数据加载与文本分块基础
学习加载非结构化数据，并采用有效的文本分块策略，以获得最佳检索性能。
数据加载与文本分块基础 是 CoddyKit 上的免费 LLM Apps in Production (RAG + Vector DB + Caching) 课时。 这是第 2 节课，共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 LLM Apps in Production (RAG + Vector DB + Caching) 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 LLM Apps in Production (RAG + Vector DB + Caching) 课程共包含 4 节课。
本课时的部分内容尚未翻译，以英文显示。
Loading Data for RAG
Welcome to Lesson 2! In Retrieval Augmented Generation (RAG), the first step is always to get your data ready. This means loading your information and preparing it for the Large Language Model (LLM).
Most real-world data is unstructured, meaning it doesn't fit neatly into rows and columns like a spreadsheet. Think of documents, web pages, or books.
Common Unstructured Data Sources
RAG systems can work with many types of unstructured data. Here are some common examples:
- Text files (.txt): Simple, plain text documents.
- PDFs (.pdf): Often contain text, images, and complex layouts.
- Word Documents (.docx): Rich text with formatting.
- Web Pages (.html): Content from websites.
- Databases/APIs: Text extracted from various fields.
The goal is to extract the raw text content from these sources.
Basic Text File Loading
Let's start with the simplest form: loading a plain text file. In Python, you can easily read the entire content of a file into a string.
This example creates a small sample.txt and then reads its content.
import os
def load_text_file(filepath):
with open(filepath, 'r', encoding='utf-8') as f:
return f.read()
if __name__ == "__main__":
# Create a dummy file for demonstration
file_content = "This is the first line.\nThis is the second line.\nAnd a final line of text."
with open("sample.txt", "w", encoding="utf-8") as f:
f.write(file_content)
# Load and print the content
loaded_data = load_text_file("sample.txt")
print("--- Loaded Content ---")
print(loaded_data)
# Clean up the dummy file
os.remove("sample.txt")Why Text Chunking is Essential
Once you've loaded your data, you can't usually send an entire book or long document directly to an LLM. Why not?
- Context Window Limits: LLMs have a maximum amount of text they can process at once.
- Cost: Sending very long texts is expensive, as you're typically charged per token.
- Relevance: Shorter, focused pieces of text are often more relevant for retrieval.
This is where text chunking comes in.
Understanding the Context Window
The context window is like an LLM's short-term memory. It's the maximum number of tokens (words or sub-words) it can consider when generating a response.
- If your input text is too long, it gets truncated.
- The LLM only 'sees' what's in its context window.
Chunking breaks your big document into smaller, manageable pieces that fit within this window.
Basic Chunking: Fixed Size
The simplest chunking strategy is fixed-size chunking. You define a specific number of characters or tokens, and then split your document into chunks of that exact size.
For example, if you have a 1000-character document and a chunk size of 100, you'll get 10 chunks.
- Pros: Easy to implement.
- Cons: Can cut sentences or paragraphs in half, losing context.
Fixed-Size Chunking in Action
Here's a Python example demonstrating fixed-size chunking. Notice how the text is simply cut at regular intervals, which might sometimes break words or sentences.
def fixed_size_chunker(text, chunk_size):
chunks = []
for i in range(0, len(text), chunk_size):
chunks.append(text[i : i + chunk_size])
return chunks
if __name__ == "__main__":
sample_text = "Large language models are powerful tools for text generation and understanding. However, they have limitations, especially with very long inputs due to their context window size."
chunk_size = 40
chunks = fixed_size_chunker(sample_text, chunk_size)
print(f"Original text length: {len(sample_text)}")
print(f"Chunk size: {chunk_size}")
print("--- Chunks ---")
for i, chunk in enumerate(chunks):
print(f"Chunk {i+1} ({len(chunk)} chars): '{chunk}'")Improving Context with Overlap
Fixed-size chunking can be problematic if important context is split across two chunks. To mitigate this, we use overlapping chunks.
With overlap, each new chunk starts a bit before the previous one ended. This ensures that some text appears in multiple chunks, preserving continuity.
- A common overlap size is 10-20% of the chunk size.
- It helps the LLM connect ideas even if they span chunk boundaries.
Overlapping Chunking Example
See how adding an overlap helps maintain context. The start of each new chunk includes some text from the end of the previous one.
def overlapping_chunker(text, chunk_size, overlap_size):
chunks = []
start = 0
while start < len(text):
end = start + chunk_size
chunk = text[start:end]
chunks.append(chunk)
start += chunk_size - overlap_size
if start < 0: # Handle cases where overlap > chunk_size initially
start = 0
return chunks
if __name__ == "__main__":
text_data = "The quick brown fox jumps over the lazy dog. Dogs are mammals and often friendly animals."
chunk_size = 30
overlap_size = 10
chunks = overlapping_chunker(text_data, chunk_size, overlap_size)
print(f"Original text length: {len(text_data)}")
print(f"Chunk size: {chunk_size}, Overlap size: {overlap_size}")
print("--- Overlapping Chunks ---")
for i, chunk in enumerate(chunks):
print(f"Chunk {i+1} ({len(chunk)} chars): '{chunk}'")Check Your Understanding
You've learned about loading data and basic chunking strategies. Now, let's test your knowledge!
Recap: Data Loading & Chunking
Great job! In this lesson, we covered the foundational steps of preparing data for RAG applications:
- Data Loading: Extracting raw text from various unstructured sources like text files, PDFs, and web pages.
- Text Chunking: The essential process of breaking down long documents into smaller, manageable pieces.
- Context Window: Understanding the LLM's limitation on input text length (measured in tokens).
- Chunking Strategies: Explored basic fixed-size chunking and the improved method of fixed-size chunking with overlap to preserve context.
Next, we'll see how these chunks are used to build a simple RAG pipeline!
用 AI 导师学习 LLM Apps in Production (RAG + Vector DB + Caching) — 免费
在浏览器中编写并运行真实代码，获得全天候 AI 导师的即时帮助，并在网页或应用中继续学习。
- 课程
- 12
- 课程
- 48
常见问题解答
「数据加载与文本分块基础」课时是免费的吗？
是的 — 「数据加载与文本分块基础」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 LLM Apps in Production (RAG + Vector DB + Caching) 课程的其余内容，请升级到 CoddyKit PRO。 LLM Apps in Production (RAG + Vector DB + Caching) 课程共包含 4 节课。
「数据加载与文本分块基础」这节课中我会学到什么？
学习加载非结构化数据，并采用有效的文本分块策略，以获得最佳检索性能。 你通过在浏览器中直接运行的动手代码来练习 LLM Apps in Production (RAG + Vector DB + Caching)，全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 LLM Apps in Production (RAG + Vector DB + Caching) 需要有经验吗？
无需任何先前经验。CoddyKit 上的 LLM Apps in Production (RAG + Vector DB + Caching) 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。 这是第 2 节课，共 4 节。
「数据加载与文本分块基础」课时需要多长时间？
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。
我能在这节 LLM Apps in Production (RAG + Vector DB + Caching) 课中编写并运行代码吗？
能。每节 LLM Apps in Production (RAG + Vector DB + Caching) 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。
此课程中的所有课时
- 选择 LLM 提供商
- 数据加载与文本分块基础
- 构建简单的 RAG 流水线
- 测试与评估您的 RAG 应用