testing/SKILL.md +0 -187
- package/.agent/skills/webapp-testing/scripts/playwright_runner.py +0 -173
- package/.agent/workflows/brainstorm.md +0 -113
- package/.agent/workflows/create.md +0 -59
- package/.agent/workflows/debug.md +0 -103
- package/.agent/workflows/deploy.md +0 -176
- package/.agent/workflows/enhance.md +0 -63
- package/.agent/workflows/orchestrate.md +0 -237
- package/.agent/workflows/plan.md +0 -89
- package/.agent/workflows/preview.md +0 -81
- package/.agent/workflows/status.md +0 -86
- package/.agent/workflows/test.md +0 -144
- package/.agent/workflows/ui-ux-pro-max.md +0 -296
|
@@ -1,172 +0,0 @@
|
|
|
1
|
-
#!/usr/bin/env python3
|
|
2
|
-
"""
|
|
3
|
-
Lint Runner - Unified linting and type checking
|
|
4
|
-
Runs appropriate linters based on project type.
|
|
5
|
-
|
|
6
|
-
Usage:
|
|
7
|
-
python lint_runner.py <project_path>
|
|
8
|
-
|
|
9
|
-
Supports:
|
|
10
|
-
- Node.js: npm run lint, npx tsc --noEmit
|
|
11
|
-
- Python: ruff check, mypy
|
|
12
|
-
"""
|
|
13
|
-
|
|
14
|
-
import subprocess
|
|
15
|
-
import sys
|
|
16
|
-
import json
|
|
17
|
-
from pathlib import Path
|
|
18
|
-
from datetime import datetime
|
|
19
|
-
|
|
20
|
-
# Fix Windows console encoding
|
|
21
|
-
try:
|
|
22
|
-
sys.stdout.reconfigure(encoding='utf-8', errors='replace')
|
|
23
|
-
except:
|
|
24
|
-
pass
|
|
25
|
-
|
|
26
|
-
|
|
27
|
-
def detect_project_type(project_path: Path) -> dict:
|
|
28
|
-
"""Detect project type and available linters."""
|
|
29
|
-
result = {
|
|
30
|
-
"type": "unknown",
|
|
31
|
-
"linters": []
|
|
32
|
-
}
|
|
33
|
-
|
|
34
|
-
# Node.js project
|
|
35
|
-
package_json = project_path / "package.json"
|
|
36
|
-
if package_json.exists():
|
|
37
|
-
result["type"] = "node"
|
|
38
|
-
try:
|
|
39
|
-
pkg = json.loads(package_json.read_text(encoding='utf-8'))
|
|
40
|
-
scripts = pkg.get("scripts", {})
|
|
41
|
-
deps = {**pkg.get("dependencies", {}), **pkg.get("devDependencies", {})}
|
|
42
|
-
|
|
43
|
-
# Check for lint script
|
|
44
|
-
if "lint" in scripts:
|
|
45
|
-
result["linters"].append({"name": "npm lint", "cmd": ["npm", "run", "lint"]})
|
|
46
|
-
elif "eslint" in deps:
|
|
47
|
-
result["linters"].append({"name": "eslint", "cmd": ["npx", "eslint", "."]})
|
|
48
|
-
|
|
49
|
-
# Check for TypeScript
|
|
50
|
-
if "typescript" in deps or (project_path / "tsconfig.json").exists():
|
|
51
|
-
result["linters"].append({"name": "tsc", "cmd": ["npx", "tsc", "--noEmit"]})
|
|
52
|
-
|
|
53
|
-
except:
|
|
54
|
-
pass
|
|
55
|
-
|
|
56
|
-
# Python project
|
|
57
|
-
if (project_path / "pyproject.toml").exists() or (project_path / "requirements.txt").exists():
|
|
58
|
-
result["type"] = "python"
|
|
59
|
-
|
|
60
|
-
# Check for ruff
|
|
61
|
-
result["linters"].append({"name": "ruff", "cmd": ["ruff", "check", "."]})
|
|
62
|
-
|
|
63
|
-
# Check for mypy
|
|
64
|
-
if (project_path / "mypy.ini").exists() or (project_path / "pyproject.toml").exists():
|
|
65
|
-
result["linters"].append({"name": "mypy", "cmd": ["mypy", "."]})
|
|
66
|
-
|
|
67
|
-
return result
|
|
68
|
-
|
|
69
|
-
|
|
70
|
-
def run_linter(linter: dict, cwd: Path) -> dict:
|
|
71
|
-
"""Run a single linter and return results."""
|
|
72
|
-
result = {
|
|
73
|
-
"name": linter["name"],
|
|
74
|
-
"passed": False,
|
|
75
|
-
"output": "",
|
|
76
|
-
"error": ""
|
|
77
|
-
}
|
|
78
|
-
|
|
79
|
-
try:
|
|
80
|
-
proc = subprocess.run(
|
|
81
|
-
linter["cmd"],
|
|
82
|
-
cwd=str(cwd),
|
|
83
|
-
capture_output=True,
|
|
84
|
-
text=True,
|
|
85
|
-
encoding='utf-8',
|
|
86
|
-
errors='replace',
|
|
87
|
-
timeout=120
|
|
88
|
-
)
|
|
89
|
-
|
|
90
|
-
result["output"] = proc.stdout[:2000] if proc.stdout else ""
|
|
91
|
-
result["error"] = proc.stderr[:500] if proc.stderr else ""
|
|
92
|
-
result["passed"] = proc.returncode == 0
|
|
93
|
-
|
|
94
|
-
except FileNotFoundError:
|
|
95
|
-
result["error"] = f"Command not found: {linter['cmd'][0]}"
|
|
96
|
-
except subprocess.TimeoutExpired:
|
|
97
|
-
result["error"] = "Timeout after 120s"
|
|
98
|
-
except Exception as e:
|
|
99
|
-
result["error"] = str(e)
|
|
100
|
-
|
|
101
|
-
return result
|
|
102
|
-
|
|
103
|
-
|
|
104
|
-
def main():
|
|
105
|
-
project_path = Path(sys.argv[1] if len(sys.argv) > 1 else ".").resolve()
|
|
106
|
-
|
|
107
|
-
print(f"\n{'='*60}")
|
|
108
|
-
print(f"[LINT RUNNER] Unified Linting")
|
|
109
|
-
print(f"{'='*60}")
|
|
110
|
-
print(f"Project: {project_path}")
|
|
111
|
-
print(f"Time: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}")
|
|
112
|
-
|
|
113
|
-
# Detect project type
|
|
114
|
-
project_info = detect_project_type(project_path)
|
|
115
|
-
print(f"Type: {project_info['type']}")
|
|
116
|
-
print(f"Linters: {len(project_info['linters'])}")
|
|
117
|
-
print("-"*60)
|
|
118
|
-
|
|
119
|
-
if not project_info["linters"]:
|
|
120
|
-
print("No linters found for this project type.")
|
|
121
|
-
output = {
|
|
122
|
-
"script": "lint_runner",
|
|
123
|
-
"project": str(project_path),
|
|
124
|
-
"type": project_info["type"],
|
|
125
|
-
"checks": [],
|
|
126
|
-
"passed": True,
|
|
127
|
-
"message": "No linters configured"
|
|
128
|
-
}
|
|
129
|
-
print(json.dumps(output, indent=2))
|
|
130
|
-
sys.exit(0)
|
|
131
|
-
|
|
132
|
-
# Run each linter
|
|
133
|
-
results = []
|
|
134
|
-
all_passed = True
|
|
135
|
-
|
|
136
|
-
for linter in project_info["linters"]:
|
|
137
|
-
print(f"\nRunning: {linter['name']}...")
|
|
138
|
-
result = run_linter(linter, project_path)
|
|
139
|
-
results.append(result)
|
|
140
|
-
|
|
141
|
-
if result["passed"]:
|
|
142
|
-
print(f" [PASS] {linter['name']}")
|
|
143
|
-
else:
|
|
144
|
-
print(f" [FAIL] {linter['name']}")
|
|
145
|
-
if result["error"]:
|
|
146
|
-
print(f" Error: {result['error'][:200]}")
|
|
147
|
-
all_passed = False
|
|
148
|
-
|
|
149
|
-
# Summary
|
|
150
|
-
print("\n" + "="*60)
|
|
151
|
-
print("SUMMARY")
|
|
152
|
-
print("="*60)
|
|
153
|
-
|
|
154
|
-
for r in results:
|
|
155
|
-
icon = "[PASS]" if r["passed"] else "[FAIL]"
|
|
156
|
-
print(f"{icon} {r['name']}")
|
|
157
|
-
|
|
158
|
-
output = {
|
|
159
|
-
"script": "lint_runner",
|
|
160
|
-
"project": str(project_path),
|
|
161
|
-
"type": project_info["type"],
|
|
162
|
-
"checks": results,
|
|
163
|
-
"passed": all_passed
|
|
164
|
-
}
|
|
165
|
-
|
|
166
|
-
print("\n" + json.dumps(output, indent=2))
|
|
167
|
-
|
|
168
|
-
sys.exit(0 if all_passed else 1)
|
|
169
|
-
|
|
170
|
-
|
|
171
|
-
if __name__ == "__main__":
|
|
172
|
-
main()
|
|
@@ -1,173 +0,0 @@
|
|
|
1
|
-
#!/usr/bin/env python3
|
|
2
|
-
"""
|
|
3
|
-
Type Coverage Checker - Measures TypeScript/Python type coverage.
|
|
4
|
-
Identifies untyped functions, any usage, and type safety issues.
|
|
5
|
-
"""
|
|
6
|
-
import sys
|
|
7
|
-
import re
|
|
8
|
-
import subprocess
|
|
9
|
-
from pathlib import Path
|
|
10
|
-
|
|
11
|
-
# Fix Windows console encoding for Unicode output
|
|
12
|
-
try:
|
|
13
|
-
sys.stdout.reconfigure(encoding='utf-8', errors='replace')
|
|
14
|
-
sys.stderr.reconfigure(encoding='utf-8', errors='replace')
|
|
15
|
-
except AttributeError:
|
|
16
|
-
pass # Python < 3.7
|
|
17
|
-
|
|
18
|
-
def check_typescript_coverage(project_path: Path) -> dict:
|
|
19
|
-
"""Check TypeScript type coverage."""
|
|
20
|
-
issues = []
|
|
21
|
-
passed = []
|
|
22
|
-
stats = {'any_count': 0, 'untyped_functions': 0, 'total_functions': 0}
|
|
23
|
-
|
|
24
|
-
ts_files = list(project_path.rglob("*.ts")) + list(project_path.rglob("*.tsx"))
|
|
25
|
-
ts_files = [f for f in ts_files if 'node_modules' not in str(f) and '.d.ts' not in str(f)]
|
|
26
|
-
|
|
27
|
-
if not ts_files:
|
|
28
|
-
return {'type': 'typescript', 'files': 0, 'passed': [], 'issues': ["[!] No TypeScript files found"], 'stats': stats}
|
|
29
|
-
|
|
30
|
-
for file_path in ts_files[:30]: # Limit
|
|
31
|
-
try:
|
|
32
|
-
content = file_path.read_text(encoding='utf-8', errors='ignore')
|
|
33
|
-
|
|
34
|
-
# Count 'any' usage
|
|
35
|
-
any_matches = re.findall(r':\s*any\b', content)
|
|
36
|
-
stats['any_count'] += len(any_matches)
|
|
37
|
-
|
|
38
|
-
# Find functions without return types
|
|
39
|
-
# function name(params) { - no return type
|
|
40
|
-
untyped = re.findall(r'function\s+\w+\s*\([^)]*\)\s*{', content)
|
|
41
|
-
# Arrow functions without types: const fn = (x) => or (x) =>
|
|
42
|
-
untyped += re.findall(r'=\s*\([^:)]*\)\s*=>', content)
|
|
43
|
-
stats['untyped_functions'] += len(untyped)
|
|
44
|
-
|
|
45
|
-
# Count typed functions
|
|
46
|
-
typed = re.findall(r'function\s+\w+\s*\([^)]*\)\s*:\s*\w+', content)
|
|
47
|
-
typed += re.findall(r':\s*\([^)]*\)\s*=>\s*\w+', content)
|
|
48
|
-
stats['total_functions'] += len(typed) + len(untyped)
|
|
49
|
-
|
|
50
|
-
except Exception:
|
|
51
|
-
continue
|
|
52
|
-
|
|
53
|
-
# Analyze results
|
|
54
|
-
if stats['any_count'] == 0:
|
|
55
|
-
passed.append("[OK] No 'any' types found")
|
|
56
|
-
elif stats['any_count'] <= 5:
|
|
57
|
-
issues.append(f"[!] {stats['any_count']} 'any' types found (acceptable)")
|
|
58
|
-
else:
|
|
59
|
-
issues.append(f"[X] {stats['any_count']} 'any' types found (too many)")
|
|
60
|
-
|
|
61
|
-
if stats['total_functions'] > 0:
|
|
62
|
-
typed_ratio = (stats['total_functions'] - stats['untyped_functions']) / stats['total_functions'] * 100
|
|
63
|
-
if typed_ratio >= 80:
|
|
64
|
-
passed.append(f"[OK] Type coverage: {typed_ratio:.0f}%")
|
|
65
|
-
elif typed_ratio >= 50:
|
|
66
|
-
issues.append(f"[!] Type coverage: {typed_ratio:.0f}% (improve)")
|
|
67
|
-
else:
|
|
68
|
-
issues.append(f"[X] Type coverage: {typed_ratio:.0f}% (too low)")
|
|
69
|
-
|
|
70
|
-
passed.append(f"[OK] Analyzed {len(ts_files)} TypeScript files")
|
|
71
|
-
|
|
72
|
-
return {'type': 'typescript', 'files': len(ts_files), 'passed': passed, 'issues': issues, 'stats': stats}
|
|
73
|
-
|
|
74
|
-
def check_python_coverage(project_path: Path) -> dict:
|
|
75
|
-
"""Check Python type hints coverage."""
|
|
76
|
-
issues = []
|
|
77
|
-
passed = []
|
|
78
|
-
stats = {'untyped_functions': 0, 'typed_functions': 0, 'any_count': 0}
|
|
79
|
-
|
|
80
|
-
py_files = list(project_path.rglob("*.py"))
|
|
81
|
-
py_files = [f for f in py_files if not any(x in str(f) for x in ['venv', '__pycache__', '.git', 'node_modules'])]
|
|
82
|
-
|
|
83
|
-
if not py_files:
|
|
84
|
-
return {'type': 'python', 'files': 0, 'passed': [], 'issues': ["[!] No Python files found"], 'stats': stats}
|
|
85
|
-
|
|
86
|
-
for file_path in py_files[:30]: # Limit
|
|
87
|
-
try:
|
|
88
|
-
content = file_path.read_text(encoding='utf-8', errors='ignore')
|
|
89
|
-
|
|
90
|
-
# Count Any usage
|
|
91
|
-
any_matches = re.findall(r':\s*Any\b', content)
|
|
92
|
-
stats['any_count'] += len(any_matches)
|
|
93
|
-
|
|
94
|
-
# Find functions with type hints
|
|
95
|
-
typed_funcs = re.findall(r'def\s+\w+\s*\([^)]*:[^)]+\)', content)
|
|
96
|
-
typed_funcs += re.findall(r'def\s+\w+\s*\([^)]*\)\s*->', content)
|
|
97
|
-
stats['typed_functions'] += len(typed_funcs)
|
|
98
|
-
|
|
99
|
-
# Find functions without type hints
|
|
100
|
-
all_funcs = re.findall(r'def\s+\w+\s*\(', content)
|
|
101
|
-
stats['untyped_functions'] += len(all_funcs) - len(typed_funcs)
|
|
102
|
-
|
|
103
|
-
except Exception:
|
|
104
|
-
continue
|
|
105
|
-
|
|
106
|
-
total = stats['typed_functions'] + stats['untyped_functions']
|
|
107
|
-
|
|
108
|
-
if total > 0:
|
|
109
|
-
typed_ratio = stats['typed_functions'] / total * 100
|
|
110
|
-
if typed_ratio >= 70:
|
|
111
|
-
passed.append(f"[OK] Type hints coverage: {typed_ratio:.0f}%")
|
|
112
|
-
elif typed_ratio >= 40:
|
|
113
|
-
issues.append(f"[!] Type hints coverage: {typed_ratio:.0f}%")
|
|
114
|
-
else:
|
|
115
|
-
issues.append(f"[X] Type hints coverage: {typed_ratio:.0f}% (add type hints)")
|
|
116
|
-
|
|
117
|
-
if stats['any_count'] == 0:
|
|
118
|
-
passed.append("[OK] No 'Any' types found")
|
|
119
|
-
elif stats['any_count'] <= 3:
|
|
120
|
-
issues.append(f"[!] {stats['any_count']} 'Any' types found")
|
|
121
|
-
else:
|
|
122
|
-
issues.append(f"[X] {stats['any_count']} 'Any' types found")
|
|
123
|
-
|
|
124
|
-
passed.append(f"[OK] Analyzed {len(py_files)} Python files")
|
|
125
|
-
|
|
126
|
-
return {'type': 'python', 'files': len(py_files), 'passed': passed, 'issues': issues, 'stats': stats}
|
|
127
|
-
|
|
128
|
-
def main():
|
|
129
|
-
target = sys.argv[1] if len(sys.argv) > 1 else "."
|
|
130
|
-
project_path = Path(target)
|
|
131
|
-
|
|
132
|
-
print("\n" + "=" * 60)
|
|
133
|
-
print(" TYPE COVERAGE CHECKER")
|
|
134
|
-
print("=" * 60 + "\n")
|
|
135
|
-
|
|
136
|
-
results = []
|
|
137
|
-
|
|
138
|
-
# Check TypeScript
|
|
139
|
-
ts_result = check_typescript_coverage(project_path)
|
|
140
|
-
if ts_result['files'] > 0:
|
|
141
|
-
results.append(ts_result)
|
|
142
|
-
|
|
143
|
-
# Check Python
|
|
144
|
-
py_result = check_python_coverage(project_path)
|
|
145
|
-
if py_result['files'] > 0:
|
|
146
|
-
results.append(py_result)
|
|
147
|
-
|
|
148
|
-
if not results:
|
|
149
|
-
print("[!] No TypeScript or Python files found.")
|
|
150
|
-
sys.exit(0)
|
|
151
|
-
|
|
152
|
-
# Print results
|
|
153
|
-
critical_issues = 0
|
|
154
|
-
for result in results:
|
|
155
|
-
print(f"\n[{result['type'].upper()}]")
|
|
156
|
-
print("-" * 40)
|
|
157
|
-
for item in result['passed']:
|
|
158
|
-
print(f" {item}")
|
|
159
|
-
for item in result['issues']:
|
|
160
|
-
print(f" {item}")
|
|
161
|
-
if item.startswith("[X]"):
|
|
162
|
-
critical_issues += 1
|
|
163
|
-
|
|
164
|
-
print("\n" + "=" * 60)
|
|
165
|
-
if critical_issues == 0:
|
|
166
|
-
print("[OK] TYPE COVERAGE: ACCEPTABLE")
|
|
167
|
-
sys.exit(0)
|
|
168
|
-
else:
|
|
169
|
-
print(f"[X] TYPE COVERAGE: {critical_issues} critical issues")
|
|
170
|
-
sys.exit(1)
|
|
171
|
-
|
|
172
|
-
if __name__ == "__main__":
|
|
173
|
-
main()
|
|
@@ -1,41 +0,0 @@
|
|
|
1
|
-
---
|
|
2
|
-
description: "Kinh nghiệm xương máu khi định tuyến LLM: 9Router (Gemini) vs Ollama (Bartowski/Reasoning)"
|
|
3
|
-
skills: ["clean-code"]
|
|
4
|
-
---
|
|
5
|
-
|
|
6
|
-
# LLM Routing Quirks (Kinh Nghiệm Phẫu Thuật Não LLM)
|
|
7
|
-
|
|
8
|
-
Tài liệu này lưu trữ những **chú ý sinh tử (quirks)** khi gọi các model LLM khác nhau trong kiến trúc Antigravity Orchestrator. Nó được đúc kết từ quá trình cày cuốc miệt mài vá lỗi của Quản Đốc AI và Sếp.
|
|
9
|
-
|
|
10
|
-
Nếu bạn đang viết Agent hoặc Script gọi LLM, **BẮT BUỘC ĐỌC** để tránh hệ thống bị đơ, văng lỗi JSON hoặc model bị "ngu" đi.
|
|
11
|
-
|
|
12
|
-
---
|
|
13
|
-
|
|
14
|
-
## 1. 9Router & Gemini (Cloud Models - API tương thích OpenAI)
|
|
15
|
-
|
|
16
|
-
**Mô hình áp dụng:** `ag/gemini-3-flash`, `ag/gemini-pro-1.5`, v.v.
|
|
17
|
-
|
|
18
|
-
### Các Quy tắc Vàng:
|
|
19
|
-
1. **JSON Format:** ĐƯỢC PHÉP (và nên) sử dụng `response_format: { type: 'json_object' }` để ép model trả về cục JSON chuẩn chỉ. Gemini xử lý JSON format tự nhiên rất tốt.
|
|
20
|
-
2. **Stream Override (Lỗi SSE):** Mặc dù truyền `stream: false`, 9Router đôi khi BỊ LỖI và vẫn trả về luồng Server-Sent Events (SSE) (các dòng `data: {...}\n\ndata: [DONE]`).
|
|
21
|
-
👉 **Cách Fix:** Code call API (`axios.post`) phải luôn chuẩn bị sẵn tinh thần parse cả 2 trường hợp: JSON Object bình thường VÀ Chuỗi SSE gộp lại. (Xem hàm `OrchestratorBrain.callAI`).
|
|
22
|
-
3. **Timeout Threshold:** Bắt buộc đặt `timeout` cao (ví dụ: `120000ms` - 2 phút). Đừng set quá thấp (30s), vì proxy API Cloud thường xuyên bị trễ nhịp do routing.
|
|
23
|
-
|
|
24
|
-
---
|
|
25
|
-
|
|
26
|
-
## 2. Ollama & Bartowski (Local Reasoning Models - Chạy Offline)
|
|
27
|
-
|
|
28
|
-
**Mô hình áp dụng:** `hf.co/bartowski/nvidia_Orchestrator-8B-GGUF:Q4_K_M`, `DeepSeek-R1`, các model có thẻ `<think>`.
|
|
29
|
-
|
|
30
|
-
### Các Quy tắc Vàng (SINH TỬ):
|
|
31
|
-
1. **🔥 TUYỆT ĐỐI KHÔNG DÙNG TÍNH NĂNG ÉP JSON:** KHÔNG BAO GIỜ truyền `format: "json"` vào request của Ollama đối với các Reasoning Models!
|
|
32
|
-
- **Vì sao?** Các reasoning models cần viết tư duy vào thẻ `<think>...</think>` trước khi ra kết quả. Việc ép JSON sẽ làm hệ thống của Ollama cắt gọt thẻ suy nghĩ của nó, khiến model bị "ngăn não", sinh ra ảo giác (hallucination), trả về rác (token rác) hoặc phá hỏng format JSON cuối cùng.
|
|
33
|
-
2. **Xử lý Thẻ `<think>` Thủ Công:** Khi model trả về chuỗi text, ta phải chủ động dùng Regex để bóc mảng `<think>...</think>` ra rồi mới quăng phần còn lại vào `JSON.parse()`.
|
|
34
|
-
- Code mài giũa: `str.replace(/<think>[\s\S]*?<\/think>/g, '').trim()`
|
|
35
|
-
3. **Mở Hộp Sọ (Context Window - `num_ctx`):** Mặc định Ollama cấp `num_ctx` rất bé (2048 token) để tiết kiệm VRAM. Với các Task Orchestration quăng cả lịch sử chat và code, **BẮT BUỘC** phải truyền `"options": { "num_ctx": 4096 }` (hoặc 8192) vào `axios.post`. Nếu không, Ollama sẽ ngắt đuôi JSON làm sập hệ thống.
|
|
36
|
-
4. **Tên Model Mặc Định:** Chú ý khi tải model HuggingFace từ Ollama (ví dụ `ollama run hf.co/bartowski/...`), Ollama sẽ giữ NGHIÊM NGẶT toàn bộ đường dẫn làm `ID Model`. Luôn truyền full tên vào trường `model` (ví dụ `model: "hf.co/bartowski/nvidia_Orchestrator-8B-GGUF:Q4_K_M"`).
|
|
37
|
-
|
|
38
|
-
---
|
|
39
|
-
|
|
40
|
-
## 3. Khuyến nghị Tự Chữa Lành (Mini-Antigravity Check)
|
|
41
|
-
Bất cứ khi nào gọi API nội bộ (để tự sửa mã) và nhận cảnh báo `SKIP Lỗi LLM` (tức 9Router bị câm), hệ thống không được gọi lại chính thằng sụp mạng, mà phải rẽ nhánh BẮN THẲNG LỆNH GỌI OLLAMA LOCALLY (tắt bypass Router) để bảo tồn mạng sống.
|
|
@@ -1,176 +0,0 @@
|
|
|
1
|
-
---
|
|
2
|
-
name: mcp-builder
|
|
3
|
-
description: MCP (Model Context Protocol) server building principles. Tool design, resource patterns, best practices.
|
|
4
|
-
allowed-tools: Read, Write, Edit, Glob, Grep
|
|
5
|
-
---
|
|
6
|
-
|
|
7
|
-
# MCP Builder
|
|
8
|
-
|
|
9
|
-
> Principles for building MCP servers.
|
|
10
|
-
|
|
11
|
-
---
|
|
12
|
-
|
|
13
|
-
## 1. MCP Overview
|
|
14
|
-
|
|
15
|
-
### What is MCP?
|
|
16
|
-
|
|
17
|
-
Model Context Protocol - standard for connecting AI systems with external tools and data sources.
|
|
18
|
-
|
|
19
|
-
### Core Concepts
|
|
20
|
-
|
|
21
|
-
| Concept | Purpose |
|
|
22
|
-
|---------|---------|
|
|
23
|
-
| **Tools** | Functions AI can call |
|
|
24
|
-
| **Resources** | Data AI can read |
|
|
25
|
-
| **Prompts** | Pre-defined prompt templates |
|
|
26
|
-
|
|
27
|
-
---
|
|
28
|
-
|
|
29
|
-
## 2. Server Architecture
|
|
30
|
-
|
|
31
|
-
### Project Structure
|
|
32
|
-
|
|
33
|
-
```
|
|
34
|
-
my-mcp-server/
|
|
35
|
-
├── src/
|
|
36
|
-
│ └── index.ts # Main entry
|
|
37
|
-
├── package.json
|
|
38
|
-
└── tsconfig.json
|
|
39
|
-
```
|
|
40
|
-
|
|
41
|
-
### Transport Types
|
|
42
|
-
|
|
43
|
-
| Type | Use |
|
|
44
|
-
|------|-----|
|
|
45
|
-
| **Stdio** | Local, CLI-based |
|
|
46
|
-
| **SSE** | Web-based, streaming |
|
|
47
|
-
| **WebSocket** | Real-time, bidirectional |
|
|
48
|
-
|
|
49
|
-
---
|
|
50
|
-
|
|
51
|
-
## 3. Tool Design Principles
|
|
52
|
-
|
|
53
|
-
### Good Tool Design
|
|
54
|
-
|
|
55
|
-
| Principle | Description |
|
|
56
|
-
|-----------|-------------|
|
|
57
|
-
| Clear name | Action-oriented (get_weather, create_user) |
|
|
58
|
-
| Single purpose | One thing well |
|
|
59
|
-
| Validated input | Schema with types and descriptions |
|
|
60
|
-
| Structured output | Predictable response format |
|
|
61
|
-
|
|
62
|
-
### Input Schema Design
|
|
63
|
-
|
|
64
|
-
| Field | Required? |
|
|
65
|
-
|-------|-----------|
|
|
66
|
-
| Type | Yes - object |
|
|
67
|
-
| Properties | Define each param |
|
|
68
|
-
| Required | List mandatory params |
|
|
69
|
-
| Description | Human-readable |
|
|
70
|
-
|
|
71
|
-
---
|
|
72
|
-
|
|
73
|
-
## 4. Resource Patterns
|
|
74
|
-
|
|
75
|
-
### Resource Types
|
|
76
|
-
|
|
77
|
-
| Type | Use |
|
|
78
|
-
|------|-----|
|
|
79
|
-
| Static | Fixed data (config, docs) |
|
|
80
|
-
| Dynamic | Generated on request |
|
|
81
|
-
| Template | URI with parameters |
|
|
82
|
-
|
|
83
|
-
### URI Patterns
|
|
84
|
-
|
|
85
|
-
| Pattern | Example |
|
|
86
|
-
|---------|---------|
|
|
87
|
-
| Fixed | `docs://readme` |
|
|
88
|
-
| Parameterized | `users://{userId}` |
|
|
89
|
-
| Collection | `files://project/*` |
|
|
90
|
-
|
|
91
|
-
---
|
|
92
|
-
|
|
93
|
-
## 5. Error Handling
|
|
94
|
-
|
|
95
|
-
### Error Types
|
|
96
|
-
|
|
97
|
-
| Situation | Response |
|
|
98
|
-
|-----------|----------|
|
|
99
|
-
| Invalid params | Validation error message |
|
|
100
|
-
| Not found | Clear "not found" |
|
|
101
|
-
| Server error | Generic error, log details |
|
|
102
|
-
|
|
103
|
-
### Best Practices
|
|
104
|
-
|
|
105
|
-
- Return structured errors
|
|
106
|
-
- Don't expose internal details
|
|
107
|
-
- Log for debugging
|
|
108
|
-
- Provide actionable messages
|
|
109
|
-
|
|
110
|
-
---
|
|
111
|
-
|
|
112
|
-
## 6. Multimodal Handling
|
|
113
|
-
|
|
114
|
-
### Supported Types
|
|
115
|
-
|
|
116
|
-
| Type | Encoding |
|
|
117
|
-
|------|----------|
|
|
118
|
-
| Text | Plain text |
|
|
119
|
-
| Images | Base64 + MIME type |
|
|
120
|
-
| Files | Base64 + MIME type |
|
|
121
|
-
|
|
122
|
-
---
|
|
123
|
-
|
|
124
|
-
## 7. Security Principles
|
|
125
|
-
|
|
126
|
-
### Input Validation
|
|
127
|
-
|
|
128
|
-
- Validate all tool inputs
|
|
129
|
-
- Sanitize user-provided data
|
|
130
|
-
- Limit resource access
|
|
131
|
-
|
|
132
|
-
### API Keys
|
|
133
|
-
|
|
134
|
-
- Use environment variables
|
|
135
|
-
- Don't log secrets
|
|
136
|
-
- Validate permissions
|
|
137
|
-
|
|
138
|
-
---
|
|
139
|
-
|
|
140
|
-
## 8. Configuration
|
|
141
|
-
|
|
142
|
-
### Claude Desktop Config
|
|
143
|
-
|
|
144
|
-
| Field | Purpose |
|
|
145
|
-
|-------|---------|
|
|
146
|
-
| command | Executable to run |
|
|
147
|
-
| args | Command arguments |
|
|
148
|
-
| env | Environment variables |
|
|
149
|
-
|
|
150
|
-
---
|
|
151
|
-
|
|
152
|
-
## 9. Testing
|
|
153
|
-
|
|
154
|
-
### Test Categories
|
|
155
|
-
|
|
156
|
-
| Type | Focus |
|
|
157
|
-
|------|-------|
|
|
158
|
-
| Unit | Tool logic |
|
|
159
|
-
| Integration | Full server |
|
|
160
|
-
| Contract | Schema validation |
|
|
161
|
-
|
|
162
|
-
---
|
|
163
|
-
|
|
164
|
-
## 10. Best Practices Checklist
|
|
165
|
-
|
|
166
|
-
- [ ] Clear, action-oriented tool names
|
|
167
|
-
- [ ] Complete input schemas with descriptions
|
|
168
|
-
- [ ] Structured JSON output
|
|
169
|
-
- [ ] Error handling for all cases
|
|
170
|
-
- [ ] Input validation
|
|
171
|
-
- [ ] Environment-based configuration
|
|
172
|
-
- [ ] Logging for debugging
|
|
173
|
-
|
|
174
|
-
---
|
|
175
|
-
|
|
176
|
-
> **Remember:** MCP tools should be simple, focused, and well-documented. The AI relies on descriptions to use them correctly.
|
|
@@ -1,107 +0,0 @@
|
|
|
1
|
-
---
|
|
2
|
-
name: memory-architecture
|
|
3
|
-
description: >
|
|
4
|
-
Hệ thống Bộ nhớ Đồ thị Đa Tầng (Multi-layered Graph Memory) cho Agent.
|
|
5
|
-
Hợp nhất giữa Progressive Disclosure (L1, L2, L3) và Temporal Knowledge Graph (Node-Edge).
|
|
6
|
-
---
|
|
7
|
-
|
|
8
|
-
# Multi-layered Graph Memory Architecture
|
|
9
|
-
|
|
10
|
-
> 🔴 **CRITICAL RULE**: Trí nhớ của Agent không còn là Text phẳng. Nó là một **Đồ thị Tri thức Đa tầng (Graph)** cho phép "Zoom-out" nhìn toàn cảnh và "Zoom-in" đọc chi tiết lõi.
|
|
11
|
-
|
|
12
|
-
---
|
|
13
|
-
|
|
14
|
-
## 1. Nguyên tắc cốt lõi: 3 Tầng Ký ức (L1/L2/L3)
|
|
15
|
-
|
|
16
|
-
```text
|
|
17
|
-
L1 (Zoom Out) → session_summary & Core_Nodes — "Toàn cảnh dự án, các khái niệm lõi"
|
|
18
|
-
L2 (Mid) → category-indexed & Edges — "Mạng lưới quan hệ, mạch sự kiện"
|
|
19
|
-
L3 (Zoom In) → detail_record & Solutions — "Chi tiết lỗi, code snippet gốc"
|
|
20
|
-
```
|
|
21
|
-
|
|
22
|
-
---
|
|
23
|
-
|
|
24
|
-
## 2. Hình thái Dữ liệu cấp thấp (Node-Edge Graph)
|
|
25
|
-
|
|
26
|
-
Mọi dữ liệu trong tầng L1/L2/L3 đều được lưu dưới dạng Đồ thị (Graph):
|
|
27
|
-
|
|
28
|
-
### A. NODES (Hạt nhân Kiến thức)
|
|
29
|
-
Các khái niệm định danh tĩnh. Vd: `[Node:DB_PORT] => 5432 (Energy: 100)`
|
|
30
|
-
|
|
31
|
-
### B. EDGES / EPISODES (Mạch Câu chuyện)
|
|
32
|
-
Luồng sự kiện liên kết. Vd: `[Node:USER] --(muốn mở port)--> [Node:DB_PORT]`
|
|
33
|
-
|
|
34
|
-
### Progressive Disclosure (Tiết kiệm Token):
|
|
35
|
-
Thay vì load toàn bộ Memory, hệ thống truy vấn tự động theo dạng Phóng to/Thu nhỏ:
|
|
36
|
-
1. **Search (Zoom Out):** Chỉ lấy mỏ neo 1-2 Node lõi + Quan hệ (L1/L2) (Tốn rất ít token).
|
|
37
|
-
2. **Details (Zoom In):** Nếu AI cần phân tích sâu 1 đoạn code phức tạp, gọi L3 (Token cao hơn nhưng cực kỳ chính xác).
|
|
38
|
-
|
|
39
|
-
---
|
|
40
|
-
|
|
41
|
-
## 3. Vòng đời Trí nhớ (Biocybernetic Lifecycle)
|
|
42
|
-
|
|
43
|
-
1. **Temporal Decay (Phai mờ):** Mỗi Node và Edge có `Energy`. Bị hao hụt tự động mỗi 24h. L1 siêu bền, L3 hao nhanh.
|
|
44
|
-
2. **Reinforcement (Củng cố):** Mỗi lần "bò" (traversal) qua Node, Energy phục hồi.
|
|
45
|
-
3. **Consolidation (Ngủ):** Cứ đêm đến, AI gộp các mạch nhớ L3/L2 rườm rà thành 1 Node L1 khái quát siêu bền.
|
|
46
|
-
4. **Forgetting (Quên lãng):** Node có `Energy < 0` HOẶC mồ côi (không ai gọi tới) sẽ BỊ XOÁ (Quét rác sinh học).
|
|
47
|
-
|
|
48
|
-
---
|
|
49
|
-
|
|
50
|
-
## 4. Bảng Phân loại Khối Lượng Ký ức
|
|
51
|
-
|
|
52
|
-
| Target / Kiểu Node | Tầng | Giải thích | Energy Khởi tạo | Mức độ Tự Phân Hủy (Decay) |
|
|
53
|
-
|---|---|---|---|---|
|
|
54
|
-
| `session_summary` | L1 | Toàn cảnh phiên làm việc | 100 | Lâu dài (Gần như không) |
|
|
55
|
-
| `Core_Fact` | L1 | Cấu hình, IP, Convention cốt lõi | 100 | Gần như không |
|
|
56
|
-
| `User_Pref` | L2 | Thói quen user | 80 | Cực chậm |
|
|
57
|
-
| `project_context` | L2 | Ngữ cảnh dự án hiện tại | 80 | Trung hạn (Cập nhật đổi version) |
|
|
58
|
-
| `Episode_Edge` | L2 | "Hôm qua đã làm gì" | 30 | Nhanh (Rất mau phai mờ) |
|
|
59
|
-
| `error_solution` | L3 | Bug log + Code fix cụ thể | 50 | Trung bình (Công nghệ đổi -> Xóa) |
|
|
60
|
-
|
|
61
|
-
---
|
|
62
|
-
|
|
63
|
-
## 5. Workflow Tương tác Trí Nhớ (Mô phỏng GraphRAG)
|
|
64
|
-
|
|
65
|
-
### KHI BẮT ĐẦU task mới (Retrieve):
|
|
66
|
-
```text
|
|
67
|
-
1. Bắn 1 tia Vector Search thả mỏ neo vào Đồ thị (Tìm L1/L2).
|
|
68
|
-
2. Khôi phục lại Mạch Câu Chuyện bằng cách "Bò" dọc theo Edges lân cận (BFS).
|
|
69
|
-
3. Nếu cần code chính xác của 1 Bug, gọi lệnh lấy Detail (L3).
|
|
70
|
-
```
|
|
71
|
-
|
|
72
|
-
### KHI KẾT THÚC task (Commit):
|
|
73
|
-
❌ **Sai:** Lưu text rác `memory_save("Fix xong bug X")` vào trong hư không.
|
|
74
|
-
✅ **Đúng:**
|
|
75
|
-
```text
|
|
76
|
-
1. Khởi tạo Node mới: Node[Fix_CORS] = "Thiếu accept headers"
|
|
77
|
-
2. Nối Mạch: Node[Env_Prod] --(gặp lỗi)=> Node[Fix_CORS]
|
|
78
|
-
3. (Tùy chọn) Gắn thêm Detail record vào L3 nếu fix cực phức tạp cần giữ code snippet.
|
|
79
|
-
```
|
|
80
|
-
|
|
81
|
-
---
|
|
82
|
-
|
|
83
|
-
## 6. Anti-Patterns Kịch Độc
|
|
84
|
-
|
|
85
|
-
```text
|
|
86
|
-
❌ Gọi L3 load toàn bộ chi tiết khi không cần thiết (Tràn token).
|
|
87
|
-
❌ Cố gắng chèn 200 dòng code vào Tên của Node (Node chỉ giữ Context lõi, Code đẩy xuống L3).
|
|
88
|
-
❌ Tạo Node rác không gán Edge quan hệ. (Garbage Collector sẽ chém thẳng tay).
|
|
89
|
-
```
|
|
90
|
-
|
|
91
|
-
---
|
|
92
|
-
|
|
93
|
-
## 7. Giao thức Tự động Dịch chuyển Ký ức (Self-Migration Protocol)
|
|
94
|
-
|
|
95
|
-
> **Dành cho các AI Agent hoạt động trong dự án cũ (Legacy Projects) vừa được cập nhật `ag-kit` thông qua `npx`.**
|
|
96
|
-
|
|
97
|
-
Khi bạn (AI) nhận thức được mình đang ở trong một dự án đã có sẵn dữ liệu ký ức cũ (chứa trong các file `.jsonl` hoặc `memory.db` phẳng), bạn **CÓ TRÁCH NHIỆM TỰ ĐỘNG** thực hiện quá trình "Memory Migration" (Dịch chuyển Ký ức) mà không cần User yêu cầu, theo các bước sau:
|
|
98
|
-
|
|
99
|
-
1. **Phát hiện:** Dùng lệnh đọc file hoặc MCP để quét nội dung các ký ức cũ trong thư mục `data/memories/` (nếu có).
|
|
100
|
-
2. **Băm nhỏ (Chunking & Entity Extraction):** Phân tích các đoạn log cũ xem chúng nói về chủ đề gì (vd: Lịch sử fix bug, Convention code cũ, Sở thích User).
|
|
101
|
-
3. **Tái cấu trúc (Graph Construction):**
|
|
102
|
-
- Từ các đoạn text đó, rút trích ra các `Nodes` cốt lõi mang tính sự thật (Core Facts).
|
|
103
|
-
- Thiết lập các `Edges` (Mạch kết nối) mô tả luồng logic giữa các Nodes đó.
|
|
104
|
-
4. **Lưu trữ (Commit):** Lưu bộ Graph vừa tạo vào cơ sở dữ liệu mới (thông qua `GraphMemoryTool` nếu dự án đã hỗ trợ, hoặc lưu ngược lại bằng `memory_save` dưới cấu trúc **Pseudo-Graph** cực đoan: `[Node:A] --[rel]--> [Node:B]` nếu nền tảng chưa được nâng cấp DB SQLite).
|
|
105
|
-
|
|
106
|
-
Bằng giao thức này, kiến thức của dự án sẽ KHÔNG BAO GIỜ bị đứt gãy sau khi cập nhật kiến trúc. Mọi ký ức quá khứ sẽ tự động "tiến hóa" thành dạng Đồ thị tương lai!
|
|
107
|
-
|