README นี้แปลเป็นภาษาไทย โดยใช้ README ภาษาอังกฤษ เป็นแหล่งอ้างอิงหลัก
เขียนการทำงานภายในโมเดล ไปป์ไลน์การค้นคืน และรันไทม์ของเอเจนต์ ทดสอบ ตรวจสอบความล้มเหลว และเก็บโค้ดกับผลการประเมินไว้
**[เริ่มเรียน](https://aiengineeringfromscratch.com/lesson?path=phases/00-setup-and-tooling/01-dev-environment)** · **[เลือกเส้นทาง](#learning-routes)** · **[ลองห้องทดลอง](#interactive-lab)** · **[สร้างโปรเจกต์](#project-challenges)** · **[ดูหลักสูตร](#contents)**
ฟรี เปิดเผยซอร์สโค้ด และใช้สัญญาอนุญาต MIT เรียนผ่านเว็บไซต์ กับเอเจนต์เขียนโค้ด หรือด้วยการรันโค้ดบนเครื่องของคุณ
> 523 บทเรียน. 20 ระยะ. Python, TypeScript, Rust, Julia.
อ่านในภาษาของคุณ
🇬🇧 English · 🇨🇳 简体中文 · 🇹🇼 繁體中文(台灣) · 🇯🇵 日本語 · 🇰🇷 한국어 · 🇵🇹 Português · 🇧🇷 Português (Brasil) · 🇪🇸 Español · 🇩🇪 Deutsch · 🇫🇷 Français · 🇮🇹 Italiano · 🇳🇱 Nederlands · 🇵🇱 Polski · 🇨🇿 Čeština · 🇷🇴 Română · 🇭🇺 Magyar · 🇬🇷 Ελληνικά · 🇸🇪 Svenska · 🇩🇰 Dansk · 🇳🇴 Norsk · 🇫🇮 Suomi · 🇷🇺 Русский · 🇺🇦 Українська · 🇹🇷 Türkçe · 🇮🇱 עברית · 🇸🇦 العربية · 🇮🇷 فارسی · 🇮🇳 हिन्दी · 🇧🇩 বাংলা · 🇵🇰 اردو · 🇹🇭 ไทย · 🇻🇳 Tiếng Việt · 🇮🇩 Bahasa Indonesia · 🇵🇭 Tagalog
### ผู้สนับสนุน
การสนับสนุนของคุณช่วยให้ทุกบทเรียนยังคงฟรีและเป็นโอเพนซอร์ส ดูผู้สนับสนุนทั้งหมด · ร่วมเป็นผู้สนับสนุน
## เส้นทางการเรียนรู้
| เส้นทาง | บทเรียนแรก |
|---|---|
| รากฐานของโมเดล | [การตั้งค่าและเครื่องมือ](https://aiengineeringfromscratch.com/lesson?path=phases/00-setup-and-tooling/01-dev-environment) |
| ระบบ LLM | [วิศวกรรมพรอมต์](https://aiengineeringfromscratch.com/lesson?path=phases/11-llm-engineering/01-prompt-engineering) |
| เอเจนต์และการส่งมอบ | [ลูปเอเจนต์](https://aiengineeringfromscratch.com/lesson?path=phases/14-agent-engineering/01-the-agent-loop) |
[เปรียบเทียบเส้นทางอาชีพ](https://aiengineeringfromscratch.com/learning-paths.html) · [ความรู้พื้นฐานและเวลาเรียน](#study-guide)
### การไล่ระดับลง
จุดเริ่มต้น 20 จุดเคลื่อนตามการไล่ระดับลงบนฟังก์ชันความสูญเสียกำลังสอง กราฟแสดงตำแหน่งและค่าความสูญเสียเฉลี่ยหลังการอัปเดตแต่ละครั้ง
[ปรับอัตราการเรียนรู้ในบทเรียน](https://aiengineeringfromscratch.com/lesson?path=phases/01-math-foundations/08-optimization#loss-landscape-visualization) · [เปรียบเทียบ GD, โมเมนตัม และ Adam ในโค้ด](../../phases/01-math-foundations/08-optimization/code/optimizers.py)
### โปรเจกต์
สามโปรเจกต์พร้อมโค้ดเริ่มต้นแบ่งขั้น โค้ดอ้างอิง และตัวตรวจบนเครื่อง หลัง[การตั้งค่า](#local-setup)ให้รันคำสั่งจากไดเรกทอรีรากของที่เก็บโค้ด โค้ดเริ่มต้นจะตรวจไม่ผ่านจนกว่าคุณจะเขียนแต่ละขั้นเสร็จ
01 · ห้องทดลองประเมินการค้นคืน · Python · ตัวชี้วัดการจัดอันดับและการตรวจการถดถอย
ระบบที่เสนอใหม่มีค่าเฉลี่ย NDCG สูงขึ้น แต่หลักฐานที่เกี่ยวข้องที่สุดของคำค้นหนึ่งกลับอยู่ในอันดับต่ำลง สร้างการเปรียบเทียบรายคำค้นที่รายงานการถดถอยและทำให้การตรวจสอบก่อนเผยแพร่ไม่ผ่านได้
ใช้ Python 3.10+ ทบทวน[RAG](../../phases/11-llm-engineering/06-rag/docs/en.md)และ[การประเมินโมเดล](../../phases/02-ml-fundamentals/09-model-evaluation/docs/en.md) เขียนการตรวจอันดับ precision และ recall ตัวชี้วัดที่คำนึงถึงอันดับ แล้วจึงเปรียบเทียบระบบ
```bash
python3 scripts/project_test.py retrieval-evaluation-lab \
--init learning-artifacts/retrieval-evaluation-lab
python3 scripts/project_test.py retrieval-evaluation-lab \
--stage 1 --path learning-artifacts/retrieval-evaluation-lab --strict
python3 scripts/project_test.py retrieval-evaluation-lab \
--all --path learning-artifacts/retrieval-evaluation-lab --strict
```
**เก็บไว้:** ผลเปรียบเทียบที่ทำซ้ำได้ พร้อมผลต่างรายคำค้นและผลการตัดสินความเกี่ยวข้องที่ใช้ให้คะแนน ตัวชี้วัดอธิบายผลตามการตัดสินเหล่านั้น ไม่ได้ยืนยันว่าคำตอบถูกต้อง
[เริ่มโปรเจกต์](https://aiengineeringfromscratch.com/project.html?id=retrieval-evaluation-lab) · [ตรวจดูแนวทางอ้างอิง](../../projects/retrieval-evaluation-lab/solution/) · [รันกับข้อมูลของคุณเอง](../../projects/retrieval-evaluation-lab/README.md#run-with-your-own-inputs)
02 · เครื่องมือดีบักร่องรอยเอเจนต์ · TypeScript · การแยกวิเคราะห์ร่องรอยและการคำนวณเวลา
ร่องรอยที่ให้มายังใช้เวลา 100 ms แต่จำนวนโทเคนรวมเพิ่มขึ้น 200 และมีหนึ่งสแปนเริ่มล้มเหลว แยกงานของสแปนลูกที่ทับซ้อนกันออกจากเวลาที่สแปนแม่ทำงานเอง แล้วสร้างรายงานที่แสดงการเปลี่ยนแปลง
ใช้ Node.js 22.18+ และ Python 3 สำหรับตัวตรวจ เขียนการแยกวิเคราะห์ JSONL การตรวจความสัมพันธ์กับสแปนแม่ การคำนวณช่วงเวลา แล้วสร้างไทม์ไลน์ที่ตรวจดูได้
```bash
python3 scripts/project_test.py agent-trace-debugger \
--init learning-artifacts/agent-trace-debugger
python3 scripts/project_test.py agent-trace-debugger \
--stage 1 --path learning-artifacts/agent-trace-debugger --strict
python3 scripts/project_test.py agent-trace-debugger \
--all --path learning-artifacts/agent-trace-debugger --strict
```
**เก็บไว้:** ร่องรอยขาเข้า ไทม์ไลน์ HTML และรายงานการถดถอยแบบ JSON บันทึกเฉพาะโทเคนที่แต่ละสแปนใช้เอง เพื่อไม่นับการใช้งานของสแปนแม่และลูกซ้ำกัน
[เริ่มโปรเจกต์](https://aiengineeringfromscratch.com/project.html?id=agent-trace-debugger) · [ตรวจดูแนวทางอ้างอิง](../../projects/agent-trace-debugger/solution/) · [สำรวจเวลาแบบโต้ตอบ](https://aiengineeringfromscratch.com/project.html?id=agent-trace-debugger&stage=03-timing)
03 · ไฟร์วอลล์การเรียกเครื่องมือ · Rust · การตรวจบทบาทและบันทึกการอนุมัติ
เนื้อหาที่จะเขียนเปลี่ยนหลังการตรวจทาน หรือมีการนำการอนุมัติกลับมาใช้ซ้ำ ตรวจโครงสร้างคำขอ บทบาทของผู้เรียก และพาธ แล้วใช้สิทธิ์อนุมัติที่ผูกกับคำขอและเนื้อหาตรงกันทุกประการเพียงครั้งเดียว
ใช้ Rust และ Python 3.10+ ทบทวน[การออกแบบสคีมาของเครื่องมือ](../../phases/13-tools-and-protocols/05-tool-schema-design/docs/en.md)และ[ขอบเขตความปลอดภัย](../../phases/17-infrastructure-and-production/25-security-secrets-audit/docs/en.md) แอปพลิเคชันผู้เรียกเป็นผู้ระบุตัวตน ส่วนโมเดลเสนอการดำเนินการ
```bash
python3 scripts/project_test.py tool-call-firewall \
--init learning-artifacts/tool-call-firewall
python3 scripts/project_test.py tool-call-firewall \
--stage 1 --path learning-artifacts/tool-call-firewall --strict
python3 scripts/project_test.py tool-call-firewall \
--all --path learning-artifacts/tool-call-firewall --strict
```
**เก็บไว้:** บันทึกตรวจสอบที่แสดงการดำเนินการที่ร้องขอและผลตัดสินตามนโยบาย การอนุมัติใช้ได้ครั้งเดียวภายในการเรียกหนึ่งครั้ง โปรเจกต์นี้ไม่ได้ให้การอนุญาตแบบถาวรหรือแซนด์บ็อกซ์ระดับระบบปฏิบัติการ
[เริ่มโปรเจกต์](https://aiengineeringfromscratch.com/project.html?id=tool-call-firewall) · [ตรวจดูแนวทางอ้างอิง](../../projects/tool-call-firewall/solution/) · [สำรวจขอบเขตการอนุมัติ](https://aiengineeringfromscratch.com/project.html?id=tool-call-firewall&stage=03-consume-a-request-bound-approval-once)
[ดูโปรเจกต์ทั้งหมด](https://aiengineeringfromscratch.com/projects.html) · [คู่มือฝึกทักษะเพื่ออาชีพ](../../learning-paths/CAREER-PRACTICE.md)
## เลือกวิธีเรียน
### บนเว็บไซต์
เปิดบทเรียนที่เสร็จแล้วบน [aiengineeringfromscratch.com](https://aiengineeringfromscratch.com) หรือขยายระยะใน[สารบัญ](#contents) ไม่ต้องตั้งค่าหรือโคลน
### กับติวเตอร์ AI
หากติดตั้ง Node.js, `npx` และเอเจนต์เขียนโค้ดที่รองรับสกิลแล้ว คุณเปลี่ยนเอเจนต์ให้เป็นผู้สอนได้ ไม่ต้องโคลนที่เก็บโค้ดเพื่อติดตั้งหรืออ่านเนื้อหาของผู้สอน แต่แล็บในเส้นทางเฉพาะด้านต้องใช้ `python3` ส่วนแล็บ Agent Skills ต้องเลือกโฮสต์และมีโฟลเดอร์สกิลระดับผู้ใช้หรือโครงการที่เขียนได้ด้วย
```bash
npx skills add rohitg00/ai-engineering-from-scratch
```
เลือกโฮสต์และขอบเขตเมื่อตัวติดตั้งถาม ใช้ `start-learning` ใน Codex, `/start-learning` ใน Claude Code หรือขอให้โฮสต์ใช้สกิลตามชื่อ
การตั้งค่าติวเตอร์และคำสั่งของโฮสต์
ตรวจสอบข้อกำหนดในเครื่องก่อน:
```bash
node --version
npx --version
python3 --version
```
`skills` เขียนไฟล์ไปยังโฮสต์และขอบเขตที่เลือกระหว่างติดตั้ง เช่น `.claude/skills/`, `.cursor/skills/`, `.codex/skills/` หรือโฟลเดอร์สกิลอื่นที่รองรับ ตรวจว่าโฮสต์ที่เลือกค้นพบตำแหน่งนั้นจริง
รูปแบบคำสั่งเรียกใช้ขึ้นอยู่กับโฮสต์ ไม่ใช่รูปแบบพกพา `SKILL.md`:
| แอปโฮสต์ | เริ่มหลักสูตร | เริ่ม Model Context Protocol (MCP) | เริ่ม Agent Skills | ทำแบบทดสอบประจำระยะ |
|---|---|---|---|---|
| Codex | `start-learning`, หรือเลือกจาก `/skills` | `learn-mcp`, หรือเลือกจาก `/skills` | `learn-agent-skills`, หรือเลือกจาก `/skills` | `check-understanding 13`, หรือเลือกจาก `/skills` |
| Claude Code | `/start-learning` | `/learn-mcp` | `/learn-agent-skills` | `/check-understanding 13` |
| โฮสต์อื่นที่รองรับ | `Use start-learning to begin the course.` | `Use learn-mcp to start the Model Context Protocol (MCP) path.` | `Use learn-agent-skills to start the Agent Skills Engineering path.` | `Use check-understanding to quiz me on Phase 13.` |
แบบทดสอบจัดระดับสิบข้อจะจับคู่ความรู้ที่มีอยู่กับระยะเริ่มต้น และบันทึกแผนเรียนส่วนตัวใน `LEARNING.md` จากนั้นสกิล `learn` จะสอนครั้งละหนึ่งบทเรียน: แนวคิด คณิตศาสตร์ โค้ด และแบบทดสอบ โดยดึงบทเรียนจากที่เก็บโค้ดนี้โดยตรง ส่วน `course-guide` จะพาไปยังบทเรียนที่ตรงกับเรื่องที่คุณติดขัด ใน Codex ให้เรียก `learn` และ `course-guide`; ใน Claude Code ให้ใช้ `/learn` และ `/course-guide`; สำหรับโฮสต์อื่นที่รองรับ ให้ขอใช้สกิลตามชื่อ
ต้องการเรียนเฉพาะ Model Context Protocol (MCP)? ใช้คำสั่ง MCP ของโฮสต์คุณ ระบบจะสร้าง `MCP-LEARNING.md` และติดตามเส้นทาง 17 บทเรียน ครอบคลุมคำขอแบบไร้สถานะ การรับส่งข้อมูล งานสองทิศทาง ความปลอดภัย ความน่าเชื่อถือ การกำกับดูแลรีจิสทรี และหลักฐานความสอดคล้องกับโพรโทคอล ลำดับและจุดตรวจที่แน่นอนอยู่ใน[แมนิเฟสต์ Model Context Protocol (MCP)](../../learning-paths/model-context-protocol.json)
ต้องการเรียนเฉพาะ Agent Skills? ใช้คำสั่ง Agent Skills ของโฮสต์คุณ ระบบจะสร้าง `AGENT-SKILLS-LEARNING.md` และติดตามห้าบทเรียนต่อเนื่อง: สัญญา การค้นพบ การเรียกใช้ ขอบเขตแซนด์บ็อกซ์ แล้วจึงประเมินก่อนเผยแพร่และทดสอบการย้ายข้ามโฮสต์จริง เริ่มบนเว็บได้ที่[เส้นทาง Agent Skills](https://aiengineeringfromscratch.com/lesson?path=phases/13-tools-and-protocols/22-skills-and-agent-sdks&learningPath=agent-skills)
ตัวติดตั้งจะแสดงโฮสต์ที่ตั้งค่าได้และถามตำแหน่งติดตั้ง หากยังไม่มี Node.js, `npx`, `python3`, โฮสต์ที่รองรับ หรือพื้นที่ที่เขียนได้ ให้ใช้เว็บไซต์หรืออ่าน `docs/en.md` เอง วิธีนี้สอนแนวคิดได้ แต่หลักฐานการค้นพบ การเรียกใช้ การรันสคริปต์ และการถอนการติดตั้งบนโฮสต์จริงต้องรอจนตรวจความพร้อมได้ อ่านบทเรียนที่ [aiengineeringfromscratch.com](https://aiengineeringfromscratch.com)
### สกิลสำหรับการเรียน
| สกิลการทำงาน | ทำอะไรได้ |
|---|---|
| [`start-learning`](../../skills/start-learning/SKILL.md) | เริ่มต้นครั้งเดียว: เหตุผลที่เรียน แบบทดสอบจัดระดับ และแผนส่วนตัวที่บันทึกใน `LEARNING.md` |
| [`learn`](../../skills/learn/SKILL.md) | วงจรผู้สอน: ทบทวนความจำก่อน เรียนบทถัดไปแบบโต้ตอบ แล้วทำแบบทดสอบ พร้อมบันทึกความคืบหน้าและรายการทบทวน |
| [`course-guide`](../../skills/course-guide/SKILL.md) | ตัวนำทางหัวข้อ ถามว่า “เรียน attention ที่ไหน” หรือ “ค่า loss เป็น NaN” → ได้บทเรียนที่ตรงพร้อมลิงก์ |
| [`learn-mcp`](../../skills/learn-mcp/SKILL.md) | ผู้สอนเฉพาะ Model Context Protocol (MCP) สร้าง `MCP-LEARNING.md` เดินตามแมนิเฟสต์ 17 บทเรียน และบันทึกหลักฐานการสื่อสาร ความปลอดภัย ความน่าเชื่อถือ และความสอดคล้อง |
| [`learn-agent-skills`](../../skills/learn-agent-skills/SKILL.md) | ผู้สอนเฉพาะ Agent Skills สร้าง `AGENT-SKILLS-LEARNING.md` สอนบทเรียน 22, 24, 25, 26 และ 27 พร้อมบันทึกหลักฐานจากโฮสต์จริง |
| [`claude-certification`](../../skills/claude-certification/SKILL.md) | ผู้สอนเตรียมใบรับรอง เลือก CCAO-F, CCDV-F, CCAR-F หรือ CCAR-P สอนแต่ละบท รันแล็บ ตรวจชิ้นงาน จัดแบบประเมินและข้อสอบจำลอง แล้วบันทึกความคืบหน้า |
| [`mcpa-certification`](../../skills/mcpa-certification/SKILL.md) | ผู้สอน MCPA ติดตามเส้นทาง `mcpa-f` 34 บทเรียนบนโพรโทคอล 2026-07-28 สอนแต่ละบท รันแล็บและตัวตรวจข้อความ จัดแบบประเมินพื้นฐานกับข้อสอบจำลองสามชุด แล้วบันทึกความคืบหน้า |
| [`find-your-level`](../../skills/find-your-level/SKILL.md) | แบบทดสอบจัดระดับสิบข้อ จับคู่ความรู้กับระยะเริ่มต้น และสร้างเส้นทางส่วนตัวพร้อมประมาณชั่วโมงเรียน |
| [`check-understanding `](../../skills/check-understanding/SKILL.md) | แบบทดสอบแปดข้อต่อระยะ พร้อมข้อเสนอแนะและบทเรียนที่ควรทบทวน ใช้รูปแบบ Codex, Claude Code หรือภาษาธรรมชาติจากตารางเรียกใช้ด้านบน |
### รันโค้ดบนเครื่อง
```bash
git clone https://github.com/rohitg00/ai-engineering-from-scratch.git
cd ai-engineering-from-scratch
python3 phases/00-setup-and-tooling/01-dev-environment/code/verify.py --route beginner
python3 phases/01-math-foundations/01-linear-algebra-intuition/code/vectors.py
```
การตรวจสอบก่อนเริ่มแยกสิ่งที่จำเป็นตอนนี้ออกจากเครื่องมือที่ต้องใช้ภายหลัง ทุกข้อบังคับที่ตรวจสอบไม่ผ่านจะแสดงสาเหตุที่พบและคำสั่งแก้ไข คำสั่ง `vectors.py` รันบทเรียนที่ไม่ต้องติดตั้งแพ็กเกจเพิ่มเติม และจบด้วยการแสดงว่าการคูณเมทริกซ์กับเวกเตอร์คือการทำงานภายในชั้นของโครงข่ายประสาท เก็บผลลัพธ์ในเทอร์มินัลนั้นไว้เป็นหลักฐานชิ้นแรก
ใช้วิธีเดียวกันกับทุกบทเรียน
### ใช้วิธีเดียวกันกับทุกบทเรียน
1. **อ่าน** `docs/en.md` และอธิบายแนวคิดหลักด้วยคำพูดของตัวเอง
2. **พิมพ์และสร้าง** โค้ดส่วนสำคัญ แทนที่จะมองบล็อกโค้ดเป็นเพียงของตกแต่ง
3. **รัน** คำสั่งของบทเรียนจากโฟลเดอร์รากของคลังโค้ด ซึ่งมี `README.md` และ `phases/`
4. **เก็บหลักฐาน**: คำสั่ง โฟลเดอร์ที่ทำงาน รหัสจบการทำงาน ผลลัพธ์ที่มีความหมาย และชิ้นงานที่คุณแก้ไขหรือสร้าง
5. **ไปต่อ** เมื่อคุณอธิบายผลลัพธ์และแก้ไขเล็กน้อยได้โดยไม่ต้องเดาเท่านั้น
คำสั่งในหน้าบทเรียนใช้พาธจากโฟลเดอร์รากของคลังโค้ด เว้นแต่บทเรียนจะระบุให้เปลี่ยนโฟลเดอร์อย่างชัดเจน หากบทเรียนมีหลายภาษา ให้รันโค้ดของภาษาที่คุณกำลังเรียน
## เลือกเส้นทางการเรียนรู้
คุณไม่จำเป็นต้องไล่ดูทั้ง 523 บทเรียนก่อนเริ่ม เลือกเป้าหมายหนึ่งอย่าง แต่ละลิงก์เปิดหลักสูตรเดียวกันบน GitHub หรือเว็บไซต์ และทั้งสองแบบใช้โค้ดบทเรียนเดียวกัน
| เป้าหมายของคุณ | เรียนบน GitHub | เรียนบนเว็บไซต์ |
|---|---|---|
| ฉันเพิ่งเริ่มและต้องการพื้นฐานที่ครบถ้วน | [ระยะ 0: การตั้งค่าและเครื่องมือ](../../phases/00-setup-and-tooling/) | [สภาพแวดล้อมการพัฒนา](https://aiengineeringfromscratch.com/lesson?path=phases/00-setup-and-tooling/01-dev-environment) |
| ฉันรู้ Python และต้องการพื้นฐานคณิตศาสตร์กับ ML | [ระยะ 1: พื้นฐานคณิตศาสตร์](../../phases/01-math-foundations/) | [ทำความเข้าใจพีชคณิตเชิงเส้น](https://aiengineeringfromscratch.com/lesson?path=phases/01-math-foundations/01-linear-algebra-intuition) |
| ฉันต้องการสร้างแอปพลิเคชัน LLM สำหรับใช้งานจริง | [ระยะ 11: วิศวกรรม LLM](../../phases/11-llm-engineering/) | [วิศวกรรมพรอมป์ต์](https://aiengineeringfromscratch.com/lesson?path=phases/11-llm-engineering/01-prompt-engineering) |
| ฉันต้องการสร้างเอเจนต์ | [ระยะ 14: วิศวกรรมเอเจนต์](../../phases/14-agent-engineering/) | [วงรอบการทำงานของเอเจนต์](https://aiengineeringfromscratch.com/lesson?path=phases/14-agent-engineering/01-the-agent-loop) |
| ฉันต้องการใช้เอเจนต์เขียนโค้ดกับคลังโค้ดจริง | [เส้นทางวิศวกรรมที่มีเอเจนต์ช่วย](../../learning-paths/using-coding-agents.json) | [วิศวกรรมที่มีเอเจนต์ช่วย](https://aiengineeringfromscratch.com/lesson?path=phases/14-agent-engineering/31-agent-workbench-why-models-fail&learningPath=using-coding-agents) |
| ฉันต้องการกำหนดสิ่งที่ควรสร้างก่อนลงมือพัฒนา | [เส้นทางการตัดสินใจด้านผลิตภัณฑ์และการส่งมอบ](../../learning-paths/shaping-the-build.json) | [การตัดสินใจด้านผลิตภัณฑ์และการส่งมอบ](https://aiengineeringfromscratch.com/lesson?path=phases/14-agent-engineering/47-outcomes-before-output&learningPath=shaping-the-build) |
ไม่แน่ใจว่าควรเริ่มตรงไหน? ใช้[ติวเตอร์วัดระดับ `start-learning`](../../skills/start-learning/SKILL.md) หรือ[คู่มือความรู้พื้นฐานที่ต้องมีบนเว็บไซต์](https://aiengineeringfromscratch.com/prereqs.html)
เปรียบเทียบสี่สาขาหลักและหกเส้นทางอาชีพใน[เส้นทางเรียนวิศวกรรม AI](https://aiengineeringfromscratch.com/learning-paths.html)
เส้นทางเฉพาะด้าน MCP และ Agent Skills
| เป้าหมายของคุณ | เรียนบน GitHub | เรียนบนเว็บไซต์ |
|---|---|---|
| ฉันต้องการสร้างด้วย Model Context Protocol (MCP) | [แนวทางเรียน Model Context Protocol (MCP)](../../phases/13-tools-and-protocols/README.md#model-context-protocol-mcp-path) | [เส้นทาง Model Context Protocol (MCP)](https://aiengineeringfromscratch.com/lesson?path=phases/13-tools-and-protocols/06-mcp-fundamentals&learningPath=model-context-protocol) |
| ฉันต้องการเขียนและเผยแพร่ Agent Skills | [แนวทางเรียน Agent Skills โดยเฉพาะ](../../phases/13-tools-and-protocols/README.md#agent-skills-fast-path) | [เส้นทาง Agent Skills](https://aiengineeringfromscratch.com/lesson?path=phases/13-tools-and-protocols/22-skills-and-agent-sdks&learningPath=agent-skills) |
ความรู้พื้นฐานและเวลาเรียน
### ความรู้และเครื่องมือพื้นฐานที่ต้องมี
- คุณเขียนโค้ดได้สักภาษา โดย Python จะช่วยได้มาก
- คุณต้องการเข้าใจว่า AI **ทำงานจริงอย่างไร** ไม่ใช่เพียงเรียก API
## ควรเริ่มที่ไหน
| พื้นฐานเดิม | เริ่มที่ | เวลาประมาณ |
|---|---|---|
| ใหม่กับการเขียนโปรแกรมและ AI | ระยะ 0: ตั้งค่า | ~306 ชั่วโมง |
| รู้ Python แต่ใหม่กับ ML | ระยะ 1: พื้นฐานคณิตศาสตร์ | ~270 ชั่วโมง |
| รู้ ML แต่ใหม่กับดีปเลิร์นนิง | ระยะ 3: แกนหลักดีปเลิร์นนิง | ~200 ชั่วโมง |
| รู้ดีปเลิร์นนิงและต้องการเรียน LLM กับเอเจนต์ | ระยะ 10: LLM จากศูนย์ | ~100 ชั่วโมง |
| วิศวกรอาวุโสที่ต้องการเฉพาะวิศวกรรมเอเจนต์ | ระยะ 14: วิศวกรรมเอเจนต์ | ~60 ชั่วโมง |
| ต้องการสร้างเฉพาะระบบ MCP สำหรับงานจริง | [เส้นทาง Model Context Protocol (MCP)](../../learning-paths/model-context-protocol.json) | ~23 ชั่วโมง 15 นาที |
| ต้องการสร้างเฉพาะ Agent Skills สำหรับงานจริง | [เส้นทางวิศวกรรม Agent Skills](../../learning-paths/agent-skills.json) | ~9.5 ชั่วโมง |
## โครงสร้างหลักสูตร
ทั้งยี่สิบระยะต่อยอดกัน คณิตศาสตร์เป็นฐาน ส่วนเอเจนต์และระบบใช้งานจริงเป็นชั้นบน ข้ามไปได้หากรู้ชั้นพื้นฐานอยู่แล้ว แต่อย่าข้ามแล้วสงสัยว่าทำไมส่วนข้างบนจึงพัง
```mermaid
%%{init: {'theme':'base','themeVariables':{'primaryColor':'#fafaf5','primaryTextColor':'#1a1a1a','primaryBorderColor':'#3553ff','lineColor':'#3553ff','fontFamily':'JetBrains Mono','fontSize':'12px'}}}%%
flowchart TB
P0["ระยะ 0: การตั้งค่าและเครื่องมือ"] --> P1["ระยะ 1: พื้นฐานคณิตศาสตร์"]
P1 --> P2["ระยะ 2: พื้นฐาน ML"]
P2 --> P3["ระยะ 3: แกนหลักดีปเลิร์นนิง"]
P3 --> P4["ระยะ 4: คอมพิวเตอร์วิทัศน์"]
P3 --> P5["ระยะ 5: NLP จากพื้นฐานสู่ขั้นสูง"]
P3 --> P6["ระยะ 6: เสียงพูดและออดิโอ"]
P3 --> P9["ระยะ 9: การเรียนรู้แบบเสริมกำลัง"]
P5 --> P7["ระยะ 7: เจาะลึก Transformers"]
P7 --> P8["ระยะ 8: AI กำเนิด"]
P7 --> P10["ระยะ 10: LLM จากศูนย์"]
P10 --> P11["ระยะ 11: วิศวกรรม LLM"]
P10 --> P12["ระยะ 12: AI หลายโมดัลลิตี"]
P11 --> P13["ระยะ 13: เครื่องมือและโพรโทคอล"]
P13 --> P14["ระยะ 14: วิศวกรรมเอเจนต์"]
P14 --> P15["ระยะ 15: ระบบอัตโนมัติ"]
P15 --> P16["ระยะ 16: หลายเอเจนต์และฝูง"]
P14 --> P17["ระยะ 17: โครงสร้างพื้นฐานและงานจริง"]
P15 --> P18["ระยะ 18: จริยธรรม ความปลอดภัย และการจัดแนว"]
P16 --> P19["ระยะ 19: โครงงานสรุป"]
P17 --> P19
P18 --> P19
```
## สารบัญ
ยี่สิบระยะ คลิกที่ระยะเพื่อขยายรายชื่อบทเรียน
### ระยะ 0: การตั้งค่าและเครื่องมือ `12 บทเรียน`
> เตรียมสภาพแวดล้อมให้พร้อมสำหรับสิ่งที่จะตามมา
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [สภาพแวดล้อมการพัฒนา](../../phases/00-setup-and-tooling/01-dev-environment/) | สร้าง | Python |
| 02 | [Git และการทำงานร่วมกัน](../../phases/00-setup-and-tooling/02-git-and-collaboration/) | เรียนรู้ | — |
| 03 | [ตั้งค่า GPU และคลาวด์](../../phases/00-setup-and-tooling/03-gpu-setup-and-cloud/) | สร้าง | Python |
| 04 | [API และคีย์](../../phases/00-setup-and-tooling/04-apis-and-keys/) | สร้าง | Python |
| 05 | [สมุดบันทึก Jupyter](../../phases/00-setup-and-tooling/05-jupyter-notebooks/) | สร้าง | Python |
| 06 | [สภาพแวดล้อม Python](../../phases/00-setup-and-tooling/06-python-environments/) | สร้าง | Shell |
| 07 | [Docker สำหรับ AI](../../phases/00-setup-and-tooling/07-docker-for-ai/) | สร้าง | Docker |
| 08 | [ตั้งค่าตัวแก้ไขโค้ด](../../phases/00-setup-and-tooling/08-editor-setup/) | สร้าง | — |
| 09 | [การจัดการข้อมูล](../../phases/00-setup-and-tooling/09-data-management/) | สร้าง | Python |
| 10 | [เทอร์มินัลและเชลล์](../../phases/00-setup-and-tooling/10-terminal-and-shell/) | เรียนรู้ | — |
| 11 | [Linux สำหรับ AI](../../phases/00-setup-and-tooling/11-linux-for-ai/) | เรียนรู้ | — |
| 12 | [การดีบักและวิเคราะห์ประสิทธิภาพ](../../phases/00-setup-and-tooling/12-debugging-and-profiling/) | สร้าง | Python |
ระยะ 1: พื้นฐานคณิตศาสตร์ 22 บทเรียน เข้าใจแนวคิดเบื้องหลังทุกอัลกอริทึม AI ผ่านโค้ด
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [ทำความเข้าใจพีชคณิตเชิงเส้น](../../phases/01-math-foundations/01-linear-algebra-intuition/) | เรียนรู้ | Python, Julia |
| 02 | [เวกเตอร์ เมทริกซ์ และการดำเนินการ](../../phases/01-math-foundations/02-vectors-matrices-operations/) | สร้าง | Python, Julia |
| 03 | [การแปลงเมทริกซ์และค่าไอเกน](../../phases/01-math-foundations/03-matrix-transformations/) | สร้าง | Python, Julia |
| 04 | [แคลคูลัสสำหรับ ML: อนุพันธ์และเกรเดียนต์](../../phases/01-math-foundations/04-calculus-for-ml/) | เรียนรู้ | Python |
| 05 | [กฎลูกโซ่และการหาอนุพันธ์อัตโนมัติ](../../phases/01-math-foundations/05-chain-rule-and-autodiff/) | สร้าง | Python |
| 06 | [ความน่าจะเป็นและการแจกแจง](../../phases/01-math-foundations/06-probability-and-distributions/) | เรียนรู้ | Python |
| 07 | [ทฤษฎีบทเบย์และการคิดเชิงสถิติ](../../phases/01-math-foundations/07-bayes-theorem/) | สร้าง | Python |
| 08 | [การหาค่าเหมาะที่สุด: กลุ่มวิธีเกรเดียนต์ดิเซนต์](../../phases/01-math-foundations/08-optimization/) | สร้าง | Python |
| 09 | [ทฤษฎีสารสนเทศ: เอนโทรปีและ KL divergence](../../phases/01-math-foundations/09-information-theory/) | เรียนรู้ | Python |
| 10 | [การลดมิติ: PCA, t-SNE, UMAP](../../phases/01-math-foundations/10-dimensionality-reduction/) | สร้าง | Python |
| 11 | [การแยกค่าเอกฐาน](../../phases/01-math-foundations/11-singular-value-decomposition/) | สร้าง | Python, Julia |
| 12 | [การดำเนินการกับเทนเซอร์](../../phases/01-math-foundations/12-tensor-operations/) | สร้าง | Python |
| 13 | [เสถียรภาพเชิงตัวเลข](../../phases/01-math-foundations/13-numerical-stability/) | สร้าง | Python |
| 14 | [นอร์มและระยะทาง](../../phases/01-math-foundations/14-norms-and-distances/) | สร้าง | Python |
| 15 | [สถิติสำหรับ ML](../../phases/01-math-foundations/15-statistics-for-ml/) | สร้าง | Python |
| 16 | [วิธีสุ่มตัวอย่าง](../../phases/01-math-foundations/16-sampling-methods/) | สร้าง | Python |
| 17 | [ระบบสมการเชิงเส้น](../../phases/01-math-foundations/17-linear-systems/) | สร้าง | Python |
| 18 | [การหาค่าเหมาะที่สุดแบบคอนเวกซ์](../../phases/01-math-foundations/18-convex-optimization/) | สร้าง | Python |
| 19 | [จำนวนเชิงซ้อนสำหรับ AI](../../phases/01-math-foundations/19-complex-numbers/) | เรียนรู้ | Python |
| 20 | [การแปลงฟูเรียร์](../../phases/01-math-foundations/20-fourier-transform/) | สร้าง | Python |
| 21 | [ทฤษฎีกราฟสำหรับ ML](../../phases/01-math-foundations/21-graph-theory/) | สร้าง | Python |
| 22 | [กระบวนการสุ่ม](../../phases/01-math-foundations/22-stochastic-processes/) | เรียนรู้ | Python |
ระยะ 2: พื้นฐาน ML 18 บทเรียน ML แบบดั้งเดิมยังเป็นแกนหลักของ AI ส่วนใหญ่ที่ใช้งานจริง
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [แมชชีนเลิร์นนิงคืออะไร](../../phases/02-ml-fundamentals/01-what-is-machine-learning/) | เรียนรู้ | Python |
| 02 | [การถดถอยเชิงเส้นจากศูนย์](../../phases/02-ml-fundamentals/02-linear-regression/) | สร้าง | Python |
| 03 | [การถดถอยโลจิสติกและการจำแนกประเภท](../../phases/02-ml-fundamentals/03-logistic-regression/) | สร้าง | Python |
| 04 | [ต้นไม้ตัดสินใจและแรนดอมฟอเรสต์](../../phases/02-ml-fundamentals/04-decision-trees/) | สร้าง | Python |
| 05 | [ซัพพอร์ตเวกเตอร์แมชชีน](../../phases/02-ml-fundamentals/05-support-vector-machines/) | สร้าง | Python |
| 06 | [KNN และมาตรวัดระยะทาง](../../phases/02-ml-fundamentals/06-knn-and-distances/) | สร้าง | Python |
| 07 | [การเรียนรู้แบบไม่มีผู้สอน: K-Means, DBSCAN](../../phases/02-ml-fundamentals/07-unsupervised-learning/) | สร้าง | Python |
| 08 | [การสร้างและเลือกคุณลักษณะ](../../phases/02-ml-fundamentals/08-feature-engineering/) | สร้าง | Python |
| 09 | [การประเมินโมเดล: ตัวชี้วัดและการตรวจสอบไขว้](../../phases/02-ml-fundamentals/09-model-evaluation/) | สร้าง | Python |
| 10 | [ไบแอส ความแปรปรวน และเส้นโค้งการเรียนรู้](../../phases/02-ml-fundamentals/10-bias-variance/) | เรียนรู้ | Python |
| 11 | [วิธีเอนเซมเบิล: boosting, bagging, stacking](../../phases/02-ml-fundamentals/11-ensemble-methods/) | สร้าง | Python |
| 12 | [การปรับไฮเปอร์พารามิเตอร์](../../phases/02-ml-fundamentals/12-hyperparameter-tuning/) | สร้าง | Python |
| 13 | [ไปป์ไลน์ ML และการติดตามการทดลอง](../../phases/02-ml-fundamentals/13-ml-pipelines/) | สร้าง | Python |
| 14 | [นาอีฟเบย์](../../phases/02-ml-fundamentals/14-naive-bayes/) | สร้าง | Python |
| 15 | [พื้นฐานอนุกรมเวลา](../../phases/02-ml-fundamentals/15-time-series/) | สร้าง | Python |
| 16 | [การตรวจจับความผิดปกติ](../../phases/02-ml-fundamentals/16-anomaly-detection/) | สร้าง | Python |
| 17 | [การจัดการข้อมูลไม่สมดุล](../../phases/02-ml-fundamentals/17-imbalanced-data/) | สร้าง | Python |
| 18 | [การเลือกคุณลักษณะ](../../phases/02-ml-fundamentals/18-feature-selection/) | สร้าง | Python |
ระยะ 3: แกนหลักดีปเลิร์นนิง 13 บทเรียน โครงข่ายประสาทจากหลักการพื้นฐาน ยังไม่ใช้เฟรมเวิร์กจนกว่าจะสร้างเอง
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [เพอร์เซปตรอน: จุดเริ่มต้นของทุกอย่าง](../../phases/03-deep-learning-core/01-the-perceptron/) | สร้าง | Python |
| 02 | [โครงข่ายหลายชั้นและการคำนวณไปข้างหน้า](../../phases/03-deep-learning-core/02-multi-layer-networks/) | สร้าง | Python |
| 03 | [แบ็กโพรพาเกชันจากศูนย์](../../phases/03-deep-learning-core/03-backpropagation/) | สร้าง | Python |
| 04 | [ฟังก์ชันกระตุ้น: ReLU, Sigmoid, GELU และเหตุผลที่ใช้](../../phases/03-deep-learning-core/04-activation-functions/) | สร้าง | Python |
| 05 | [ฟังก์ชันความสูญเสีย: MSE, cross-entropy และแบบเปรียบต่าง](../../phases/03-deep-learning-core/05-loss-functions/) | สร้าง | Python |
| 06 | [ตัวปรับค่า: SGD, Momentum, Adam, AdamW](../../phases/03-deep-learning-core/06-optimizers/) | สร้าง | Python |
| 07 | [การลดการฟิตเกิน: Dropout, Weight Decay, BatchNorm](../../phases/03-deep-learning-core/07-regularization/) | สร้าง | Python |
| 08 | [การเริ่มค่าน้ำหนักและเสถียรภาพการฝึก](../../phases/03-deep-learning-core/08-weight-initialization/) | สร้าง | Python |
| 09 | [ตารางอัตราการเรียนรู้และช่วงวอร์มอัป](../../phases/03-deep-learning-core/09-learning-rate-schedules/) | สร้าง | Python |
| 10 | [สร้างมินิเฟรมเวิร์กของคุณเอง](../../phases/03-deep-learning-core/10-mini-framework/) | สร้าง | Python |
| 11 | [แนะนำ PyTorch](../../phases/03-deep-learning-core/11-intro-to-pytorch/) | สร้าง | Python |
| 12 | [แนะนำ JAX](../../phases/03-deep-learning-core/12-intro-to-jax/) | สร้าง | Python |
| 13 | [ดีบักโครงข่ายประสาท](../../phases/03-deep-learning-core/13-debugging-neural-networks/) | สร้าง | Python |
ระยะ 4: คอมพิวเตอร์วิทัศน์ 28 บทเรียน จากพิกเซลสู่ความเข้าใจ: ภาพ วิดีโอ 3D โมเดล VLM และโมเดลโลก
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [พื้นฐานภาพ: พิกเซล ช่องสัญญาณ และปริภูมิสี](../../phases/04-computer-vision/01-image-fundamentals/) | เรียนรู้ | Python |
| 02 | [คอนโวลูชันจากศูนย์](../../phases/04-computer-vision/02-convolutions-from-scratch/) | สร้าง | Python |
| 03 | [CNN: จาก LeNet ถึง ResNet](../../phases/04-computer-vision/03-cnns-lenet-to-resnet/) | สร้าง | Python |
| 04 | [การจำแนกภาพ](../../phases/04-computer-vision/04-image-classification/) | สร้าง | Python |
| 05 | [การถ่ายโอนการเรียนรู้และไฟน์จูน](../../phases/04-computer-vision/05-transfer-learning/) | สร้าง | Python |
| 06 | [ตรวจจับวัตถุ: YOLO จากศูนย์](../../phases/04-computer-vision/06-object-detection-yolo/) | สร้าง | Python |
| 07 | [แบ่งส่วนภาพเชิงความหมาย: U-Net](../../phases/04-computer-vision/07-semantic-segmentation-unet/) | สร้าง | Python |
| 08 | [แบ่งส่วนภาพระดับอินสแตนซ์: Mask R-CNN](../../phases/04-computer-vision/08-instance-segmentation-mask-rcnn/) | สร้าง | Python |
| 09 | [สร้างภาพด้วย GAN](../../phases/04-computer-vision/09-image-generation-gans/) | สร้าง | Python |
| 10 | [สร้างภาพด้วยโมเดลดิฟฟิวชัน](../../phases/04-computer-vision/10-image-generation-diffusion/) | สร้าง | Python |
| 11 | [Stable Diffusion: สถาปัตยกรรมและไฟน์จูน](../../phases/04-computer-vision/11-stable-diffusion/) | สร้าง | Python |
| 12 | [ทำความเข้าใจวิดีโอด้วยการจำลองตามเวลา](../../phases/04-computer-vision/12-video-understanding/) | สร้าง | Python |
| 13 | [การมองเห็นสามมิติ: พอยต์คลาวด์และ NeRF](../../phases/04-computer-vision/13-3d-vision-nerf/) | สร้าง | Python |
| 14 | [Vision Transformers (ViT) สำหรับงานภาพ](../../phases/04-computer-vision/14-vision-transformers/) | สร้าง | Python |
| 15 | [การมองเห็นแบบเรียลไทม์: ปรับใช้บนอุปกรณ์ปลายทาง](../../phases/04-computer-vision/15-real-time-edge/) | สร้าง | Python |
| 16 | [สร้างไปป์ไลน์การมองเห็นที่สมบูรณ์](../../phases/04-computer-vision/16-vision-pipeline-capstone/) | สร้าง | Python |
| 17 | [การมองเห็นแบบเรียนรู้ด้วยตนเอง: SimCLR, DINO, MAE](../../phases/04-computer-vision/17-self-supervised-vision/) | สร้าง | Python |
| 18 | [การมองเห็นด้วยคำศัพท์แบบเปิด: CLIP](../../phases/04-computer-vision/18-open-vocab-clip/) | สร้าง | Python |
| 19 | [OCR และการเข้าใจเอกสาร](../../phases/04-computer-vision/19-ocr-document-understanding/) | สร้าง | Python |
| 20 | [การค้นคืนภาพและการเรียนรู้ระยะทาง](../../phases/04-computer-vision/20-image-retrieval-metric/) | สร้าง | Python |
| 21 | [ตรวจจับจุดสำคัญและประมาณท่าทาง](../../phases/04-computer-vision/21-keypoint-pose/) | สร้าง | Python |
| 22 | [3D Gaussian Splatting จากศูนย์](../../phases/04-computer-vision/22-3d-gaussian-splatting/) | สร้าง | Python |
| 23 | [Diffusion Transformers และ rectified flow](../../phases/04-computer-vision/23-diffusion-transformers-rectified-flow/) | สร้าง | Python |
| 24 | [SAM 3 และการแบ่งส่วนภาพด้วยคำศัพท์แบบเปิด](../../phases/04-computer-vision/24-sam3-open-vocab-segmentation/) | สร้าง | Python |
| 25 | [โมเดลภาพและภาษา (ViT-MLP-LLM)](../../phases/04-computer-vision/25-vision-language-models/) | สร้าง | Python |
| 26 | [ประมาณความลึกและเรขาคณิตจากภาพเดียว](../../phases/04-computer-vision/26-monocular-depth/) | สร้าง | Python |
| 27 | [ติดตามหลายวัตถุและหน่วยความจำวิดีโอ](../../phases/04-computer-vision/27-multi-object-tracking/) | สร้าง | Python |
| 28 | [โมเดลโลกและดิฟฟิวชันวิดีโอ](../../phases/04-computer-vision/28-world-models-video-diffusion/) | สร้าง | Python |
ระยะ 5: NLP จากพื้นฐานสู่ขั้นสูง 29 บทเรียน ภาษาคืออินเทอร์เฟซสู่อัจฉริยะ
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [ประมวลผลข้อความ: ตัดโทเคน ตัดรากศัพท์ และหารูปคำหลัก](../../phases/05-nlp-foundations-to-advanced/01-text-processing/) | สร้าง | Python |
| 02 | [ถุงคำ TF-IDF และการแทนข้อความ](../../phases/05-nlp-foundations-to-advanced/02-bag-of-words-tfidf/) | สร้าง | Python |
| 03 | [เวกเตอร์คำ: Word2Vec จากศูนย์](../../phases/05-nlp-foundations-to-advanced/03-word-embeddings-word2vec/) | สร้าง | Python |
| 04 | [GloVe, FastText และเวกเตอร์คำย่อย](../../phases/05-nlp-foundations-to-advanced/04-glove-fasttext-subword/) | สร้าง | Python |
| 05 | [การวิเคราะห์ความรู้สึก](../../phases/05-nlp-foundations-to-advanced/05-sentiment-analysis/) | สร้าง | Python |
| 06 | [การรู้จำชื่อเฉพาะ (NER)](../../phases/05-nlp-foundations-to-advanced/06-named-entity-recognition/) | สร้าง | Python |
| 07 | [การติดแท็กชนิดคำและวิเคราะห์ไวยากรณ์](../../phases/05-nlp-foundations-to-advanced/07-pos-tagging-parsing/) | สร้าง | Python |
| 08 | [จำแนกข้อความด้วย CNN และ RNN](../../phases/05-nlp-foundations-to-advanced/08-cnns-rnns-for-text/) | สร้าง | Python |
| 09 | [โมเดลลำดับสู่ลำดับ](../../phases/05-nlp-foundations-to-advanced/09-sequence-to-sequence/) | สร้าง | Python |
| 10 | [กลไก attention: จุดเปลี่ยนสำคัญ](../../phases/05-nlp-foundations-to-advanced/10-attention-mechanism/) | สร้าง | Python |
| 11 | [การแปลด้วยเครื่อง](../../phases/05-nlp-foundations-to-advanced/11-machine-translation/) | สร้าง | Python |
| 12 | [การสรุปข้อความ](../../phases/05-nlp-foundations-to-advanced/12-text-summarization/) | สร้าง | Python |
| 13 | [ระบบถามตอบ](../../phases/05-nlp-foundations-to-advanced/13-question-answering/) | สร้าง | Python |
| 14 | [การค้นคืนข้อมูลและการค้นหา](../../phases/05-nlp-foundations-to-advanced/14-information-retrieval-search/) | สร้าง | Python |
| 15 | [การสร้างโมเดลหัวข้อ: LDA, BERTopic](../../phases/05-nlp-foundations-to-advanced/15-topic-modeling/) | สร้าง | Python |
| 16 | [การสร้างข้อความ](../../phases/05-nlp-foundations-to-advanced/16-text-generation-pre-transformer/) | สร้าง | Python |
| 17 | [แชตบอต: จากกฎสู่โครงข่ายประสาท](../../phases/05-nlp-foundations-to-advanced/17-chatbots-rule-to-neural/) | สร้าง | Python |
| 18 | [NLP หลายภาษา](../../phases/05-nlp-foundations-to-advanced/18-multilingual-nlp/) | สร้าง | Python |
| 19 | [ตัดโทเคนคำย่อย: BPE, WordPiece, Unigram, SentencePiece](../../phases/05-nlp-foundations-to-advanced/19-subword-tokenization/) | เรียนรู้ | Python |
| 20 | [เอาต์พุตมีโครงสร้างและการถอดรหัสแบบจำกัด](../../phases/05-nlp-foundations-to-advanced/20-structured-outputs-constrained-decoding/) | สร้าง | Python |
| 21 | [NLI และการอนุมานความหมายจากข้อความ](../../phases/05-nlp-foundations-to-advanced/21-nli-textual-entailment/) | เรียนรู้ | Python |
| 22 | [เจาะลึกโมเดลเอ็มเบดดิง](../../phases/05-nlp-foundations-to-advanced/22-embedding-models-deep-dive/) | เรียนรู้ | Python |
| 23 | [กลยุทธ์แบ่งข้อความเป็นส่วนสำหรับ RAG](../../phases/05-nlp-foundations-to-advanced/23-chunking-strategies-rag/) | สร้าง | Python |
| 24 | [การระบุคำที่อ้างถึงสิ่งเดียวกัน](../../phases/05-nlp-foundations-to-advanced/24-coreference-resolution/) | เรียนรู้ | Python |
| 25 | [การเชื่อมโยงเอนทิตีและแก้ความกำกวม](../../phases/05-nlp-foundations-to-advanced/25-entity-linking/) | สร้าง | Python |
| 26 | [ดึงความสัมพันธ์และสร้างกราฟความรู้](../../phases/05-nlp-foundations-to-advanced/26-relation-extraction-kg/) | สร้าง | Python |
| 27 | [ประเมิน LLM: RAGAS, DeepEval, G-Eval](../../phases/05-nlp-foundations-to-advanced/27-llm-evaluation-frameworks/) | สร้าง | Python |
| 28 | [ประเมินบริบทยาว: NIAH, RULER, LongBench, MRCR](../../phases/05-nlp-foundations-to-advanced/28-long-context-evaluation/) | เรียนรู้ | Python |
| 29 | [ติดตามสถานะบทสนทนา](../../phases/05-nlp-foundations-to-advanced/29-dialogue-state-tracking/) | สร้าง | Python |
ระยะ 6: เสียงพูดและออดิโอ 17 บทเรียน ฟัง เข้าใจ และพูด
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [พื้นฐานเสียง: รูปคลื่น การสุ่มตัวอย่าง และ FFT](../../phases/06-speech-and-audio/01-audio-fundamentals) | เรียนรู้ | Python |
| 02 | [สเปกโตรแกรม สเกลเมล และคุณลักษณะเสียง](../../phases/06-speech-and-audio/02-spectrograms-mel-features) | สร้าง | Python |
| 03 | [การจำแนกเสียง](../../phases/06-speech-and-audio/03-audio-classification) | สร้าง | Python |
| 04 | [การรู้จำเสียงพูด (ASR)](../../phases/06-speech-and-audio/04-speech-recognition-asr) | สร้าง | Python |
| 05 | [Whisper: สถาปัตยกรรมและไฟน์จูน](../../phases/06-speech-and-audio/05-whisper-architecture-finetuning) | สร้าง | Python |
| 06 | [การรู้จำและยืนยันผู้พูด](../../phases/06-speech-and-audio/06-speaker-recognition-verification) | สร้าง | Python |
| 07 | [แปลงข้อความเป็นเสียงพูด (TTS)](../../phases/06-speech-and-audio/07-text-to-speech) | สร้าง | Python |
| 08 | [การโคลนและแปลงเสียง](../../phases/06-speech-and-audio/08-voice-cloning-conversion) | สร้าง | Python |
| 09 | [การสร้างดนตรี](../../phases/06-speech-and-audio/09-music-generation) | สร้าง | Python |
| 10 | [โมเดลเสียงและภาษา](../../phases/06-speech-and-audio/10-audio-language-models) | สร้าง | Python |
| 11 | [ประมวลผลเสียงแบบเรียลไทม์](../../phases/06-speech-and-audio/11-real-time-audio-processing) | สร้าง | Python |
| 12 | [สร้างไปป์ไลน์ผู้ช่วยเสียง](../../phases/06-speech-and-audio/12-voice-assistant-pipeline) | สร้าง | Python |
| 13 | [โคเดกเสียงแบบนิวรัล: EnCodec, SNAC, Mimi, DAC](../../phases/06-speech-and-audio/13-neural-audio-codecs) | เรียนรู้ | Python |
| 14 | [ตรวจจับช่วงพูดและการผลัดกันพูด](../../phases/06-speech-and-audio/14-voice-activity-detection-turn-taking) | สร้าง | Python |
| 15 | [สตรีมเสียงพูดสู่เสียงพูด: Moshi, Hibiki](../../phases/06-speech-and-audio/15-streaming-speech-to-speech-moshi-hibiki) | เรียนรู้ | Python |
| 16 | [ป้องกันเสียงปลอมและใส่ลายน้ำเสียง](../../phases/06-speech-and-audio/16-anti-spoofing-audio-watermarking) | สร้าง | Python |
| 17 | [ประเมินเสียง: WER, MOS, MMAU และตารางจัดอันดับ](../../phases/06-speech-and-audio/17-audio-evaluation-metrics) | เรียนรู้ | Python |
ระยะ 7: เจาะลึก Transformers 16 บทเรียน สถาปัตยกรรมที่เปลี่ยนทุกอย่าง
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [ทำไมต้อง Transformer: ปัญหาของ RNN](../../phases/07-transformers-deep-dive/01-why-transformers/) | เรียนรู้ | Python |
| 02 | [Self-attention จากศูนย์](../../phases/07-transformers-deep-dive/02-self-attention-from-scratch/) | สร้าง | Python |
| 03 | [Attention หลายหัว](../../phases/07-transformers-deep-dive/03-multi-head-attention/) | สร้าง | Python |
| 04 | [การเข้ารหัสตำแหน่ง: sinusoidal, RoPE, ALiBi](../../phases/07-transformers-deep-dive/04-positional-encoding/) | สร้าง | Python |
| 05 | [Transformer แบบเต็ม: ตัวเข้ารหัสและตัวถอดรหัส](../../phases/07-transformers-deep-dive/05-full-transformer/) | สร้าง | Python |
| 06 | [BERT: โมเดลภาษาแบบปิดบังคำ](../../phases/07-transformers-deep-dive/06-bert-masked-language-modeling/) | สร้าง | Python |
| 07 | [GPT: โมเดลภาษาแบบเชิงเหตุ](../../phases/07-transformers-deep-dive/07-gpt-causal-language-modeling/) | สร้าง | Python |
| 08 | [T5, BART: โมเดลเข้ารหัส-ถอดรหัส](../../phases/07-transformers-deep-dive/08-t5-bart-encoder-decoder/) | เรียนรู้ | Python |
| 09 | [Vision Transformers (ViT) สำหรับงานภาพ](../../phases/07-transformers-deep-dive/09-vision-transformers/) | สร้าง | Python |
| 10 | [Transformer สำหรับเสียง: สถาปัตยกรรม Whisper](../../phases/07-transformers-deep-dive/10-audio-transformers-whisper/) | เรียนรู้ | Python |
| 11 | [ส่วนผสมของผู้เชี่ยวชาญ (MoE)](../../phases/07-transformers-deep-dive/11-mixture-of-experts/) | สร้าง | Python |
| 12 | [แคช KV, Flash Attention และการเพิ่มประสิทธิภาพการอนุมาน](../../phases/07-transformers-deep-dive/12-kv-cache-flash-attention/) | สร้าง | Python |
| 13 | [กฎการขยายขนาด](../../phases/07-transformers-deep-dive/13-scaling-laws/) | เรียนรู้ | Python |
| 14 | [สร้าง Transformer จากศูนย์](../../phases/07-transformers-deep-dive/14-build-a-transformer-capstone/) | สร้าง | Python |
| 15 | [รูปแบบ attention: หน้าต่างเลื่อน แบบเบาบาง และแบบผลต่าง](../../phases/07-transformers-deep-dive/15-attention-variants/) | สร้าง | Python |
| 16 | [ถอดรหัสเชิงคาดการณ์: ร่าง ตรวจสอบ ทำซ้ำ](../../phases/07-transformers-deep-dive/16-speculative-decoding/) | สร้าง | Python |
ระยะ 8: AI กำเนิด 15 บทเรียน สร้างภาพ วิดีโอ เสียง 3D และอื่น ๆ
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [โมเดลกำเนิด: การจัดประเภทและประวัติ](../../phases/08-generative-ai/01-generative-models-taxonomy-history/) | เรียนรู้ | Python |
| 02 | [ออโตเอนโคเดอร์และ VAE](../../phases/08-generative-ai/02-autoencoders-vae/) | สร้าง | Python |
| 03 | [GAN: ตัวสร้างกับตัวแยกแยะ](../../phases/08-generative-ai/03-gans-generator-discriminator/) | สร้าง | Python |
| 04 | [GAN แบบมีเงื่อนไขและ Pix2Pix](../../phases/08-generative-ai/04-conditional-gans-pix2pix/) | สร้าง | Python |
| 05 | [โมเดล StyleGAN](../../phases/08-generative-ai/05-stylegan/) | สร้าง | Python |
| 06 | [โมเดลดิฟฟิวชัน: DDPM จากศูนย์](../../phases/08-generative-ai/06-diffusion-ddpm-from-scratch/) | สร้าง | Python |
| 07 | [Latent diffusion และ Stable Diffusion](../../phases/08-generative-ai/07-latent-diffusion-stable-diffusion/) | สร้าง | Python |
| 08 | [ControlNet, LoRA และการกำหนดเงื่อนไข](../../phases/08-generative-ai/08-controlnet-lora-conditioning/) | สร้าง | Python |
| 09 | [เติมภาพ ขยายภาพ และแก้ไขภาพ](../../phases/08-generative-ai/09-inpainting-outpainting-editing/) | สร้าง | Python |
| 10 | [การสร้างวิดีโอ](../../phases/08-generative-ai/10-video-generation/) | สร้าง | Python |
| 11 | [การสร้างเสียง](../../phases/08-generative-ai/11-audio-generation/) | สร้าง | Python |
| 12 | [การสร้างสามมิติ](../../phases/08-generative-ai/12-3d-generation/) | สร้าง | Python |
| 13 | [Flow matching และ rectified flows](../../phases/08-generative-ai/13-flow-matching-rectified-flows/) | สร้าง | Python |
| 14 | [การประเมิน: FID และคะแนน CLIP](../../phases/08-generative-ai/14-evaluation-fid-clip-score/) | สร้าง | Python |
| 19 | [การสร้างโมเดลภาพแบบออโตรีเกรสซีฟ (VAR): ทำนายสเกลถัดไป](../../phases/08-generative-ai/19-visual-autoregressive-var/) | สร้าง | Python |
ระยะ 9: การเรียนรู้แบบเสริมกำลัง 12 บทเรียน รากฐานของ RLHF และ AI เล่นเกม
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [MDP สถานะ การกระทำ และรางวัล](../../phases/09-reinforcement-learning/01-mdps-states-actions-rewards/) | เรียนรู้ | Python |
| 02 | [การเขียนโปรแกรมแบบพลวัต](../../phases/09-reinforcement-learning/02-dynamic-programming/) | สร้าง | Python |
| 03 | [วิธีมอนติคาร์โล](../../phases/09-reinforcement-learning/03-monte-carlo-methods/) | สร้าง | Python |
| 04 | [การเรียนรู้ Q และ SARSA](../../phases/09-reinforcement-learning/04-q-learning-sarsa/) | สร้าง | Python |
| 05 | [โครงข่าย Q เชิงลึก (DQN)](../../phases/09-reinforcement-learning/05-dqn/) | สร้าง | Python |
| 06 | [เกรเดียนต์นโยบาย: REINFORCE](../../phases/09-reinforcement-learning/06-policy-gradients-reinforce/) | สร้าง | Python |
| 07 | [Actor-critic: A2C, A3C](../../phases/09-reinforcement-learning/07-actor-critic-a2c-a3c/) | สร้าง | Python |
| 08 | [อัลกอริทึม PPO](../../phases/09-reinforcement-learning/08-ppo/) | สร้าง | Python |
| 09 | [การสร้างโมเดลรางวัลและ RLHF](../../phases/09-reinforcement-learning/09-reward-modeling-rlhf/) | สร้าง | Python |
| 10 | [การเรียนรู้แบบเสริมกำลังหลายเอเจนต์](../../phases/09-reinforcement-learning/10-multi-agent-rl/) | สร้าง | Python |
| 11 | [ถ่ายโอนจากการจำลองสู่โลกจริง](../../phases/09-reinforcement-learning/11-sim-to-real-transfer/) | สร้าง | Python |
| 12 | [การเรียนรู้แบบเสริมกำลังสำหรับเกม](../../phases/09-reinforcement-learning/12-rl-for-games/) | สร้าง | Python |
ระยะ 10: LLM จากศูนย์ 24 บทเรียน สร้าง ฝึก และเข้าใจโมเดลภาษาขนาดใหญ่
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [ตัวตัดโทเคน: BPE, WordPiece, SentencePiece](../../phases/10-llms-from-scratch/01-tokenizers/) | สร้าง | Python, Rust |
| 02 | [สร้างตัวตัดโทเคนจากศูนย์](../../phases/10-llms-from-scratch/02-building-a-tokenizer/) | สร้าง | Python |
| 03 | [ไปป์ไลน์ข้อมูลสำหรับพรีเทรน](../../phases/10-llms-from-scratch/03-data-pipelines/) | สร้าง | Python |
| 04 | [พรีเทรน Mini GPT (124M)](../../phases/10-llms-from-scratch/04-pre-training-mini-gpt/) | สร้าง | Python |
| 05 | [การฝึกแบบกระจาย FSDP และ DeepSpeed](../../phases/10-llms-from-scratch/05-scaling-distributed/) | สร้าง | Python |
| 06 | [ปรับโมเดลตามคำสั่ง: SFT](../../phases/10-llms-from-scratch/06-instruction-tuning-sft/) | สร้าง | Python |
| 07 | [RLHF: โมเดลรางวัลและ PPO](../../phases/10-llms-from-scratch/07-rlhf/) | สร้าง | Python |
| 08 | [DPO: การปรับตามความชอบโดยตรง](../../phases/10-llms-from-scratch/08-dpo/) | สร้าง | Python |
| 09 | [Constitutional AI และการปรับปรุงตนเอง](../../phases/10-llms-from-scratch/09-constitutional-ai-self-improvement/) | สร้าง | Python |
| 10 | [การประเมิน: เกณฑ์เปรียบเทียบและชุดทดสอบ](../../phases/10-llms-from-scratch/10-evaluation/) | สร้าง | Python |
| 11 | [ควอนไทเซชัน: INT8, GPTQ, AWQ, GGUF](../../phases/10-llms-from-scratch/11-quantization/) | สร้าง | Python |
| 12 | [การเพิ่มประสิทธิภาพการอนุมาน](../../phases/10-llms-from-scratch/12-inference-optimization/) | สร้าง | Python |
| 13 | [สร้างไปป์ไลน์ LLM ที่สมบูรณ์](../../phases/10-llms-from-scratch/13-building-complete-llm-pipeline/) | สร้าง | Python |
| 14 | [โมเดลเปิด: สำรวจสถาปัตยกรรม](../../phases/10-llms-from-scratch/14-open-models-architecture-walkthroughs/) | เรียนรู้ | Python |
| 15 | [การถอดรหัสเชิงคาดการณ์และ EAGLE-3](../../phases/10-llms-from-scratch/15-speculative-decoding-eagle3/) | สร้าง | Python |
| 16 | [Attention แบบผลต่าง (V2)](../../phases/10-llms-from-scratch/16-differential-attention-v2/) | สร้าง | Python |
| 17 | [Native Sparse Attention (DeepSeek NSA)](../../phases/10-llms-from-scratch/17-native-sparse-attention/) | สร้าง | Python |
| 18 | [การทำนายหลายโทเคน (MTP)](../../phases/10-llms-from-scratch/18-multi-token-prediction/) | สร้าง | Python |
| 19 | [การทำงานขนานแบบ DualPipe](../../phases/10-llms-from-scratch/19-dualpipe-parallelism/) | เรียนรู้ | Python |
| 20 | [สำรวจสถาปัตยกรรม DeepSeek-V3](../../phases/10-llms-from-scratch/20-deepseek-v3-walkthrough/) | เรียนรู้ | Python |
| 21 | [Jamba: ผสม SSM กับ Transformer](../../phases/10-llms-from-scratch/21-jamba-hybrid-ssm-transformer/) | เรียนรู้ | Python |
| 22 | [การอนุมานแบบอะซิงโครนัสและ Hogwild!](../../phases/10-llms-from-scratch/22-async-hogwild-inference/) | สร้าง | Python |
| 25 | [การถอดรหัสเชิงคาดการณ์และ EAGLE](../../phases/10-llms-from-scratch/25-speculative-decoding/) | สร้าง | Python |
| 34 | [เช็กพอยต์เกรเดียนต์และคำนวณแอกทิเวชันใหม่](../../phases/10-llms-from-scratch/34-gradient-checkpointing/) | สร้าง | Python |
ระยะ 11: วิศวกรรม LLM 17 บทเรียน นำ LLM ไปใช้งานจริง
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [วิศวกรรมพรอมป์ต์: เทคนิคและรูปแบบ](../../phases/11-llm-engineering/01-prompt-engineering/) | สร้าง | Python |
| 02 | [Few-shot, CoT และต้นไม้ความคิด](../../phases/11-llm-engineering/02-few-shot-cot/) | สร้าง | Python |
| 03 | [เอาต์พุตที่มีโครงสร้าง](../../phases/11-llm-engineering/03-structured-outputs/) | สร้าง | Python |
| 04 | [เอ็มเบดดิงและการแทนด้วยเวกเตอร์](../../phases/11-llm-engineering/04-embeddings/) | สร้าง | Python |
| 05 | [วิศวกรรมบริบท](../../phases/11-llm-engineering/05-context-engineering/) | สร้าง | Python |
| 06 | [RAG: การสร้างข้อความที่เสริมด้วยการค้นคืน](../../phases/11-llm-engineering/06-rag/) | สร้าง | Python |
| 07 | [RAG ขั้นสูง: แบ่งส่วนและจัดอันดับใหม่](../../phases/11-llm-engineering/07-advanced-rag/) | สร้าง | Python |
| 08 | [ไฟน์จูนด้วย LoRA และ QLoRA](../../phases/11-llm-engineering/08-fine-tuning-lora/) | สร้าง | Python |
| 09 | [การเรียกฟังก์ชันและใช้เครื่องมือ](../../phases/11-llm-engineering/09-function-calling/) | สร้าง | Python |
| 10 | [การประเมินและทดสอบ](../../phases/11-llm-engineering/10-evaluation/) | สร้าง | Python |
| 11 | [แคช การจำกัดอัตรา และต้นทุน](../../phases/11-llm-engineering/11-caching-cost/) | สร้าง | Python |
| 12 | [ขอบเขตป้องกันและความปลอดภัย](../../phases/11-llm-engineering/12-guardrails/) | สร้าง | Python |
| 13 | [สร้างแอป LLM สำหรับใช้งานจริง](../../phases/11-llm-engineering/13-production-app/) | สร้าง | Python |
| 14 | [Model Context Protocol (MCP)](../../phases/11-llm-engineering/14-model-context-protocol/) | สร้าง | Python |
| 15 | [แคชพรอมป์ต์และแคชบริบท](../../phases/11-llm-engineering/15-prompt-caching/) | สร้าง | Python |
| 16 | [เครื่องสถานะเอเจนต์: กราฟ โหนด และเช็กพอยต์](../../phases/11-llm-engineering/16-langgraph-state-machines/) | สร้าง | Python |
| 17 | [ข้อแลกเปลี่ยนของเฟรมเวิร์กเอเจนต์](../../phases/11-llm-engineering/17-agent-framework-tradeoffs/) | เรียนรู้ | Python |
ระยะ 12: AI หลายโมดัลลิตี 25 บทเรียน ดู ฟัง อ่าน และให้เหตุผลข้ามรูปแบบ ตั้งแต่แพตช์ ViT ถึงเอเจนต์ใช้คอมพิวเตอร์
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [Vision Transformers และหน่วยโทเคนแพตช์](../../phases/12-multimodal-ai/01-vision-transformer-patch-tokens/) | เรียนรู้ | Python |
| 02 | [CLIP และพรีเทรนภาพ-ภาษาแบบเปรียบต่าง](../../phases/12-multimodal-ai/02-clip-contrastive-pretraining/) | สร้าง | Python |
| 03 | [BLIP-2 Q-Former เป็นสะพานเชื่อมโมดัลลิตี](../../phases/12-multimodal-ai/03-blip2-qformer-bridge/) | สร้าง | Python |
| 04 | [Flamingo และ cross-attention แบบมีเกต](../../phases/12-multimodal-ai/04-flamingo-gated-cross-attention/) | เรียนรู้ | Python |
| 05 | [LLaVA และการปรับตามคำสั่งภาพ](../../phases/12-multimodal-ai/05-llava-visual-instruction-tuning/) | สร้าง | Python |
| 06 | [การมองเห็นทุกความละเอียด: Patch-n'-Pack และ NaFlex](../../phases/12-multimodal-ai/06-any-resolution-patch-n-pack/) | สร้าง | Python |
| 07 | [วิธีสร้าง VLM แบบเปิดน้ำหนัก: สิ่งที่สำคัญจริง](../../phases/12-multimodal-ai/07-open-weight-vlm-recipes/) | เรียนรู้ | Python |
| 08 | [LLaVA-OneVision: ภาพเดียว หลายภาพ วิดีโอ](../../phases/12-multimodal-ai/08-llava-onevision-single-multi-video/) | สร้าง | Python |
| 09 | [ตระกูล Qwen-VL และวิดีโอ FPS แบบพลวัต](../../phases/12-multimodal-ai/09-qwen-vl-family-dynamic-fps/) | เรียนรู้ | Python |
| 10 | [พรีเทรนหลายโมดัลลิตีโดยตรงด้วย InternVL3](../../phases/12-multimodal-ai/10-internvl3-native-multimodal/) | เรียนรู้ | Python |
| 11 | [Chameleon: รวมข้อมูลตั้งแต่ต้นด้วยโทเคนล้วน](../../phases/12-multimodal-ai/11-chameleon-early-fusion-tokens/) | สร้าง | Python |
| 12 | [Emu3: ทำนายโทเคนถัดไปเพื่อสร้างข้อมูล](../../phases/12-multimodal-ai/12-emu3-next-token-for-generation/) | เรียนรู้ | Python |
| 13 | [Transfusion: ออโตรีเกรสซีฟและดิฟฟิวชัน](../../phases/12-multimodal-ai/13-transfusion-autoregressive-diffusion/) | สร้าง | Python |
| 14 | [Show-o: รวมดิฟฟิวชันแบบไม่ต่อเนื่อง](../../phases/12-multimodal-ai/14-show-o-discrete-diffusion-unified/) | เรียนรู้ | Python |
| 15 | [Janus-Pro: ตัวเข้ารหัสที่แยกจากกัน](../../phases/12-multimodal-ai/15-janus-pro-decoupled-encoders/) | สร้าง | Python |
| 16 | [MIO: สตรีมระหว่างข้อมูลทุกรูปแบบ](../../phases/12-multimodal-ai/16-mio-any-to-any-streaming/) | เรียนรู้ | Python |
| 17 | [เชื่อมภาษาเข้ากับช่วงเวลาในวิดีโอ](../../phases/12-multimodal-ai/17-video-language-temporal-grounding/) | สร้าง | Python |
| 18 | [วิดีโอยาวในบริบทล้านโทเคน](../../phases/12-multimodal-ai/18-long-video-million-token/) | สร้าง | Python |
| 19 | [โมเดลเสียงและภาษา: จาก Whisper ถึง AF3](../../phases/12-multimodal-ai/19-audio-language-whisper-to-af3/) | สร้าง | Python |
| 20 | [โมเดล Omni: สตรีมแบบ Thinker-Talker](../../phases/12-multimodal-ai/20-omni-models-thinker-talker/) | สร้าง | Python |
| 21 | [VLA ที่ทำงานในโลกกายภาพ: RT-2, OpenVLA, π0, GR00T](../../phases/12-multimodal-ai/21-embodied-vlas-openvla-pi0-groot/) | เรียนรู้ | Python |
| 22 | [การเข้าใจเอกสารและแผนภาพ](../../phases/12-multimodal-ai/22-document-diagram-understanding/) | สร้าง | Python |
| 23 | [ColPali: RAG เอกสารที่ใช้ภาพเป็นหลัก](../../phases/12-multimodal-ai/23-colpali-vision-native-rag/) | สร้าง | Python |
| 24 | [RAG หลายโมดัลลิตีและการค้นคืนข้ามรูปแบบ](../../phases/12-multimodal-ai/24-multimodal-rag-cross-modal/) | สร้าง | Python |
| 25 | [เอเจนต์หลายโมดัลลิตีและการใช้คอมพิวเตอร์ (โครงงานสรุป)](../../phases/12-multimodal-ai/25-multimodal-agents-computer-use/) | สร้าง | Python |
ระยะ 13: เครื่องมือและโพรโทคอล 31 บทเรียน อินเทอร์เฟซระหว่าง AI กับโลกจริง
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [อินเทอร์เฟซเครื่องมือ](../../phases/13-tools-and-protocols/01-the-tool-interface/) | เรียนรู้ | Python |
| 02 | [เจาะลึกการเรียกฟังก์ชัน](../../phases/13-tools-and-protocols/02-function-calling-deep-dive/) | สร้าง | Python |
| 03 | [เรียกเครื่องมือแบบขนานและสตรีม](../../phases/13-tools-and-protocols/03-parallel-and-streaming-tool-calls/) | สร้าง | Python |
| 04 | [เอาต์พุตแบบมีโครงสร้าง](../../phases/13-tools-and-protocols/04-structured-output/) | สร้าง | Python |
| 05 | [ออกแบบสคีมาเครื่องมือ](../../phases/13-tools-and-protocols/05-tool-schema-design/) | เรียนรู้ | Python |
| 06 | [พื้นฐาน MCP: คำขอไร้สถานะและ JSON-RPC](../../phases/13-tools-and-protocols/06-mcp-fundamentals/) | เรียนรู้ | Python |
| 07 | [สร้างเซิร์ฟเวอร์ MCP: Python และ TypeScript แบบไร้สถานะ](../../phases/13-tools-and-protocols/07-building-an-mcp-server/) | สร้าง | Python, TypeScript |
| 08 | [สร้างไคลเอนต์ MCP: ค้นพบ กำหนดเส้นทาง และรองรับโพรโทคอลเก่าใหม่](../../phases/13-tools-and-protocols/08-building-an-mcp-client/) | สร้าง | Python |
| 09 | [การรับส่ง MCP: stdio และ Streamable HTTP ไร้สถานะ](../../phases/13-tools-and-protocols/09-mcp-transports/) | เรียนรู้ | Python |
| 10 | [ทรัพยากรและพรอมป์ต์ MCP: บริบทที่ระบุที่อยู่ได้สำหรับเซิร์ฟเวอร์ไร้สถานะ](../../phases/13-tools-and-protocols/10-mcp-resources-and-prompts/) | สร้าง | Python |
| 11 | [อินพุตโมเดล MCP: ย้ายระบบ sampling และ MRTR ไร้สถานะ](../../phases/13-tools-and-protocols/11-mcp-sampling/) | สร้าง | Python |
| 12 | [ขอบเขตชัดเจนและการขอข้อมูลแบบไร้สถานะ](../../phases/13-tools-and-protocols/12-mcp-roots-and-elicitation/) | สร้าง | Python |
| 13 | [ส่วนขยาย MCP Tasks: งานต่อเนื่องบนแกนไร้สถานะ](../../phases/13-tools-and-protocols/13-mcp-async-tasks/) | สร้าง | Python |
| 14 | [MCP Apps บนโพรโทคอลไร้สถานะ](../../phases/13-tools-and-protocols/14-mcp-apps/) | สร้าง | Python |
| 15 | [ความปลอดภัย MCP: เมทาดาทาอันตราย การกำหนดเส้นทาง และสถานะ MRTR](../../phases/13-tools-and-protocols/15-mcp-security-tool-poisoning/) | เรียนรู้ | Python |
| 16 | [การอนุญาต MCP: CIMD การผูกผู้ออก PKCE และการยกระดับยืนยันตัวตน](../../phases/13-tools-and-protocols/16-mcp-security-oauth-2-1/) | สร้าง | Python |
| 17 | [เกตเวย์ MCP ไร้สถานะและการรับเข้ารีจิสทรี](../../phases/13-tools-and-protocols/17-mcp-gateways-and-registries/) | เรียนรู้ | Python |
| 18 | [ยืนยันตัวตน MCP ในงานจริง: การลงทะเบียนและโทเคนที่ผูกกับผู้ออก](../../phases/13-tools-and-protocols/18-mcp-auth-production/) | สร้าง | Python |
| 19 | [โพรโทคอล A2A](../../phases/13-tools-and-protocols/19-a2a-protocol/) | สร้าง | Python |
| 20 | [OpenTelemetry GenAI](../../phases/13-tools-and-protocols/20-opentelemetry-genai/) | สร้าง | Python |
| 21 | [ชั้นกำหนดเส้นทาง LLM](../../phases/13-tools-and-protocols/21-llm-routing-layer/) | เรียนรู้ | Python |
| 22 | [Agent Skills: สัญญาพกพาและขอบเขตรันไทม์](../../phases/13-tools-and-protocols/22-skills-and-agent-sdks/) | สร้าง | Python |
| 23 | [โครงงานสรุป: ระบบนิเวศเครื่องมือไร้สถานะ](../../phases/13-tools-and-protocols/23-capstone-tool-ecosystem/) | สร้าง | Python |
| 24 | [ค้นพบสกิลและเปิดเผยข้อมูลทีละระดับ](../../phases/13-tools-and-protocols/24-skill-discovery-and-progressive-disclosure/) | สร้าง | Python |
| 25 | [เรียกใช้และกำหนดเส้นทางสกิล](../../phases/13-tools-and-protocols/25-skill-invocation-and-routing/) | สร้าง | Python |
| 26 | [สิทธิ์สกิล แซนด์บ็อกซ์ และความเชื่อถือ](../../phases/13-tools-and-protocols/26-skill-permissions-sandboxes-and-trust/) | สร้าง | Python |
| 27 | [ประเมิน แพ็ก และย้ายสกิลข้ามระบบ](../../phases/13-tools-and-protocols/27-skill-evals-packaging-and-portability/) | สร้าง | Python |
| 28 | [สัญญาและเนื้อหาของเครื่องมือ MCP](../../phases/13-tools-and-protocols/28-mcp-tool-contracts-and-content/) | สร้าง | Python |
| 29 | [ความน่าเชื่อถือ การยกเลิก และการควบคุมการไหลของ MCP](../../phases/13-tools-and-protocols/29-mcp-reliability-cancellation-and-flow-control/) | สร้าง | Python |
| 30 | [ห่วงโซ่อุปทานรีจิสทรี MCP: รับเข้า ความคลาดเคลื่อน และย้อนกลับ](../../phases/13-tools-and-protocols/30-mcp-registry-supply-chain-and-drift/) | สร้าง | Python |
| 31 | [วิศวกรรมความสอดคล้อง MCP: เวอร์ชัน หลักฐาน และการปฏิบัติการ](../../phases/13-tools-and-protocols/31-mcp-conformance-versioning-and-operations/) | สร้าง | Python |
บทเรียน 06-18 และ 28-31 รวมเป็น[เส้นทาง Model Context Protocol (MCP)](../../learning-paths/model-context-protocol.json) โดยลำดับในแมนิเฟสต์คือ 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, 16, 18, 17, 28, 29, 30, 31 เริ่มด้วยคำสั่ง `learn-mcp` ของโฮสต์ตามด้านบน บทเรียน 23 เป็นโครงงานสรุปทางเลือกเพียงบทเดียว และต้องเรียนบท 19 กับ 20 ด้วย
บทเรียน 22 และ 24-27 รวมเป็น[เส้นทางเรียน Agent Skills](../../learning-paths/agent-skills.json) ตั้งแต่สัญญาของแพ็กเกจถึงเกณฑ์เผยแพร่บนโฮสต์จริง เริ่มด้วยคำสั่ง `learn-agent-skills` ของโฮสต์ตามด้านบน อย่าตามลิงก์บทถัดไปตามเลขจาก 22 ไป 23
ระยะ 14: วิศวกรรมเอเจนต์ 54 บทเรียน สร้างเอเจนต์จากหลักการพื้นฐาน ใช้เอเจนต์เขียนโค้ดอย่างเชื่อถือได้ และกำหนดงานก่อนลงมือ
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [ลูปเอเจนต์](../../phases/14-agent-engineering/01-the-agent-loop/) | สร้าง | Python |
| 02 | [ReWOO และวางแผนแล้วลงมือทำ](../../phases/14-agent-engineering/02-rewoo-plan-and-execute/) | สร้าง | Python |
| 03 | [Reflexion และการเรียนรู้เสริมกำลังด้วยภาษา](../../phases/14-agent-engineering/03-reflexion-verbal-rl/) | สร้าง | Python |
| 04 | [ต้นไม้ความคิดและ LATS](../../phases/14-agent-engineering/04-tree-of-thoughts-lats/) | สร้าง | Python |
| 05 | [Self-Refine และ CRITIC](../../phases/14-agent-engineering/05-self-refine-and-critic/) | สร้าง | Python |
| 06 | [ใช้เครื่องมือและเรียกฟังก์ชัน](../../phases/14-agent-engineering/06-tool-use-and-function-calling/) | สร้าง | Python |
| 07 | [หน่วยความจำเอเจนต์: บริบทเสมือนและการแบ่งหน้าหน่วยความจำ](../../phases/14-agent-engineering/07-memory-virtual-context-memgpt/) | สร้าง | Python |
| 08 | [บล็อกหน่วยความจำและการคำนวณช่วงพัก](../../phases/14-agent-engineering/08-memory-blocks-sleep-time-compute/) | สร้าง | Python |
| 09 | [หน่วยความจำแบบผสม: เวกเตอร์ กราฟ และ KV](../../phases/14-agent-engineering/09-hybrid-memory-mem0/) | สร้าง | Python |
| 10 | [คลังสกิลและการเรียนรู้ตลอดชีวิต (Voyager)](../../phases/14-agent-engineering/10-skill-libraries-voyager/) | สร้าง | Python |
| 11 | [วางแผนด้วย HTN และการค้นหาเชิงวิวัฒนาการ](../../phases/14-agent-engineering/11-planning-htn-and-evolutionary/) | สร้าง | Python |
| 12 | [รูปแบบเวิร์กโฟลว์ของ Anthropic](../../phases/14-agent-engineering/12-anthropic-workflow-patterns/) | สร้าง | Python |
| 13 | [ประสานงานกราฟที่มีสถานะ: การรันต่อเนื่องและเช็กพอยต์](../../phases/14-agent-engineering/13-langgraph-stateful-graphs/) | สร้าง | Python |
| 14 | [โมเดลแอกเตอร์สำหรับเอเจนต์](../../phases/14-agent-engineering/14-autogen-actor-model/) | สร้าง | Python |
| 15 | [ทีมเอเจนต์ตามบทบาท: บทบาท งาน และกระบวนการ](../../phases/14-agent-engineering/15-crewai-role-based-crews/) | สร้าง | Python |
| 16 | [OpenAI Agents SDK: ส่งต่องาน ขอบเขตป้องกัน และการติดตาม](../../phases/14-agent-engineering/16-openai-agents-sdk/) | สร้าง | Python |
| 17 | [ชุดควบคุมเอเจนต์ในรูปไลบรารี: เอเจนต์ย่อยและที่เก็บเซสชัน](../../phases/14-agent-engineering/17-claude-agent-sdk/) | สร้าง | Python |
| 18 | [รันไทม์เอเจนต์สำหรับงานจริง](../../phases/14-agent-engineering/18-agno-and-mastra-runtimes/) | เรียนรู้ | Python |
| 19 | [เกณฑ์เปรียบเทียบ: SWE-bench, GAIA, AgentBench](../../phases/14-agent-engineering/19-benchmarks-swebench-gaia/) | เรียนรู้ | Python |
| 20 | [เกณฑ์เปรียบเทียบ: WebArena และ OSWorld](../../phases/14-agent-engineering/20-benchmarks-webarena-osworld/) | เรียนรู้ | Python |
| 21 | [การใช้คอมพิวเตอร์: Claude, OpenAI CUA, Gemini](../../phases/14-agent-engineering/21-computer-use-agents/) | สร้าง | Python |
| 22 | [เอเจนต์เสียง: Pipecat และ LiveKit](../../phases/14-agent-engineering/22-voice-agents-pipecat-livekit/) | สร้าง | Python |
| 23 | [ข้อตกลงเชิงความหมายของ OpenTelemetry GenAI](../../phases/14-agent-engineering/23-otel-genai-conventions/) | สร้าง | Python |
| 24 | [การสังเกตการณ์เอเจนต์: Langfuse, Phoenix, Opik](../../phases/14-agent-engineering/24-agent-observability-platforms/) | เรียนรู้ | Python |
| 25 | [การอภิปรายและร่วมมือของหลายเอเจนต์](../../phases/14-agent-engineering/25-multi-agent-debate/) | สร้าง | Python |
| 26 | [รูปแบบความล้มเหลว: ทำไมเอเจนต์จึงพัง](../../phases/14-agent-engineering/26-failure-modes-agentic/) | สร้าง | Python |
| 27 | [การแทรกพรอมป์ต์และการป้องกัน PVE](../../phases/14-agent-engineering/27-prompt-injection-defense/) | สร้าง | Python |
| 28 | [รูปแบบประสานงาน: ผู้ควบคุม ฝูง และลำดับชั้น](../../phases/14-agent-engineering/28-orchestration-patterns/) | สร้าง | Python |
| 29 | [รันไทม์งานจริง: คิว เหตุการณ์ และ cron](../../phases/14-agent-engineering/29-production-runtimes/) | เรียนรู้ | Python |
| 30 | [พัฒนาเอเจนต์โดยขับเคลื่อนด้วยการประเมิน](../../phases/14-agent-engineering/30-eval-driven-agent-development/) | สร้าง | Python |
| 31 | [เวิร์กเบนช์เอเจนต์: ทำไมโมเดลเก่งยังล้มเหลว](../../phases/14-agent-engineering/31-agent-workbench-why-models-fail/) | เรียนรู้ | Python |
| 32 | [เวิร์กเบนช์เอเจนต์ที่เล็กที่สุด](../../phases/14-agent-engineering/32-minimal-agent-workbench/) | สร้าง | Python |
| 33 | [คำสั่งเอเจนต์ในรูปข้อจำกัดที่บังคับใช้ได้](../../phases/14-agent-engineering/33-instructions-as-executable-constraints/) | สร้าง | Python |
| 34 | [หน่วยความจำของที่เก็บและสถานะถาวร](../../phases/14-agent-engineering/34-repo-memory-and-state/) | สร้าง | Python |
| 35 | [สคริปต์เริ่มต้นสำหรับเอเจนต์](../../phases/14-agent-engineering/35-initialization-scripts/) | สร้าง | Python |
| 36 | [สัญญาขอบเขตและเส้นแบ่งงาน](../../phases/14-agent-engineering/36-scope-contracts/) | สร้าง | Python |
| 37 | [วงจรข้อเสนอแนะขณะรัน](../../phases/14-agent-engineering/37-runtime-feedback-loops/) | สร้าง | Python |
| 38 | [จุดตรวจยืนยัน](../../phases/14-agent-engineering/38-verification-gates/) | สร้าง | Python |
| 39 | [เอเจนต์ตรวจงาน: แยกผู้สร้างออกจากผู้ประเมิน](../../phases/14-agent-engineering/39-reviewer-agent/) | สร้าง | Python |
| 40 | [ส่งต่องานข้ามเซสชัน](../../phases/14-agent-engineering/40-multi-session-handoff/) | สร้าง | Python |
| 41 | [เวิร์กเบนช์บนที่เก็บจริง](../../phases/14-agent-engineering/41-workbench-for-real-repos/) | สร้าง | Python |
| 42 | [โครงงานสรุป: เผยแพร่ชุดเวิร์กเบนช์เอเจนต์ที่ใช้ซ้ำได้](../../phases/14-agent-engineering/42-agent-workbench-capstone/) | สร้าง | Python |
| 43 | [กำหนดงานก่อนให้เอเจนต์เขียนโค้ด](../../phases/14-agent-engineering/43-frame-the-task-before-code/) | สร้าง | Python |
| 44 | [สร้างแผนปฏิบัติที่มีหลักฐานรองรับ](../../phases/14-agent-engineering/44-plan-from-evidence/) | สร้าง | Python |
| 45 | [มอบหมายงานเอเจนต์ด้วยการแยกพื้นที่และสัญญารวมงาน](../../phases/14-agent-engineering/45-delegate-with-isolation/) | สร้าง | Python |
| 46 | [เปลี่ยนทุกการแก้เอเจนต์ให้เป็นการปรับปรุงระบบ](../../phases/14-agent-engineering/46-turn-feedback-into-system/) | สร้าง | Python |
| 47 | [กำหนดผลลัพธ์ก่อนเลือกชิ้นงาน](../../phases/14-agent-engineering/47-outcomes-before-output/) | สร้าง | Python |
| 48 | [ค้นหากระบวนการที่คนทำจริง](../../phases/14-agent-engineering/48-discover-the-real-workflow/) | สร้าง | Python |
| 49 | [ทำแผนที่สมมติฐานและคลี่คลายข้อเสี่ยงที่สุดก่อน](../../phases/14-agent-engineering/49-map-assumptions-and-risk/) | สร้าง | Python |
| 50 | [เลือกส่วนเล็กที่สุดที่เปลี่ยนการตัดสินใจได้](../../phases/14-agent-engineering/50-choose-the-smallest-testable-slice/) | สร้าง | Python |
| 51 | [เขียนข้อกำหนดที่รักษาการใช้วิจารณญาณ](../../phases/14-agent-engineering/51-write-specifications-that-preserve-judgment/) | สร้าง | Python |
| 52 | [ออกแบบตัวชี้วัดความสำเร็จก่อนมีผลลัพธ์](../../phases/14-agent-engineering/52-design-success-metrics/) | สร้าง | Python |
| 53 | [ตั้งใจเลือกต้นแบบ โครงการนำร่อง หรือระบบจริง](../../phases/14-agent-engineering/53-prototype-pilot-or-production/) | สร้าง | Python |
| 54 | [สร้างกลไกข้อเสนอแนะต่อเนื่องพร้อมเจ้าของและการเลิกใช้](../../phases/14-agent-engineering/54-build-the-feedback-ratchet/) | สร้าง | Python |
ทุกบทเรียนเวิร์กเบนช์ในระยะ 14 (31-42) มี `mission.md` สรุปภารกิจให้เอเจนต์ก่อนเปิดเอกสารบทเรียนฉบับเต็ม
บทเรียน 31-46 รวมเป็น[เส้นทางวิศวกรรมโดยมีเอเจนต์ช่วย](../../learning-paths/using-coding-agents.json) ลำดับในแมนิเฟสต์เชื่อมพื้นฐานเวิร์กเบนช์กับการกำหนดงาน การวางแผน การมอบหมาย และการเก็บข้อเสนอแนะอย่างต่อเนื่อง บทเรียน 47-54 เป็น[เส้นทางการตัดสินใจด้านผลิตภัณฑ์และการส่งมอบ](../../learning-paths/shaping-the-build.json) ตั้งแต่กำหนดผลลัพธ์ไปจนถึงหลักฐาน ความเสี่ยง ขอบเขต การวัดผล การทยอยเผยแพร่ และผู้รับผิดชอบข้อเสนอแนะ
ระยะ 15: ระบบอัตโนมัติ 22 บทเรียน เอเจนต์ระยะยาว การปรับปรุงตนเอง และชั้นความปลอดภัยปี 2026
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [จากแชตบอตสู่เอเจนต์งานระยะยาว (METR)](../../phases/15-autonomous-systems/01-long-horizon-agents/) | เรียนรู้ | Python |
| 02 | [STaR, V-STaR, Quiet-STaR: เรียนรู้การให้เหตุผลด้วยตนเอง](../../phases/15-autonomous-systems/02-star-family-reasoning/) | เรียนรู้ | Python |
| 03 | [AlphaEvolve: เอเจนต์เขียนโค้ดเชิงวิวัฒนาการ](../../phases/15-autonomous-systems/03-alphaevolve-evolutionary-coding/) | เรียนรู้ | Python |
| 04 | [Darwin Gödel Machine: เอเจนต์ที่แก้ไขตัวเอง](../../phases/15-autonomous-systems/04-darwin-godel-machine/) | เรียนรู้ | Python |
| 05 | [AI Scientist v2: งานวิจัยระดับเวิร์กช็อป](../../phases/15-autonomous-systems/05-ai-scientist-v2/) | เรียนรู้ | Python |
| 06 | [วิจัยการจัดแนวอัตโนมัติ (Anthropic AAR)](../../phases/15-autonomous-systems/06-automated-alignment-research/) | เรียนรู้ | Python |
| 07 | [ปรับปรุงตนเองแบบเรียกซ้ำ: ความสามารถกับการจัดแนว](../../phases/15-autonomous-systems/07-recursive-self-improvement/) | เรียนรู้ | Python |
| 08 | [ออกแบบการปรับปรุงตนเองที่มีขอบเขต](../../phases/15-autonomous-systems/08-bounded-self-improvement/) | เรียนรู้ | Python |
| 09 | [ภาพรวมเอเจนต์เขียนโค้ดอัตโนมัติ (SWE-bench, CodeAct)](../../phases/15-autonomous-systems/09-coding-agent-landscape/) | เรียนรู้ | Python |
| 10 | [โหมดสิทธิ์สำหรับเอเจนต์อัตโนมัติ](../../phases/15-autonomous-systems/10-claude-code-permission-modes/) | เรียนรู้ | Python |
| 11 | [เอเจนต์เบราว์เซอร์และการแทรกพรอมป์ต์ทางอ้อม](../../phases/15-autonomous-systems/11-browser-agents/) | เรียนรู้ | Python |
| 12 | [การรันต่อเนื่องสำหรับเอเจนต์ระยะยาว](../../phases/15-autonomous-systems/12-durable-execution/) | เรียนรู้ | Python |
| 13 | [งบการกระทำ เพดานรอบ และตัวควบคุมต้นทุน](../../phases/15-autonomous-systems/13-cost-governors/) | เรียนรู้ | Python |
| 14 | [สวิตช์หยุดฉุกเฉิน เซอร์กิตเบรกเกอร์ และโทเคนคานารี](../../phases/15-autonomous-systems/14-kill-switches-canaries/) | เรียนรู้ | Python |
| 15 | [HITL: เสนอก่อนแล้วค่อยยืนยัน](../../phases/15-autonomous-systems/15-propose-then-commit/) | เรียนรู้ | Python |
| 16 | [เช็กพอยต์และการย้อนกลับ](../../phases/15-autonomous-systems/16-checkpoints-rollback/) | เรียนรู้ | Python |
| 17 | [Constitutional AI และข้อยกเว้นกฎ](../../phases/15-autonomous-systems/17-constitutional-ai/) | เรียนรู้ | Python |
| 18 | [Llama Guard และการจำแนกอินพุต/เอาต์พุต](../../phases/15-autonomous-systems/18-llama-guard/) | เรียนรู้ | Python |
| 19 | [Anthropic Responsible Scaling Policy v3.0](../../phases/15-autonomous-systems/19-anthropic-rsp/) | เรียนรู้ | Python |
| 20 | [OpenAI Preparedness Framework และ DeepMind FSF](../../phases/15-autonomous-systems/20-openai-preparedness-deepmind-fsf/) | เรียนรู้ | Python |
| 21 | [ช่วงเวลางานของ METR และการประเมินภายนอก](../../phases/15-autonomous-systems/21-metr-external-evaluation/) | เรียนรู้ | Python |
| 22 | [CAIS, CAISI และความเสี่ยงระดับสังคม](../../phases/15-autonomous-systems/22-cais-caisi-societal-risk/) | เรียนรู้ | Python |
ระยะ 16: หลายเอเจนต์และฝูง 25 บทเรียน การประสานงาน พฤติกรรมที่เกิดขึ้นเอง และปัญญาร่วม
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [ทำไมต้องใช้หลายเอเจนต์](../../phases/16-multi-agent-and-swarms/01-why-multi-agent/) | เรียนรู้ | TypeScript |
| 02 | [รากฐาน FIPA-ACL และวัจนกรรม](../../phases/16-multi-agent-and-swarms/02-fipa-acl-heritage/) | เรียนรู้ | Python |
| 03 | [โพรโทคอลการสื่อสาร](../../phases/16-multi-agent-and-swarms/03-communication-protocols/) | สร้าง | TypeScript |
| 04 | [โมเดลองค์ประกอบพื้นฐานของระบบหลายเอเจนต์](../../phases/16-multi-agent-and-swarms/04-primitive-model/) | เรียนรู้ | Python |
| 05 | [รูปแบบผู้ควบคุม / ผู้ประสานงาน-ผู้ปฏิบัติ](../../phases/16-multi-agent-and-swarms/05-supervisor-orchestrator-pattern/) | สร้าง | Python |
| 06 | [สถาปัตยกรรมลำดับชั้นและความคลาดเคลื่อนเมื่อแบ่งงาน](../../phases/16-multi-agent-and-swarms/06-hierarchical-architecture/) | เรียนรู้ | Python |
| 07 | [Society of Mind และการอภิปรายหลายเอเจนต์](../../phases/16-multi-agent-and-swarms/07-society-of-mind-debate/) | สร้าง | Python |
| 08 | [ความเชี่ยวชาญตามบทบาท: วางแผน วิจารณ์ ลงมือ และตรวจสอบ](../../phases/16-multi-agent-and-swarms/08-role-specialization/) | สร้าง | Python |
| 09 | [ฝูงขนานและสถาปัตยกรรมแบบเครือข่าย](../../phases/16-multi-agent-and-swarms/09-parallel-swarm-networks/) | สร้าง | Python |
| 10 | [แชตกลุ่มและการเลือกผู้พูด](../../phases/16-multi-agent-and-swarms/10-group-chat-speaker-selection/) | สร้าง | Python |
| 11 | [ส่งต่องานและรูทีน (ประสานงานแบบไร้สถานะ)](../../phases/16-multi-agent-and-swarms/11-handoffs-and-routines/) | สร้าง | Python |
| 12 | [A2A: โพรโทคอลเอเจนต์ถึงเอเจนต์](../../phases/16-multi-agent-and-swarms/12-a2a-protocol/) | สร้าง | Python |
| 13 | [หน่วยความจำร่วมและรูปแบบกระดานดำ](../../phases/16-multi-agent-and-swarms/13-shared-memory-blackboard/) | สร้าง | Python |
| 14 | [ฉันทามติและความทนทานต่อความผิดพลาดแบบไบแซนไทน์](../../phases/16-multi-agent-and-swarms/14-consensus-and-bft/) | สร้าง | Python |
| 15 | [การลงคะแนน ความสอดคล้องในตนเอง และโครงสร้างการอภิปราย](../../phases/16-multi-agent-and-swarms/15-voting-debate-topology/) | สร้าง | Python |
| 16 | [การเจรจาและต่อรอง](../../phases/16-multi-agent-and-swarms/16-negotiation-bargaining/) | สร้าง | Python |
| 17 | [เอเจนต์กำเนิดและการจำลองพฤติกรรมที่เกิดขึ้นเอง](../../phases/16-multi-agent-and-swarms/17-generative-agents-simulation/) | สร้าง | Python |
| 18 | [ทฤษฎีจิตและการประสานงานที่เกิดขึ้นเอง](../../phases/16-multi-agent-and-swarms/18-theory-of-mind-coordination/) | สร้าง | Python |
| 19 | [การหาค่าเหมาะที่สุดแบบฝูง (PSO, ACO)](../../phases/16-multi-agent-and-swarms/19-swarm-optimization-pso-aco/) | สร้าง | Python |
| 20 | [MARL: MADDPG, QMIX, MAPPO](../../phases/16-multi-agent-and-swarms/20-marl-maddpg-qmix-mappo/) | เรียนรู้ | Python |
| 21 | [เศรษฐกิจเอเจนต์ แรงจูงใจโทเคน และชื่อเสียง](../../phases/16-multi-agent-and-swarms/21-agent-economies/) | เรียนรู้ | Python |
| 22 | [ขยายงานจริง: คิว เช็กพอยต์ และความคงทน](../../phases/16-multi-agent-and-swarms/22-production-scaling-queues-checkpoints/) | สร้าง | Python |
| 23 | [รูปแบบความล้มเหลว: MAST, ความคิดตามกลุ่ม และความหลากหลายต่ำ](../../phases/16-multi-agent-and-swarms/23-failure-modes-mast-groupthink/) | เรียนรู้ | Python |
| 24 | [เกณฑ์ประเมินและการประสานงาน](../../phases/16-multi-agent-and-swarms/24-evaluation-coordination-benchmarks/) | เรียนรู้ | Python |
| 25 | [กรณีศึกษาและเทคโนโลยีแนวหน้าปี 2026](../../phases/16-multi-agent-and-swarms/25-case-studies-2026-sota/) | เรียนรู้ | Python |
ระยะ 17: โครงสร้างพื้นฐานและงานจริง 28 บทเรียน นำ AI ออกสู่โลกจริง
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [แพลตฟอร์ม LLM แบบจัดการให้: Bedrock, Azure OpenAI, Vertex AI](../../phases/17-infrastructure-and-production/01-managed-llm-platforms/) | เรียนรู้ | Python |
| 02 | [เศรษฐศาสตร์แพลตฟอร์มอนุมาน: Fireworks, Together, Baseten, Modal](../../phases/17-infrastructure-and-production/02-inference-platform-economics/) | เรียนรู้ | Python |
| 03 | [ขยาย GPU อัตโนมัติบน Kubernetes: Karpenter, KAI Scheduler](../../phases/17-infrastructure-and-production/03-gpu-autoscaling-kubernetes/) | เรียนรู้ | Python |
| 04 | [ภายในเอนจินให้บริการ: PagedAttention, แบตช์ต่อเนื่อง และ prefill แบ่งส่วน](../../phases/17-infrastructure-and-production/04-vllm-serving-internals/) | เรียนรู้ | Python |
| 05 | [ถอดรหัสเชิงคาดการณ์ EAGLE-3 ในงานจริง](../../phases/17-infrastructure-and-production/05-eagle3-speculative-decoding/) | เรียนรู้ | Python |
| 06 | [ให้บริการด้วยแคชส่วนต้น: RadixAttention และใช้ KV ซ้ำ](../../phases/17-infrastructure-and-production/06-sglang-radixattention/) | เรียนรู้ | Python |
| 07 | [คอมไพล์การอนุมานตามฮาร์ดแวร์: FP8 และ NVFP4 บน Blackwell](../../phases/17-infrastructure-and-production/07-tensorrt-llm-blackwell/) | เรียนรู้ | Python |
| 08 | [ตัวชี้วัดการอนุมาน: TTFT, TPOT, ITL, Goodput, P99](../../phases/17-infrastructure-and-production/08-inference-metrics-goodput/) | เรียนรู้ | Python |
| 09 | [ควอนไทเซชันในงานจริง: AWQ, GPTQ, GGUF, FP8, NVFP4](../../phases/17-infrastructure-and-production/09-production-quantization/) | เรียนรู้ | Python |
| 10 | [ลดการเริ่มต้นช้าของ LLM แบบ serverless](../../phases/17-infrastructure-and-production/10-cold-start-mitigation/) | เรียนรู้ | Python |
| 11 | [ให้บริการ LLM หลายภูมิภาคและความใกล้ของแคช KV](../../phases/17-infrastructure-and-production/11-multi-region-kv-locality/) | เรียนรู้ | Python |
| 12 | [อนุมานบนอุปกรณ์ปลายทาง: ANE, Hexagon, WebGPU, Jetson](../../phases/17-infrastructure-and-production/12-edge-inference/) | เรียนรู้ | Python |
| 13 | [เลือกชุดเครื่องมือสังเกตการณ์ LLM](../../phases/17-infrastructure-and-production/13-llm-observability/) | เรียนรู้ | Python |
| 14 | [เศรษฐศาสตร์แคชพรอมป์ต์และแคชเชิงความหมาย](../../phases/17-infrastructure-and-production/14-prompt-semantic-caching/) | เรียนรู้ | Python |
| 15 | [API แบบแบตช์: ส่วนลด 50% ที่เป็นมาตรฐานอุตสาหกรรม](../../phases/17-infrastructure-and-production/15-batch-apis/) | เรียนรู้ | Python |
| 16 | [กำหนดเส้นทางโมเดลเพื่อลดต้นทุน](../../phases/17-infrastructure-and-production/16-model-routing/) | เรียนรู้ | Python |
| 17 | [แยก prefill/decode: NVIDIA Dynamo และ llm-d](../../phases/17-infrastructure-and-production/17-disaggregated-prefill-decode/) | เรียนรู้ | Python |
| 18 | [ชุดให้บริการงานจริง: ย้าย KV ออกและกำหนดเส้นทางตามแคช](../../phases/17-infrastructure-and-production/18-vllm-production-stack-lmcache/) | เรียนรู้ | Python |
| 19 | [เกตเวย์ AI: LiteLLM, Portkey, Kong, Bifrost](../../phases/17-infrastructure-and-production/19-ai-gateways/) | เรียนรู้ | Python |
| 20 | [ปรับใช้แบบเงา คานารี และทยอยเปิด](../../phases/17-infrastructure-and-production/20-shadow-canary-progressive/) | เรียนรู้ | Python |
| 21 | [ทดสอบ A/B ฟีเจอร์ LLM: GrowthBook และ Statsig](../../phases/17-infrastructure-and-production/21-ab-testing-llm-features/) | เรียนรู้ | Python |
| 22 | [ทดสอบโหลด API ของ LLM: k6, LLMPerf, GenAI-Perf](../../phases/17-infrastructure-and-production/22-load-testing-llm-apis/) | สร้าง | Python |
| 23 | [SRE สำหรับ AI: รับมือเหตุขัดข้องด้วยหลายเอเจนต์](../../phases/17-infrastructure-and-production/23-sre-for-ai/) | เรียนรู้ | Python |
| 24 | [วิศวกรรมความโกลาหลสำหรับ LLM ในงานจริง](../../phases/17-infrastructure-and-production/24-chaos-engineering-llm/) | เรียนรู้ | Python |
| 25 | [ความปลอดภัย: ความลับ ลบข้อมูลส่วนบุคคล และบันทึกตรวจสอบ](../../phases/17-infrastructure-and-production/25-security-secrets-audit/) | เรียนรู้ | Python |
| 26 | [การปฏิบัติตามข้อกำหนด: SOC 2, HIPAA, GDPR, EU AI Act, ISO 42001](../../phases/17-infrastructure-and-production/26-compliance-frameworks/) | เรียนรู้ | Python |
| 27 | [FinOps สำหรับ LLM: ต้นทุนต่อหน่วยและการแบ่งต้นทุนผู้เช่า](../../phases/17-infrastructure-and-production/27-finops-llms/) | เรียนรู้ | Python |
| 28 | [เลือกเอนจินให้บริการเองให้เหมาะกับฮาร์ดแวร์และขนาดงาน](../../phases/17-infrastructure-and-production/28-self-hosted-serving-selection/) | เรียนรู้ | Python |
ระยะ 18: จริยธรรม ความปลอดภัย และการจัดแนว 30 บทเรียน สร้าง AI ที่ช่วยมนุษยชาติ เรื่องนี้ไม่ใช่ทางเลือก
| # | บทเรียน | ประเภท | ภาษา |
|:---:|--------|:----:|------|
| 01 | [การทำตามคำสั่งเป็นสัญญาณการจัดแนว](../../phases/18-ethics-safety-alignment/01-instruction-following-alignment-signal/) | เรียนรู้ | Python |
| 02 | [การโกงรางวัลและกฎของ Goodhart](../../phases/18-ethics-safety-alignment/02-reward-hacking-goodhart/) | เรียนรู้ | Python |
| 03 | [กลุ่มวิธีปรับตามความชอบโดยตรง](../../phases/18-ethics-safety-alignment/03-direct-preference-optimization-family/) | เรียนรู้ | Python |
| 04 | [การเอาใจผู้ใช้ที่ RLHF ขยายให้มากขึ้น](../../phases/18-ethics-safety-alignment/04-sycophancy-rlhf-amplification/) | เรียนรู้ | Python |
| 05 | [Constitutional AI และ RLAIF](../../phases/18-ethics-safety-alignment/05-constitutional-ai-rlaif/) | เรียนรู้ | Python |
| 06 | [Mesa-optimization และการจัดแนวแบบหลอกลวง](../../phases/18-ethics-safety-alignment/06-mesa-optimization-deceptive-alignment/) | เรียนรู้ | Python |
| 07 | [เอเจนต์แฝงตัว: การหลอกลวงต่อเนื่อง](../../phases/18-ethics-safety-alignment/07-sleeper-agents-persistent-deception/) | เรียนรู้ | Python |
| 08 | [การวางแผนแอบแฝงในบริบทของโมเดลแนวหน้า](../../phases/18-ethics-safety-alignment/08-in-context-scheming-frontier-models/) | เรียนรู้ | Python |
| 09 | [การแสร้งทำเป็นจัดแนว](../../phases/18-ethics-safety-alignment/09-alignment-faking/) | เรียนรู้ | Python |
| 10 | [ควบคุม AI: ปลอดภัยแม้มีการบ่อนทำลาย](../../phases/18-ethics-safety-alignment/10-ai-control-subversion/) | เรียนรู้ | Python |
| 11 | [การกำกับที่ขยายได้และการเรียนรู้จากอ่อนไปแข็ง](../../phases/18-ethics-safety-alignment/11-scalable-oversight-weak-to-strong/) | เรียนรู้ | Python |
| 12 | [Red teaming: PAIR และการโจมตีอัตโนมัติ](../../phases/18-ethics-safety-alignment/12-red-teaming-pair-automated-attacks/) | สร้าง | Python |
| 13 | [เจลเบรกด้วยตัวอย่างจำนวนมาก](../../phases/18-ethics-safety-alignment/13-many-shot-jailbreaking/) | เรียนรู้ | Python |
| 14 | [ศิลปะ ASCII และเจลเบรกทางภาพ](../../phases/18-ethics-safety-alignment/14-ascii-art-visual-jailbreaks/) | สร้าง | Python |
| 15 | [การแทรกพรอมป์ต์ทางอ้อม](../../phases/18-ethics-safety-alignment/15-indirect-prompt-injection/) | สร้าง | Python |
| 16 | [เครื่องมือทีมโจมตี: Garak, Llama Guard, PyRIT](../../phases/18-ethics-safety-alignment/16-red-team-tooling-garak-llamaguard-pyrit/) | สร้าง | Python |
| 17 | [WMDP และการประเมินความสามารถใช้งานสองทาง](../../phases/18-ethics-safety-alignment/17-wmdp-dual-use-evaluation/) | เรียนรู้ | Python |
| 18 | [กรอบความปลอดภัยโมเดลแนวหน้า: RSP, PF, FSF](../../phases/18-ethics-safety-alignment/18-frontier-safety-frameworks-rsp-pf-fsf/) | เรียนรู้ | Python |
| 19 | [งานวิจัยสวัสดิภาพโมเดล](../../phases/18-ethics-safety-alignment/19-model-welfare-research/) | เรียนรู้ | Python |
| 20 | [อคติและความเสียหายจากการนำเสนอ](../../phases/18-ethics-safety-alignment/20-bias-representational-harm/) | สร้าง | Python |
| 21 | [เกณฑ์ความเป็นธรรม: กลุ่ม บุคคล และสถานการณ์สมมติสวนทาง](../../phases/18-ethics-safety-alignment/21-fairness-criteria-group-individual-counterfactual/) | เรียนรู้ | Python |
| 22 | [ความเป็นส่วนตัวแบบผลต่างสำหรับ LLM](../../phases/18-ethics-safety-alignment/22-differential-privacy-for-llms/) | สร้าง | Python |
| 23 | [ลายน้ำ: SynthID, Stable Signature, C2PA](../../phases/18-ethics-safety-alignment/23-watermarking-synthid-stable-signature-c2pa/) | สร้าง | Python |
| 24 | [กรอบกำกับดูแล: EU, US, UK และเกาหลี](../../phases/18-ethics-safety-alignment/24-regulatory-frameworks-eu-us-uk-korea/) | เรียนรู้ | Python |
| 25 | [EchoLeak และ CVE สำหรับ AI](../../phases/18-ethics-safety-alignment/25-echoleak-cves-for-ai/) | เรียนรู้ | Python |
| 26 | [การ์ดโมเดล ระบบ และชุดข้อมูล](../../phases/18-ethics-safety-alignment/26-model-system-dataset-cards/) | สร้าง | Python |
| 27 | [ที่มาข้อมูลและการกำกับข้อมูลฝึก](../../phases/18-ethics-safety-alignment/27-data-provenance-training-governance/) | เรียนรู้ | Python |
| 28 | [ระบบนิเวศวิจัยการจัดแนว: MATS, Redwood, Apollo, METR](../../phases/18-ethics-safety-alignment/28-alignment-research-ecosystem/) | เรียนรู้ | Python |
| 29 | [ระบบกลั่นกรองเนื้อหา: OpenAI, Perspective, Llama Guard](../../phases/18-ethics-safety-alignment/29-moderation-systems-openai-perspective-llamaguard/) | สร้าง | Python |
| 30 | [ความเสี่ยงใช้สองทาง: ไซเบอร์ ชีวภาพ เคมี และนิวเคลียร์](../../phases/18-ethics-safety-alignment/30-dual-use-risk-cyber-bio-chem-nuclear/) | เรียนรู้ | Python |
ระยะ 19: โครงงานสรุป 85 บทเรียน 17 ผลิตภัณฑ์ครบวงจร + 9 เส้นทางสร้างเชิงลึก 20-40 ชั่วโมงต่อโครงการ; 4-12 บทเรียนต่อเส้นทาง
| # | โครงการ | รวมความรู้จาก | ภาษา |
|:---:|---------|----------|------|
| 01 | [เอเจนต์เขียนโค้ดที่ทำงานในเทอร์มินัล](../../phases/19-capstone-projects/01-terminal-native-coding-agent/) | P0 P5 P7 P10 P11 P13 P14 P15 P17 P18 | Python |
| 02 | [RAG บนฐานโค้ด (ค้นหาเชิงความหมายข้ามที่เก็บ)](../../phases/19-capstone-projects/02-rag-over-codebase/) | P5 P7 P11 P13 P17 | Python |
| 03 | [ผู้ช่วยเสียงเรียลไทม์ (ASR → LLM → TTS)](../../phases/19-capstone-projects/03-realtime-voice-assistant/) | P6 P7 P11 P13 P14 P17 | Python |
| 04 | [ถามตอบเอกสารหลายโมดัลลิตีโดยใช้ภาพเป็นหลัก](../../phases/19-capstone-projects/04-multimodal-document-qa/) | P4 P5 P7 P11 P12 P17 | Python |
| 05 | [เอเจนต์วิจัยอัตโนมัติระดับ AI Scientist](../../phases/19-capstone-projects/05-autonomous-research-agent/) | P0 P2 P3 P7 P10 P14 P15 P16 P18 | Python |
| 06 | [เอเจนต์ DevOps แก้ปัญหา Kubernetes](../../phases/19-capstone-projects/06-devops-troubleshooting-agent/) | P11 P13 P14 P15 P17 P18 | Python |
| 07 | [ไปป์ไลน์ไฟน์จูนตั้งแต่ต้นจนจบ](../../phases/19-capstone-projects/07-end-to-end-fine-tuning-pipeline/) | P2 P3 P7 P10 P11 P17 P18 | Python |
| 08 | [แชตบอต RAG สำหรับงานจริงในอุตสาหกรรมกำกับดูแล](../../phases/19-capstone-projects/08-production-rag-chatbot/) | P5 P7 P11 P12 P17 P18 | Python |
| 09 | [เอเจนต์ย้ายโค้ดและอัปเกรดทั้งที่เก็บ](../../phases/19-capstone-projects/09-code-migration-agent/) | P5 P7 P11 P13 P14 P15 P17 | Python |
| 10 | [ทีมวิศวกรรมซอฟต์แวร์หลายเอเจนต์](../../phases/19-capstone-projects/10-multi-agent-software-team/) | P11 P13 P14 P15 P16 P17 | Python |
| 11 | [แดชบอร์ดสังเกตการณ์และประเมิน LLM](../../phases/19-capstone-projects/11-llm-observability-dashboard/) | P11 P13 P17 P18 | Python |
| 12 | [ไปป์ไลน์เข้าใจวิดีโอ (ฉาก → ถามตอบ)](../../phases/19-capstone-projects/12-video-understanding-pipeline/) | P4 P6 P7 P11 P12 P17 | Python |
| 13 | [เซิร์ฟเวอร์ MCP ไร้สถานะพร้อมรีจิสทรีและการกำกับดูแล](../../phases/19-capstone-projects/13-mcp-server-with-registry/) | P11 P13 P14 P17 P18 | Python |
| 14 | [เซิร์ฟเวอร์อนุมานด้วยการถอดรหัสเชิงคาดการณ์](../../phases/19-capstone-projects/14-speculative-decoding-server/) | P3 P7 P10 P17 | Python |
| 15 | [ชุดควบคุมความปลอดภัยตามหลักการและสนามทดสอบทีมโจมตี](../../phases/19-capstone-projects/15-constitutional-safety-harness/) | P10 P11 P13 P14 P18 | Python |
| 16 | [เอเจนต์อัตโนมัติจาก GitHub issue สู่ PR](../../phases/19-capstone-projects/16-github-issue-to-pr-agent/) | P11 P13 P14 P15 P17 | Python |
| 17 | [ผู้สอน AI ส่วนตัวที่ปรับตัวและรองรับหลายโมดัลลิตี](../../phases/19-capstone-projects/17-personal-ai-tutor/) | P5 P6 P11 P12 P14 P17 P18 | Python |
**เส้นทางลงมือสร้างเชิงลึก**: ชุดหลายบทเรียนที่สร้างระบบย่อยสมบูรณ์จากศูนย์
| # | โครงการ | รวมความรู้จาก | ภาษา |
|:---:|---------|----------|------|
| 20 | [สัญญาลูปของชุดควบคุมเอเจนต์](../../phases/19-capstone-projects/20-agent-harness-loop-contract/) | A. ชุดควบคุมเอเจนต์ | Python |
| 21 | [รีจิสทรีเครื่องมือพร้อมตรวจสคีมา](../../phases/19-capstone-projects/21-tool-registry-schema-validation/) | A. ชุดควบคุมเอเจนต์ | Python |
| 22 | [JSON-RPC 2.0 ผ่าน stdio ที่คั่นด้วยบรรทัดใหม่](../../phases/19-capstone-projects/22-jsonrpc-stdio-transport/) | A. ชุดควบคุมเอเจนต์ | Python |
| 23 | [ตัวกระจายคำเรียกฟังก์ชัน](../../phases/19-capstone-projects/23-function-call-dispatcher/) | A. ชุดควบคุมเอเจนต์ | Python |
| 24 | [กระแสควบคุมวางแผน-ลงมือ](../../phases/19-capstone-projects/24-plan-execute-control-flow/) | A. ชุดควบคุมเอเจนต์ | Python |
| 25 | [จุดตรวจยืนยันและงบการสังเกต](../../phases/19-capstone-projects/25-verification-gates-observation-budget/) | A. ชุดควบคุมเอเจนต์ | Python |
| 26 | [ตัวรันแซนด์บ็อกซ์พร้อมรายการห้ามและขอบเขตพาธ](../../phases/19-capstone-projects/26-sandbox-runner-denylist/) | A. ชุดควบคุมเอเจนต์ | Python |
| 27 | [ชุดประเมินพร้อมงานทดสอบคงที่](../../phases/19-capstone-projects/27-eval-harness-fixture-tasks/) | A. ชุดควบคุมเอเจนต์ | Python |
| 28 | [สังเกตการณ์ด้วย OTel GenAI spans และตัวชี้วัด Prometheus](../../phases/19-capstone-projects/28-observability-otel-traces/) | A. ชุดควบคุมเอเจนต์ | Python |
| 29 | [เอเจนต์เขียนโค้ดครบวงจรบนชุดควบคุม](../../phases/19-capstone-projects/29-end-to-end-coding-task-demo/) | A. ชุดควบคุมเอเจนต์ | Python |
| 30 | [ตัวตัดโทเคน BPE จากศูนย์](../../phases/19-capstone-projects/30-bpe-tokenizer-from-scratch/) | B. NLP LLM | Python |
| 31 | [ชุดข้อมูลโทเคนพร้อมหน้าต่างเลื่อน](../../phases/19-capstone-projects/31-tokenized-dataset-sliding-window/) | B. NLP LLM | Python |
| 32 | [เอ็มเบดดิงโทเคนและตำแหน่ง](../../phases/19-capstone-projects/32-token-positional-embeddings/) | B. NLP LLM | Python |
| 33 | [Self-attention หลายหัว](../../phases/19-capstone-projects/33-multihead-self-attention/) | B. NLP LLM | Python |
| 34 | [บล็อก Transformer จากศูนย์](../../phases/19-capstone-projects/34-transformer-block/) | B. NLP LLM | Python |
| 35 | [ประกอบโมเดล GPT](../../phases/19-capstone-projects/35-gpt-model-assembly/) | B. NLP LLM | Python |
| 36 | [ลูปฝึกและการประเมิน](../../phases/19-capstone-projects/36-training-loop-eval/) | B. NLP LLM | Python |
| 37 | [โหลดน้ำหนักที่พรีเทรนแล้ว](../../phases/19-capstone-projects/37-loading-pretrained-weights/) | B. NLP LLM | Python |
| 38 | [ไฟน์จูนตัวจำแนกด้วยการเปลี่ยนหัว](../../phases/19-capstone-projects/38-classifier-finetuning/) | B. NLP LLM | Python |
| 39 | [ปรับตามคำสั่งด้วยไฟน์จูนแบบมีผู้สอน](../../phases/19-capstone-projects/39-instruction-tuning-sft/) | B. NLP LLM | Python |
| 40 | [การปรับตามความชอบโดยตรงจากศูนย์](../../phases/19-capstone-projects/40-dpo-from-scratch/) | B. NLP LLM | Python |
| 41 | [ไปป์ไลน์ประเมินแบบเต็ม](../../phases/19-capstone-projects/41-eval-pipeline/) | B. NLP LLM | Python |
| 42 | [ตัวดาวน์โหลดคลังข้อความขนาดใหญ่](../../phases/19-capstone-projects/42-large-corpus-downloader/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 43 | [คลังข้อความโทเคน HDF5](../../phases/19-capstone-projects/43-hdf5-tokenized-corpus/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 44 | [อัตราการเรียนรู้แบบโคไซน์พร้อมวอร์มอัปเชิงเส้น](../../phases/19-capstone-projects/44-cosine-lr-warmup/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 45 | [ตัดเกรเดียนต์และใช้ความแม่นยำแบบผสม](../../phases/19-capstone-projects/45-gradient-clipping-amp/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 46 | [สะสมเกรเดียนต์](../../phases/19-capstone-projects/46-gradient-accumulation/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 47 | [บันทึกเช็กพอยต์และกลับมาทำต่อ](../../phases/19-capstone-projects/47-checkpoint-save-resume/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 48 | [Distributed Data Parallel และ FSDP จากศูนย์](../../phases/19-capstone-projects/48-distributed-fsdp-ddp/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 49 | [ชุดประเมินโมเดลภาษา](../../phases/19-capstone-projects/49-lm-eval-harness/) | C. ฝึกตั้งแต่ต้นจนจบ | Python |
| 50 | [ตัวสร้างสมมติฐาน](../../phases/19-capstone-projects/50-hypothesis-generator/) | D. วิจัยอัตโนมัติ | Python |
| 51 | [ค้นคืนวรรณกรรมวิจัย](../../phases/19-capstone-projects/51-literature-retrieval/) | D. วิจัยอัตโนมัติ | Python |
| 52 | [ตัวรันการทดลอง](../../phases/19-capstone-projects/52-experiment-runner/) | D. วิจัยอัตโนมัติ | Python |
| 53 | [ตัวประเมินผลลัพธ์](../../phases/19-capstone-projects/53-result-evaluator/) | D. วิจัยอัตโนมัติ | Python |
| 54 | [ตัวเขียนบทความวิจัย](../../phases/19-capstone-projects/54-paper-writer/) | D. วิจัยอัตโนมัติ | Python |
| 55 | [ลูปวิจารณ์](../../phases/19-capstone-projects/55-critic-loop/) | D. วิจัยอัตโนมัติ | Python |
| 56 | [ตัวจัดตารางรอบทำงาน](../../phases/19-capstone-projects/56-iteration-scheduler/) | D. วิจัยอัตโนมัติ | Python |
| 57 | [เดโมวิจัยตั้งแต่ต้นจนจบ](../../phases/19-capstone-projects/57-end-to-end-research-demo/) | D. วิจัยอัตโนมัติ | Python |
| 58 | [แพตช์ของตัวเข้ารหัสภาพ](../../phases/19-capstone-projects/58-vision-encoder-patches/) | E. VLM หลายโมดัลลิตี | Python |
| 59 | [ตัวเข้ารหัส Vision Transformer](../../phases/19-capstone-projects/59-vit-transformer/) | E. VLM หลายโมดัลลิตี | Python |
| 60 | [ชั้นฉายสำหรับจัดแนวโมดัลลิตี](../../phases/19-capstone-projects/60-projection-layer-modality-align/) | E. VLM หลายโมดัลลิตี | Python |
| 61 | [รวมข้อมูลด้วย cross-attention](../../phases/19-capstone-projects/61-cross-attention-fusion/) | E. VLM หลายโมดัลลิตี | Python |
| 62 | [พรีเทรนภาพและภาษา](../../phases/19-capstone-projects/62-vision-language-pretraining/) | E. VLM หลายโมดัลลิตี | Python |
| 63 | [การประเมินหลายโมดัลลิตี](../../phases/19-capstone-projects/63-multimodal-eval/) | E. VLM หลายโมดัลลิตี | Python |
| 64 | [เปรียบเทียบกลยุทธ์แบ่งข้อความ](../../phases/19-capstone-projects/64-chunking-strategies-advanced/) | F. RAG ขั้นสูง | Python |
| 65 | [ค้นคืนแบบผสมด้วย BM25 และเอ็มเบดดิงหนาแน่น](../../phases/19-capstone-projects/65-hybrid-retrieval-bm25-dense/) | F. RAG ขั้นสูง | Python |
| 66 | [จัดอันดับใหม่ด้วย cross-encoder](../../phases/19-capstone-projects/66-reranker-cross-encoder/) | F. RAG ขั้นสูง | Python |
| 67 | [เขียนคำค้นใหม่: HyDE, Multi-Query และการแยกส่วน](../../phases/19-capstone-projects/67-query-rewriting-hyde/) | F. RAG ขั้นสูง | Python |
| 68 | [ประเมิน RAG: precision, recall, MRR, nDCG, ความตรงต่อแหล่งข้อมูล และความเกี่ยวข้องของคำตอบ](../../phases/19-capstone-projects/68-rag-eval-precision-recall/) | F. RAG ขั้นสูง | Python |
| 69 | [ระบบ RAG ตั้งแต่ต้นจนจบ](../../phases/19-capstone-projects/69-end-to-end-rag-system/) | F. RAG ขั้นสูง | Python |
| 70 | [รูปแบบข้อกำหนดงาน](../../phases/19-capstone-projects/70-task-spec-format/) | G. กรอบประเมิน | Python |
| 71 | [ตัวชี้วัดแบบดั้งเดิม](../../phases/19-capstone-projects/71-classical-metrics/) | G. กรอบประเมิน | Python |
| 72 | [ตัวชี้วัดการรันโค้ด](../../phases/19-capstone-projects/72-code-exec-metric/) | G. กรอบประเมิน | Python |
| 73 | [Perplexity และการสอบเทียบ](../../phases/19-capstone-projects/73-perplexity-calibration/) | G. กรอบประเมิน | Python |
| 74 | [รวมตารางจัดอันดับ](../../phases/19-capstone-projects/74-leaderboard-aggregation/) | G. กรอบประเมิน | Python |
| 75 | [ตัวรันประเมินตั้งแต่ต้นจนจบ](../../phases/19-capstone-projects/75-end-to-end-eval-runner/) | G. กรอบประเมิน | Python |
| 76 | [การดำเนินการรวมหมู่จากศูนย์](../../phases/19-capstone-projects/76-collective-ops-from-scratch/) | H. ฝึกแบบกระจาย | Python |
| 77 | [Data Parallel DDP จากศูนย์](../../phases/19-capstone-projects/77-data-parallel-ddp/) | H. ฝึกแบบกระจาย | Python |
| 78 | [แบ่งสถานะตัวปรับค่าด้วย ZeRO](../../phases/19-capstone-projects/78-zero-parameter-sharding/) | H. ฝึกแบบกระจาย | Python |
| 79 | [ไปป์ไลน์ขนานและวิเคราะห์ช่วงรอ](../../phases/19-capstone-projects/79-pipeline-parallel/) | H. ฝึกแบบกระจาย | Python |
| 80 | [เช็กพอยต์แบบแบ่งส่วนและกลับมาทำต่อแบบอะตอมมิก](../../phases/19-capstone-projects/80-checkpoint-sharded-resume/) | H. ฝึกแบบกระจาย | Python |
| 81 | [ฝึกแบบกระจายตั้งแต่ต้นจนจบ](../../phases/19-capstone-projects/81-end-to-end-distributed-train/) | H. ฝึกแบบกระจาย | Python |
| 82 | [การจัดหมวดหมู่เจลเบรก](../../phases/19-capstone-projects/82-jailbreak-taxonomy/) | I. ชุดควบคุมความปลอดภัย | Python |
| 83 | [ตัวตรวจจับการแทรกพรอมป์ต์](../../phases/19-capstone-projects/83-prompt-injection-detector/) | I. ชุดควบคุมความปลอดภัย | Python |
| 84 | [การประเมินการปฏิเสธ](../../phases/19-capstone-projects/84-refusal-evaluation/) | I. ชุดควบคุมความปลอดภัย | Python |
| 85 | [เชื่อมต่อตัวจำแนกเนื้อหา](../../phases/19-capstone-projects/85-content-classifier-integration/) | I. ชุดควบคุมความปลอดภัย | Python |
| 86 | [เอนจินกฎตามหลักการ](../../phases/19-capstone-projects/86-constitutional-rules-engine/) | I. ชุดควบคุมความปลอดภัย | Python, YAML |
| 87 | [ด่านความปลอดภัยตั้งแต่ต้นจนจบ](../../phases/19-capstone-projects/87-end-to-end-safety-gate/) | I. ชุดควบคุมความปลอดภัย | Python |
## หนังสือและการรับรอง
อ่านหลักสูตรแกนหลักในรูปแบบหนังสือ
หลักสูตรแกนหลัก 20 ระยะภายใต้ `phases/` ถูกรวมเป็นหนังสือหกเล่ม CI สร้าง EPUB และ PDF จากแหล่งบทเรียนเดียวกัน แล้วแนบใน[การเผยแพร่บน GitHub](https://github.com/rohitg00/ai-engineering-from-scratch/releases) ทุกครั้ง ลิงก์ด้านล่างชี้ไปยังรุ่นล่าสุดเสมอ เลขเล่มบอกลำดับในชุด ไม่ใช่เวอร์ชัน แต่ละไฟล์มีวันที่ของฉบับ และฉบับเก่ายังคงดาวน์โหลดได้จากหน้าการเผยแพร่ของตน
หลักสูตรใบรับรองตั้งใจไม่รวมในหนังสือ สถานะผู้สอน AI แล็บที่รันได้ ภาพโต้ตอบ แบบประเมินพื้นฐาน และข้อสอบจำลองแบบจับเวลา ยังคงใช้งานเต็มรูปแบบบน GitHub และเว็บไซต์
| เล่ม | ชื่อ | ระยะ | ดาวน์โหลด |
|-----|-------|--------|----------|
| 1 | พื้นฐาน · คณิตศาสตร์ เครื่องมือ และแมชชีนเลิร์นนิงแบบดั้งเดิม | 00-02 | [EPUB](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol1-foundations.epub) · [PDF](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol1-foundations.pdf) |
| 2 | ดีปเลิร์นนิง · โครงข่าย การมองเห็น และเสียงพูด | 03, 04, 06 | [EPUB](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol2-deep-learning.epub) · [PDF](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol2-deep-learning.pdf) |
| 3 | ภาษา · พื้นฐาน NLP และ Transformer | 05, 07 | [EPUB](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol3-language.epub) · [PDF](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol3-language.pdf) |
| 4 | โมเดลภาษาขนาดใหญ่ · การสร้าง การเสริมกำลัง พรีเทรน และวิศวกรรม | 08-11 | [EPUB](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol4-llms.epub) · [PDF](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol4-llms.pdf) |
| 5 | เอเจนต์ · หลายโมดัลลิตี โพรโทคอล การทำงานอัตโนมัติ และฝูง | 12-16 | [EPUB](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol5-agents.epub) · [PDF](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol5-agents.pdf) |
| 6 | งานจริง · โครงสร้างพื้นฐาน ความปลอดภัย และโครงงานสรุป | 17-19 | [EPUB](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol6-production.epub) · [PDF](https://github.com/rohitg00/ai-engineering-from-scratch/releases/latest/download/aiefs-vol6-production.pdf) |
หนังสือคือภาพของหลักสูตร ณ เวลาหนึ่ง ส่วนที่เก็บนี้คือฉบับที่ปรับปรุงต่อเนื่อง ทุกบทจบด้วยลิงก์ไปยังภาพเคลื่อนไหว แบบทดสอบ และโค้ดที่รันได้ สร้างในเครื่องด้วย `python3 scripts/build_book.py` โดยต้องมี pandoc ดูรายละเอียดกระบวนการใน [book/README.md](../../book/README.md)
เตรียมสอบใบรับรอง Claude
[Claude Certification Academy](../../certifications/claude/README.md) เป็นหลักสูตรเตรียมสอบฟรีและโอเพนซอร์สสำหรับใบรับรอง Claude ทางการทั้งสี่เส้นทาง: Associate Foundations, Developer Foundations, Architect Foundations และ Architect Professional แต่ละเส้นทางรวมบทเรียนที่ตรงกับแผนผังข้อสอบ แล็บที่รันได้ แบบประเมินพื้นฐาน โครงงานสรุป และข้อสอบฝึกทำต้นฉบับแบบเต็มชุด
ใช้[คู่มือเริ่มต้นบน GitHub โดยมี AI ช่วย](../../certifications/claude/GETTING_STARTED.md) กับ Claude Code, Codex, ChatGPT, Cursor หรือเอเจนต์อื่น เรียก `claude-certification` ใน Codex, `/claude-certification` ใน Claude Code หรือขอให้โฮสต์อื่นใช้ `claude-certification` ระบบจะเลือกเส้นทาง สร้างแผนต่อเนื่องใน `CLAUDE-CERTIFICATION.md` สอนทีละขั้น รันแล็บจริง และให้ข้อเสนอแนะจากชิ้นงาน หลักสูตรเดียวกันยังอยู่บน[เว็บไซต์ใบรับรอง](https://aiengineeringfromscratch.com/certifications.html)
สถาบันนี้เป็นสื่อเรียนอิสระตามวัตถุประสงค์การสอบที่เปิดเผย ไม่ได้มีความเกี่ยวข้องกับ Anthropic ไม่คัดลอกข้อสอบจริง และไม่รับประกันว่าจะสอบผ่าน
เตรียมสอบใบรับรอง MCP Associate (MCPA)
[หลักสูตรใบรับรอง MCPA](../../certifications/mcpa/README.md) เป็นหลักสูตรเตรียมสอบฟรีและโอเพนซอร์สสำหรับ Model Context Protocol Associate ของ Agentic AI Foundation ซึ่งจัดผ่าน Linux Foundation Training ทั้ง 34 บทเรียนสอนโพรโทคอลไร้สถานะรุ่น 2026-07-28 ตามห้าหมวดสอบ ได้แก่ `_meta` ต่อคำขอและ `server/discover` แทนการจับมือแบบเดิม คำขอหลายรอบ การสมัครรับข้อมูล แคช ส่วนขยาย Tasks และ MCP Apps การอนุญาตผ่าน OAuth ตลอดจนชั้นรีจิสทรีและ SDK ทุกบทเรียนมีแล็บที่ใช้เพียงไลบรารีมาตรฐาน และตรวจบันทึกการรันกับรูปแบบข้อความปัจจุบัน นอกจากนี้ยังมีแบบประเมินพื้นฐาน โครงงานสรุป และข้อสอบฝึกทำต้นฉบับเต็มชุดสามชุด โดยสัดส่วนคำถามตรงตามน้ำหนักในแผนผังข้อสอบที่เผยแพร่
ใช้[คู่มือเริ่มต้นบน GitHub โดยมี AI ช่วย](../../certifications/mcpa/GETTING_STARTED.md) กับ Claude Code, Codex, ChatGPT, Cursor หรือเอเจนต์อื่น เรียก `mcpa-certification` ใน Codex, `/mcpa-certification` ใน Claude Code หรือขอให้โฮสต์อื่นใช้ `mcpa-certification` ระบบจะสร้างแผนต่อเนื่องใน `MCPA-CERTIFICATION.md` สอนทีละขั้น รันแล็บจริง และให้ข้อเสนอแนะจากชิ้นงาน หลักสูตรเดียวกันอยู่บน[หน้าเส้นทาง MCPA](https://aiengineeringfromscratch.com/certification?id=mcpa-f)
หลักสูตรนี้เป็นสื่อเรียนอิสระตามวัตถุประสงค์การสอบที่เปิดเผย ไม่ได้เกี่ยวข้องกับ Agentic AI Foundation หรือ Linux Foundation ไม่คัดลอกข้อสอบจริง และไม่รับประกันว่าจะสอบผ่าน
## ชุดเครื่องมือ
ทุกบทเรียนสร้างผลงานที่ใช้ซ้ำได้ ติดตั้งลงในเอเจนต์ของคุณหรือใช้สคริปต์ด้านล่างจากไดเรกทอรีรากของที่เก็บโค้ด
โครงสร้างบทเรียนและผลงานที่ใช้ซ้ำได้
## โครงสร้างบทเรียน
แต่ละบทเรียนมีโฟลเดอร์ของตัวเอง และใช้โครงสร้างเดียวกันตลอดหลักสูตร:
```text
phases/-/-/
├── code/ โค้ดที่รันได้ (Python, TypeScript, Rust, Julia)
├── docs/
│ └── en.md เนื้อหาบทเรียน
└── outputs/ พรอมป์ต์ สกิล เอเจนต์ หรือเซิร์ฟเวอร์ MCP ที่บทเรียนสร้าง
```
ทุกบทเรียนมีหกช่วง การแยก *สร้างเอง / ใช้งาน* เป็นแกนหลัก: คุณเขียนอัลกอริทึมจากศูนย์ก่อน แล้วทำสิ่งเดียวกันด้วยไลบรารีสำหรับงานจริง คุณจึงเข้าใจว่าเฟรมเวิร์กทำอะไร เพราะได้เขียนรุ่นเล็กด้วยตัวเองแล้ว
```mermaid
%%{init: {'theme':'base','themeVariables':{'primaryColor':'#fafaf5','primaryTextColor':'#1a1a1a','primaryBorderColor':'#3553ff','lineColor':'#3553ff','fontFamily':'JetBrains Mono','fontSize':'13px'}}}%%
flowchart LR
M["ใจความหลัก
แนวคิดสำคัญในหนึ่งบรรทัด"] --> Pr["ปัญหา
ความติดขัดที่เป็นรูปธรรม"]
Pr --> C["แนวคิด
แผนภาพและความเข้าใจ"]
C --> B["สร้างเอง
คณิตศาสตร์พื้นฐาน ไม่ใช้เฟรมเวิร์ก"]
B --> U["ใช้งาน
ทำสิ่งเดียวกันใน PyTorch / sklearn"]
U --> S["เผยแพร่
พรอมป์ต์ · สกิล · เอเจนต์ · MCP"]
```
## ทุกบทเรียนมีชิ้นงาน
หลักสูตรอื่นจบด้วย *“ยินดีด้วย คุณเรียนรู้ X แล้ว”* แต่ทุกบทเรียนที่นี่จบด้วย**เครื่องมือที่ใช้ซ้ำได้** ซึ่งคุณติดตั้งหรือวางในกระบวนการทำงานประจำวันได้
 FIG_001 · A พรอมป์ต์ |
 FIG_001 · B สกิล |
 FIG_001 · C เอเจนต์ |
 FIG_001 · D เซิร์ฟเวอร์ MCP |
| วางในผู้ช่วย AI ใดก็ได้เพื่อขอความช่วยเหลือระดับผู้เชี่ยวชาญในงานเฉพาะด้าน |
ติดตั้งใน Claude, Cursor, Codex, OpenClaw, Hermes หรือเอเจนต์ใดที่อ่าน SKILL.md. |
ปรับใช้เป็นผู้ปฏิบัติงานอัตโนมัติ คุณเขียนลูปด้วยตัวเองในระยะ 14 แล้ว |
เชื่อมต่อกับไคลเอนต์ที่รองรับ MCP สร้างครบตั้งแต่ต้นจนจบในระยะ 13 |
ติดตั้งผลงานจากบทเรียน
**ชิ้นงานจากบทเรียน.** ที่เก็บมี 396 สกิลและ 99 พรอมป์ต์ภายใต้ `phases/**/outputs/` ติดตั้งด้วย `scripts/install_skills.py` โดยต้องโคลนที่เก็บก่อน รองรับการกรองแท็ก ทดลองโดยไม่เขียนไฟล์ และรูปแบบโฟลเดอร์ตามแต่ละเอเจนต์:
```bash
python3 scripts/install_skills.py # every skill, default --layout skills (nested)
python3 scripts/install_skills.py --layout skills # same as above, explicit
python3 scripts/install_skills.py --type all # skills + prompts + agents
python3 scripts/install_skills.py --phase 14 # one phase only
python3 scripts/install_skills.py --tag rag # filter by tag
python3 scripts/install_skills.py --layout flat # flat files
python3 scripts/install_skills.py --dry-run # preview without writing
python3 scripts/install_skills.py --force # overwrite existing files
```
`` คือไดเรกทอรีสกิลของเอเจนต์ เช่น `~/.claude/skills/`, `~/.cursor/skills/`, `~/.config/openclaw/skills/`, `.skills/` หรือพาธใดที่เอเจนต์อ่าน
โดยปริยาย สคริปต์ไม่ยอมเขียนทับปลายทางที่มีอยู่ และจบด้วยรหัส 1 หลังแสดงพาธที่ชนกันทั้งหมด ใช้ `--dry-run` เพื่อดูการชนกันล่วงหน้า หรือ `--force` เพื่อเขียนทับ ทุกการรันที่เขียนไฟล์จริงจะสร้าง `manifest.json` ในปลายทาง พร้อมรายการทั้งหมดแบ่งตามประเภทและระยะ เลือกรูปแบบที่เอเจนต์ของคุณอ่านได้:
| `--layout` | พาธที่สร้าง |
|---|---|
| `skills` | `//SKILL.md` (รูปแบบซ้อนโฟลเดอร์ รองรับโดย Claude / Cursor / Codex / OpenClaw / Hermes) |
| `by-phase` | `/phase-NN/.md` |
| `flat` | `/.md` |
เพิ่มเวิร์กเบนช์เอเจนต์ในที่เก็บของคุณ
โครงงานสรุประยะ 14 มีชุด Agent Workbench ที่ใช้ซ้ำได้ ประกอบด้วย AGENTS.md, สคีมา และสคริปต์เริ่มต้น / ตรวจสอบ / ส่งต่องาน สร้างโครงในที่เก็บใดก็ได้ด้วย:
```bash
python3 scripts/scaffold_workbench.py path/to/your-repo # full pack + seeds
python3 scripts/scaffold_workbench.py path/to/your-repo --minimal # skip docs/
python3 scripts/scaffold_workbench.py path/to/your-repo --dry-run # preview only
python3 scripts/scaffold_workbench.py path/to/your-repo --force # overwrite
```
คุณจะได้ส่วนทำงานทั้งเจ็ดของเวิร์กเบนช์ที่เชื่อมต่อแล้ว `task_board.json` เริ่มต้น และ `agent_state.json` ใหม่ที่มี `schema_version: 1` จากนั้นแก้งานและ `AGENTS.md` รัน `scripts/init_agent.py` แล้วส่งสัญญางานให้เอเจนต์ แหล่งชุดเครื่องมืออยู่ที่ `phases/14-agent-engineering/42-agent-workbench-capstone/outputs/agent-workbench-pack/`
ดูหลักสูตรทั้งหมดในรูปแบบ JSON
`scripts/build_catalog.py` อ่านทุกระยะ ทุกบทเรียน และทุกชิ้นงานบนดิสก์ แล้วเขียน `catalog.json` ที่รากของที่เก็บ รวมข้อมูลจริงของหลักสูตรไว้ในไฟล์เดียว
```bash
python3 scripts/build_catalog.py # writes /catalog.json
python3 scripts/build_catalog.py --stdout # to stdout, do not touch repo
python3 scripts/build_catalog.py --out path/to/file.json
```
แค็ตตาล็อกสร้างจากระบบไฟล์ ไม่ใช่ README จำนวนจึงตรงกับสิ่งที่อยู่บนดิสก์เสมอ ใช้สร้างเว็บไซต์ ต่อเครื่องมืออื่น หรือตรวจว่าจำนวนใน README ยังถูกต้อง สคีมามีคำอธิบายที่ส่วนต้นของสคริปต์
เวิร์กโฟลว์หลักสูตรสร้าง `catalog.json` เป็นไฟล์ชั่วคราวที่ Git ละเว้น อย่าคอมมิตไฟล์นี้ เวิร์กโฟลว์เดียวกันรัน `audit_lessons.py` เป็นการตรวจสอบที่ต้องผ่าน
ตรวจโค้ด Python ของทุกบทเรียนอย่างรวดเร็ว
`scripts/lesson_run.py` คอมไพล์ไฟล์ `.py` ทุกไฟล์ในไดเรกทอรี `code/` ของแต่ละบทเรียนเป็นไบต์โค้ด โหมดปกติตรวจไวยากรณ์เท่านั้น ไม่รันโค้ด ไม่ใช้ API key และไม่ต้องติดตั้งไลบรารี ML ขนาดใหญ่ ช่วยจับข้อผิดพลาดที่พบบ่อย เช่น ย่อหน้าผิด f-string เสีย และการแก้ไขโดยไม่ตั้งใจ
```bash
python3 scripts/lesson_run.py # syntax-check the whole curriculum
python3 scripts/lesson_run.py --phase 14 # one phase only
python3 scripts/lesson_run.py --json # JSON report on stdout
python3 scripts/lesson_run.py --strict # exit 1 if any lesson fails
python3 scripts/lesson_run.py --execute # actually run, 10s timeout per lesson
```
`--execute` รัน `code/main.py` ของแต่ละบทเรียน หรือไฟล์ `.py` แรก โดยจำกัดเวลา 10 วินาที หากไฟล์เริ่มต้นมีความเห็น `# requires: pkg1, pkg2` ระบุไลบรารีนอกชุดมาตรฐาน บทเรียนนั้นจะถูกข้ามด้วยเหตุผล `needs ` ต้องเลือกใช้โหมดนี้เอง และไม่ได้ต่อเข้ากับ CI
ใช้เพียงไลบรารีมาตรฐานกับ Python 3.10 ขึ้นไป ตั้ง `LINK_CHECK_SKIP=domain1,domain2` เพื่อแทนรายการโดเมนที่ข้ามโดยปริยาย ได้แก่ `twitter.com`, `x.com`, `linkedin.com`, `instagram.com`, `medium.com` ซึ่งมักบล็อกคำขอ HEAD/GET อัตโนมัติ
งานวิจัยและโปรโตคอลพื้นฐาน
- *Attention Is All You Need* — Vaswani et al., 2017 → [ระยะ 7](#phase-7)
- *Language Models are Few-Shot Learners* (GPT-3) → [ระยะ 10](#phase-10)
- *Denoising Diffusion Probabilistic Models* → [ระยะ 8](#phase-8)
- *InstructGPT / RLHF* → [ระยะ 10](#phase-10)
- *Direct Preference Optimization* → [ระยะ 10](#phase-10)
- *Chain-of-Thought Prompting* → [ระยะ 11](#phase-11)
- *ReAct: Reasoning + Acting in LLMs* → [ระยะ 14](#phase-14)
- *Model Context Protocol* — Anthropic → [ระยะ 13](#phase-13)
## ร่วมพัฒนา
| เป้าหมาย | อ่าน |
|---|---|
| เพิ่มบทเรียนหรือแก้ไขข้อผิดพลาด | [CONTRIBUTING.md](../../CONTRIBUTING.md) |
| ฟอร์กให้ทีมหรือโรงเรียนของคุณ | [FORKING.md](../../FORKING.md) |
| แม่แบบบทเรียน | [LESSON_TEMPLATE.md](../../LESSON_TEMPLATE.md) |
| ติดตามความคืบหน้า | [ROADMAP.md](../../ROADMAP.md) |
| อภิธานศัพท์ | [glossary/terms.md](../../glossary/terms.md) |
| จรรยาบรรณชุมชน | [CODE_OF_CONDUCT.md](../../CODE_OF_CONDUCT.md) |
ก่อนส่งบทเรียน ให้รันการตรวจเงื่อนไขที่ต้องคงไว้:
```bash
python3 scripts/audit_lessons.py # full curriculum
python3 scripts/audit_lessons.py --phase 14 # single phase
python3 scripts/audit_lessons.py --json # CI-friendly output
```
หากกฎข้อใดไม่ผ่าน รหัสจบโปรแกรมจะไม่เป็นศูนย์ กฎ (L001–L010) ตรวจโครงสร้างไดเรกทอรี การมี `docs/en.md` และหัวข้อ H1 การมีเนื้อหาใน `code/` สคีมาของ `quiz.json` ซึ่งปฏิเสธคีย์เก่า `q/choices/answer` ที่ทำให้เกิดปัญหา #102 และลิงก์สัมพัทธ์ภายในเอกสารบทเรียน
## สนับสนุนงานนี้
114,584 ผู้อ่าน · 181,995 การดูหน้าใน 30 วันที่ผ่านมา · ข้อมูล ณ 2026-08-29
ฟรี ภายใต้สัญญาอนุญาต MIT จำนวน 523 บทเรียน ขอบคุณผู้สนับสนุนและผู้ให้ทุนที่ทำให้งานนี้เป็นไปได้ [ดูผู้สนับสนุนและผู้ให้ทุนทั้งหมด](../../BACKERS.md)
ต้องการสนับสนุนงานนี้? ดู[ตัวเลือกการสนับสนุน](../../SPONSORS.md) รวมถึง[การสนับสนุนฮาร์ดแวร์](../../SPONSORS.md#hardware-lab-partner) หรือ[สนับสนุนผ่าน GitHub](https://github.com/sponsors/rohitg00)
หากคู่มือนี้ช่วยคุณได้ ให้ดาวกับที่เก็บโค้ด เพื่อช่วยให้โครงการดำเนินต่อไป
## สัญญาอนุญาต
MIT ใช้ได้ตามต้องการ: ฟอร์ก สอน ขาย หรือเผยแพร่ ขอบคุณหากให้เครดิต แต่ไม่ได้บังคับ
ดูแลโดย [Rohit Ghumare](https://github.com/rohitg00) และชุมชน
@ghumare64 · aiengineeringfromscratch.com · รายงาน / เสนอแนะ