AutoRAG 0.3.8__tar.gz → 0.3.9__tar.gz

{autorag-0.3.8 → autorag-0.3.9}/AutoRAG.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: AutoRAG
-Version: 0.3.8
+Version: 0.3.9
 Summary: Automatically Evaluate RAG pipelines with your own data. Find optimal structure for new RAG product.
 Author-email: Marker-Inc <vkehfdl1@gmail.com>
 License:                                  Apache License
@@ -240,6 +240,7 @@ Requires-Dist: voyageai
 Requires-Dist: mixedbread-ai
 Requires-Dist: llama-index-llms-bedrock
 Requires-Dist: scikit-learn
+Requires-Dist: emoji
 Requires-Dist: pymilvus
 Requires-Dist: chromadb>=0.5.0
 Requires-Dist: quart
@@ -524,7 +525,7 @@ initial_qa.to_parquet('./qa.parquet', './corpus.parquet')
 
 # RAG Optimization
 
-<a href="https://huggingface.co/spaces/AutoRAG/AutoRAG-optimization">
+<a href="https://huggingface.co/spaces/AutoRAG/RAG-Pipeline-Optimization">
 <img src="https://github.com/user-attachments/assets/8c6e4b02-3938-4560-b817-c95764965b50" alt="Hugging Face Sticker" style="width:200px;height:auto;">
 </a>
 
@@ -818,3 +819,18 @@ We are developing AutoRAG as open-source.
 So this project welcomes contributions and suggestions. Feel free to contribute to this project.
 
 Plus, check out our detailed documentation at [here](https://docs.auto-rag.com/index.html).
+
+
+## Citation
+
+```bibtex
+@misc{kim2024autoragautomatedframeworkoptimization,
+      title={AutoRAG: Automated Framework for optimization of Retrieval Augmented Generation Pipeline},
+      author={Dongkyu Kim and Byoungwook Kim and Donggeon Han and Matouš Eibich},
+      year={2024},
+      eprint={2410.20878},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2410.20878},
+}
+```

{autorag-0.3.8 → autorag-0.3.9}/AutoRAG.egg-info/SOURCES.txt

@@ -225,12 +225,23 @@ docs/source/_static/trial_json.png
 docs/source/_static/trial_summary.png
 docs/source/_static/web_interface.png
 docs/source/_static/web_interface_gradio.png
+docs/source/_static/qa/data_creation_schema.png
 docs/source/_static/roadmap/RAG_paradigms.png
 docs/source/_static/roadmap/advanced_RAG.png
 docs/source/_static/roadmap/cycle.png
 docs/source/_static/roadmap/merger.png
 docs/source/_static/roadmap/node_line_modular.png
 docs/source/_static/roadmap/policy.png
+docs/source/_static/yaml/compact_structured.png
+docs/source/_static/yaml/detail_folder.png
+docs/source/_static/yaml/full_modules.png
+docs/source/_static/yaml/full_structured.png
+docs/source/_static/yaml/full_yaml_structure.png
+docs/source/_static/yaml/gpu_modules.png
+docs/source/_static/yaml/half_structured.png
+docs/source/_static/yaml/non_gpu_modules.png
+docs/source/_static/yaml/sample_yaml_folder.png
+docs/source/_static/yaml/simple_structured.png
 docs/source/api_spec/autorag.data.chunk.rst
 docs/source/api_spec/autorag.data.corpus.rst
 docs/source/api_spec/autorag.data.legacy.corpus.rst
@@ -337,6 +348,7 @@ docs/source/nodes/retrieval/vectordb.md
 docs/source/optimization/custom_config.md
 docs/source/optimization/folder_structure.md
 docs/source/optimization/optimization.md
+docs/source/optimization/sample_config.md
 docs/source/optimization/strategies.md
 docs/source/roadmap/modular_rag.md
 docs/source/vectordb/chroma.md

{autorag-0.3.8 → autorag-0.3.9}/AutoRAG.egg-info/requires.txt

@@ -19,6 +19,7 @@ voyageai
 mixedbread-ai
 llama-index-llms-bedrock
 scikit-learn
+emoji
 pymilvus
 chromadb>=0.5.0
 quart

{autorag-0.3.8 → autorag-0.3.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: AutoRAG
-Version: 0.3.8
+Version: 0.3.9
 Summary: Automatically Evaluate RAG pipelines with your own data. Find optimal structure for new RAG product.
 Author-email: Marker-Inc <vkehfdl1@gmail.com>
 License:                                  Apache License
@@ -240,6 +240,7 @@ Requires-Dist: voyageai
 Requires-Dist: mixedbread-ai
 Requires-Dist: llama-index-llms-bedrock
 Requires-Dist: scikit-learn
+Requires-Dist: emoji
 Requires-Dist: pymilvus
 Requires-Dist: chromadb>=0.5.0
 Requires-Dist: quart
@@ -524,7 +525,7 @@ initial_qa.to_parquet('./qa.parquet', './corpus.parquet')
 
 # RAG Optimization
 
-<a href="https://huggingface.co/spaces/AutoRAG/AutoRAG-optimization">
+<a href="https://huggingface.co/spaces/AutoRAG/RAG-Pipeline-Optimization">
 <img src="https://github.com/user-attachments/assets/8c6e4b02-3938-4560-b817-c95764965b50" alt="Hugging Face Sticker" style="width:200px;height:auto;">
 </a>
 
@@ -818,3 +819,18 @@ We are developing AutoRAG as open-source.
 So this project welcomes contributions and suggestions. Feel free to contribute to this project.
 
 Plus, check out our detailed documentation at [here](https://docs.auto-rag.com/index.html).
+
+
+## Citation
+
+```bibtex
+@misc{kim2024autoragautomatedframeworkoptimization,
+      title={AutoRAG: Automated Framework for optimization of Retrieval Augmented Generation Pipeline},
+      author={Dongkyu Kim and Byoungwook Kim and Donggeon Han and Matouš Eibich},
+      year={2024},
+      eprint={2410.20878},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2410.20878},
+}
+```

{autorag-0.3.8 → autorag-0.3.9}/README.md

@@ -221,7 +221,7 @@ initial_qa.to_parquet('./qa.parquet', './corpus.parquet')
 
 # RAG Optimization
 
-<a href="https://huggingface.co/spaces/AutoRAG/AutoRAG-optimization">
+<a href="https://huggingface.co/spaces/AutoRAG/RAG-Pipeline-Optimization">
 <img src="https://github.com/user-attachments/assets/8c6e4b02-3938-4560-b817-c95764965b50" alt="Hugging Face Sticker" style="width:200px;height:auto;">
 </a>
 
@@ -515,3 +515,18 @@ We are developing AutoRAG as open-source.
 So this project welcomes contributions and suggestions. Feel free to contribute to this project.
 
 Plus, check out our detailed documentation at [here](https://docs.auto-rag.com/index.html).
+
+
+## Citation
+
+```bibtex
+@misc{kim2024autoragautomatedframeworkoptimization,
+      title={AutoRAG: Automated Framework for optimization of Retrieval Augmented Generation Pipeline},
+      author={Dongkyu Kim and Byoungwook Kim and Donggeon Han and Matouš Eibich},
+      year={2024},
+      eprint={2410.20878},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2410.20878},
+}
+```

autorag-0.3.9/autorag/VERSION

@@ -0,0 +1 @@
+0.3.9

{autorag-0.3.8 → autorag-0.3.9}/autorag/__init__.py

@@ -121,6 +121,9 @@ try:
 	from llama_index.llms.huggingface import HuggingFaceLLM
 	from llama_index.llms.ollama import Ollama
 
+	generator_models["huggingfacellm"] = HuggingFaceLLM
+	generator_models["ollama"] = Ollama
+
 except ImportError:
 	logger.info(
 		"You are using API version of AutoRAG."

{autorag-0.3.8 → autorag-0.3.9}/autorag/cli.py

@@ -17,8 +17,14 @@ from autorag.validator import Validator
 
 logger = logging.getLogger("AutoRAG")
 
+autorag_dir = os.path.dirname(os.path.realpath(__file__))
+version_file = os.path.join(autorag_dir, "VERSION")
+with open(version_file, "r") as f:
+	__version__ = f.read().strip()
+
 
 @click.group()
+@click.version_option(__version__)
 def cli():
 	pass
 

{autorag-0.3.8 → autorag-0.3.9}/autorag/data/parse/clova.py

@@ -79,6 +79,7 @@ async def clova_ocr_pure(
 	table_detection: bool = False,
 ) -> Tuple[str, str, int]:
 	session = aiohttp.ClientSession()
+	table_html = ""
 	headers = {"X-OCR-SECRET": api_key, "Content-Type": "application/json"}
 
 	# Convert image data to base64

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/colbert.py

@@ -2,7 +2,6 @@ from typing import List, Tuple
 
 import numpy as np
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.utils.util import (
@@ -141,7 +140,7 @@ def get_colbert_embedding_batch(
 	input_batches = slice_tokenizer_result(encoding, batch_size)
 	result_embedding = []
 	with torch.no_grad():
-		for encoding_batch in tqdm(input_batches):
+		for encoding_batch in input_batches:
 			result_embedding.append(model(**encoding_batch).last_hidden_state)
 	total_tensor = torch.cat(
 		result_embedding, dim=0

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/flag_embedding.py

@@ -1,7 +1,6 @@
 from typing import List, Tuple, Iterable
 
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.utils.util import (
@@ -103,7 +102,7 @@ def flag_embedding_run_model(input_texts, model, batch_size: int):
 		raise ImportError("FlagEmbeddingReranker requires PyTorch to be installed.")
 	batch_input_texts = make_batch(input_texts, batch_size)
 	results = []
-	for batch_texts in tqdm(batch_input_texts):
+	for batch_texts in batch_input_texts:
 		with torch.no_grad():
 			pred_scores = model.compute_score(sentence_pairs=batch_texts)
 		if batch_size == 1 or not isinstance(pred_scores, Iterable):

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/koreranker.py

@@ -2,7 +2,6 @@ from typing import List, Tuple
 
 import numpy as np
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.utils.util import (
@@ -109,7 +108,7 @@ def koreranker_run_model(input_texts, model, tokenizer, device, batch_size: int)
 		raise ImportError("For using KoReranker, please install torch first.")
 	batch_input_texts = make_batch(input_texts, batch_size)
 	results = []
-	for batch_texts in tqdm(batch_input_texts):
+	for batch_texts in batch_input_texts:
 		inputs = tokenizer(
 			batch_texts,
 			padding=True,

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/monot5.py

@@ -2,7 +2,6 @@ from itertools import chain
 from typing import List, Tuple
 
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.utils.util import (
@@ -166,7 +165,7 @@ def monot5_run_model(
 		raise ImportError("For using MonoT5 Reranker, please install torch first.")
 	batch_input_texts = make_batch(input_texts, batch_size)
 	results = []
-	for batch_texts in tqdm(batch_input_texts):
+	for batch_texts in batch_input_texts:
 		flattened_batch_texts = list(chain.from_iterable(batch_texts))
 		input_encodings = tokenizer(
 			flattened_batch_texts,

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/openvino.py

@@ -3,7 +3,6 @@ from typing import Any, List, Tuple
 
 import numpy as np
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 
@@ -173,7 +172,7 @@ def openvino_run_model(
 ):
 	batch_input_texts = make_batch(input_texts, batch_size)
 	results = []
-	for batch_texts in tqdm(batch_input_texts):
+	for batch_texts in batch_input_texts:
 		input_tensors = tokenizer(
 			batch_texts,
 			padding=True,

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/sentence_transformer.py

@@ -1,7 +1,6 @@
 from typing import List, Tuple
 
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.utils.util import (
@@ -123,7 +122,7 @@ def sentence_transformer_run_model(input_texts, model, batch_size: int):
 		)
 	batch_input_texts = make_batch(input_texts, batch_size)
 	results = []
-	for batch_texts in tqdm(batch_input_texts):
+	for batch_texts in batch_input_texts:
 		with torch.no_grad():
 			pred_scores = model.predict(sentences=batch_texts, apply_softmax=True)
 		results.extend(pred_scores.tolist())

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/tart/tart.py

@@ -2,7 +2,6 @@ from itertools import chain
 from typing import List, Tuple
 
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.nodes.passagereranker.tart.modeling_enc_t5 import (
@@ -123,9 +122,7 @@ def tart_run_model(
 	batch_input_texts = make_batch(flattened_texts, batch_size)
 	batch_contents_list = make_batch(flattened_contents, batch_size)
 	results = []
-	for batch_texts, batch_contents in tqdm(
-		zip(batch_input_texts, batch_contents_list)
-	):
+	for batch_texts, batch_contents in zip(batch_input_texts, batch_contents_list):
 		feature = tokenizer(
 			batch_texts,
 			batch_contents,

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/passagereranker/upr.py

@@ -2,7 +2,6 @@ import logging
 from typing import List, Tuple
 
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.nodes.passagereranker.base import BasePassageReranker
 from autorag.utils import result_to_dataframe
@@ -72,7 +71,6 @@ class Upr(BasePassageReranker):
 
 		:return: tuple of lists containing the reranked contents, ids, and scores
 		"""
-		tqdm.pandas()
 		df = pd.DataFrame(
 			{
 				"query": queries,
@@ -81,7 +79,7 @@ class Upr(BasePassageReranker):
 			}
 		)
 
-		df["scores"] = df.progress_apply(
+		df["scores"] = df.apply(
 			lambda row: self.scorer.compute(
 				query=row["query"], contents=row["contents"]
 			),

{autorag-0.3.8 → autorag-0.3.9}/autorag/nodes/retrieval/run.py

@@ -6,7 +6,6 @@ from typing import List, Callable, Dict, Tuple, Union
 
 import numpy as np
 import pandas as pd
-from tqdm import tqdm
 
 from autorag.evaluation import evaluate_retrieval
 from autorag.schema.metricinput import MetricInput
@@ -512,7 +511,7 @@ def optimize_hybrid(
 	).tolist()
 
 	result_list = []
-	for weight_value in tqdm(weight_candidates):
+	for weight_value in weight_candidates:
 		result_df = hybrid_module_func.run_evaluator(
 			project_dir=project_dir,
 			previous_result=previous_result,

{autorag-0.3.8 → autorag-0.3.9}/autorag/utils/preprocess.py

@@ -3,7 +3,7 @@ from datetime import datetime
 import numpy as np
 import pandas as pd
 
-from autorag.utils.util import normalize_unicode
+from autorag.utils.util import preprocess_text
 
 
 def validate_qa_dataset(df: pd.DataFrame):
@@ -60,9 +60,9 @@ def cast_qa_dataset(df: pd.DataFrame):
 	), "query must be string type."
 	df["retrieval_gt"] = df["retrieval_gt"].apply(cast_retrieval_gt)
 	df["generation_gt"] = df["generation_gt"].apply(cast_generation_gt)
-	df["query"] = df["query"].apply(normalize_unicode)
+	df["query"] = df["query"].apply(preprocess_text)
 	df["generation_gt"] = df["generation_gt"].apply(
-		lambda x: list(map(normalize_unicode, x))
+		lambda x: list(map(preprocess_text, x))
 	)
 	return df
 
@@ -104,13 +104,13 @@ def cast_corpus_dataset(df: pd.DataFrame):
 		lambda x: make_prev_next_id_metadata(x, "next_id")
 	)
 
-	df["contents"] = df["contents"].apply(normalize_unicode)
+	df["contents"] = df["contents"].apply(preprocess_text)
 
 	def normalize_unicode_metadata(metadata: dict):
 		result = {}
 		for key, value in metadata.items():
 			if isinstance(value, str):
-				result[key] = normalize_unicode(value)
+				result[key] = preprocess_text(value)
 			else:
 				result[key] = value
 		return result

{autorag-0.3.8 → autorag-0.3.9}/autorag/utils/util.py

@@ -15,6 +15,7 @@ from json import JSONDecoder
 from typing import List, Callable, Dict, Optional, Any, Collection, Iterable
 
 from asyncio import AbstractEventLoop
+import emoji
 import numpy as np
 import pandas as pd
 import tiktoken
@@ -334,7 +335,11 @@ def save_parquet_safe(df: pd.DataFrame, filepath: str, upsert: bool = False):
 def openai_truncate_by_token(
 	texts: List[str], token_limit: int, model_name: str
 ) -> List[str]:
-	tokenizer = tiktoken.encoding_for_model(model_name)
+	try:
+		tokenizer = tiktoken.encoding_for_model(model_name)
+	except KeyError:
+		# This is not a real OpenAI model
+		return texts
 
 	def truncate_text(text: str, limit: int, tokenizer):
 		tokens = tokenizer.encode(text)
@@ -464,6 +469,14 @@ def find_node_summary_files(trial_dir: str) -> List[str]:
 	return filtered_files
 
 
+def preprocess_text(text: str) -> str:
+	return normalize_unicode(demojize(text))
+
+
+def demojize(text: str) -> str:
+	return emoji.demojize(text)
+
+
 def normalize_unicode(text: str) -> str:
 	return unicodedata.normalize("NFC", text)
 
@@ -699,10 +712,10 @@ def decode_multiple_json_from_bytes(byte_data: bytes) -> list:
 	Decode multiple JSON objects from bytes received from SSE server.
 
 	Args:
-		byte_data: Bytes containing one or more JSON objects
+	        byte_data: Bytes containing one or more JSON objects
 
 	Returns:
-		List of decoded JSON objects
+	        List of decoded JSON objects
 	"""
 	# Decode bytes to string
 	try:

{autorag-0.3.8 → autorag-0.3.9}/autorag/validator.py

@@ -42,8 +42,20 @@ class Validator:
 		self.corpus_data = cast_corpus_dataset(self.corpus_data)
 
 	def validate(self, yaml_path: str, qa_cnt: int = 5, random_state: int = 42):
-		# sample QA data
-		sample_qa_df = self.qa_data.sample(qa_cnt, random_state=random_state)
+		# Determine the sample size and log a warning if qa_cnt is larger than available records
+		available_records = len(self.qa_data)
+		safe_sample_size = min(qa_cnt, available_records)  # 먼저 safe_sample_size 계산
+
+		if safe_sample_size < qa_cnt:
+			logger.warning(
+				f"Minimal Requested sample size ({qa_cnt}) is larger than available records ({available_records}). "
+				f"Sampling will be limited to {safe_sample_size} records. "
+			)
+
+		# safe sample QA data
+		sample_qa_df = self.qa_data.sample(
+			n=safe_sample_size, random_state=random_state
+		)
 		sample_qa_df.reset_index(drop=True, inplace=True)
 
 		# get doc_id

{autorag-0.3.8 → autorag-0.3.9}/docs/source/data_creation/data_creation.md

@@ -15,6 +15,13 @@ In this new data creation pipeline, we have three schemas. `Raw`, `QA`, and `Cor
 You can use the corpus to generate the answer for the question.
 You have to make corpus data from your documents using parsing and chunking.
 
+In other words,
+`Raw` : parsed data
+`Corpus` : chunked data
+`QA` : Question & Answer dataset based on the corpus
+
+![](../_static/qa/data_creation_schema.png)
+
 To see the tutorial of the data creation, check [here](tutorial.md).
 
 ```{toctree}

{autorag-0.3.8 → autorag-0.3.9}/docs/source/index.rst

@@ -112,6 +112,7 @@ Also, feel free to ask your question at our `github issue <https://github.com/Ma
    optimization/folder_structure.md
    optimization/custom_config.md
    optimization/strategies.md
+   optimization/sample_config.md
 
 
 .. toctree::

{autorag-0.3.8 → autorag-0.3.9}/docs/source/local_model.md

@@ -34,18 +34,17 @@ The following modules can use generator module, which including `llama_index_llm
 - [multi_query_expansion](nodes/query_expansion/multi_query_expansion.md)
 - [tree_summarize](nodes/passage_compressor/tree_summarize.md)
 
-### Supporting LLM models
 
-We support most of the llm that LlamaIndex is supporting.
-To change the LLM model type, you can change the `llm` parameter to the following values:
+### Supporting LLM Models
 
-| LLM Model Type | llm parameter  |
-|:--------------:|:--------------:|
-|     OpenAI     |     openai     |
-| HuggingFaceLLM | huggingfacellm |
-|   OpenAILike   |   openailike   |
-|     Ollama     |     ollama     |
-|    Bedrock     |    bedrock     |
+We support most of the LLMs that LlamaIndex supports. You can use different types of LLM interfaces by configuring the `llm` parameter:
+
+| LLM Model Type | llm parameter  | Description |
+|:--------------:|:--------------:|-------------|
+|     OpenAI     |     openai     | For OpenAI models (GPT-3.5, GPT-4) |
+|   OpenAILike   |   openailike   | For models with OpenAI-compatible APIs (e.g., Mistral, Claude) |
+|     Ollama     |     ollama     | For locally running Ollama models |
+|    Bedrock     |    bedrock     | For AWS Bedrock models |
 
 For example, if you want to use `OpenAILike` model, you can set `llm` parameter to `openailike`.
 
@@ -69,6 +68,52 @@ The most frequently used parameters are `model`, `max_token`, and `temperature`.
 Please check what you can set for the model parameter
 at [LlamaIndex LLM](https://docs.llamaindex.ai/en/stable/module_guides/models/llms/).
 
+#### Using HuggingFace Models
+
+There are two main ways to use HuggingFace models:
+
+1. **Through OpenAILike Interface** (Recommended for hosted API endpoints):
+```yaml
+nodes:
+  - node_line_name: node_line_1
+    nodes:
+      - node_type: generator
+        modules:
+          - module_type: llama_index_llm
+            llm: openailike
+            model: mistralai/Mistral-7B-Instruct-v0.2
+            api_base: your_api_base
+            api_key: your_api_key
+```
+
+2. **Through Direct HuggingFace Integration** (For local deployment):
+```yaml
+nodes:
+  - node_line_name: node_line_1
+    nodes:
+      - node_type: generator
+        modules:
+          - module_type: llama_index_llm
+            llm: huggingface
+            model_name: mistralai/Mistral-7B-Instruct-v0.2
+            device_map: "auto"
+            model_kwargs:
+              torch_dtype: "float16"
+```
+
+#### Common Parameters
+
+The most frequently used parameters for LLM configuration are:
+
+- `model`: The model identifier or name
+- `max_tokens`: Maximum number of tokens in the response
+- `temperature`: Controls randomness in the output (0.0 to 1.0)
+- `api_base`: API endpoint URL (for hosted models)
+- `api_key`: Authentication key (if required)
+
+For a complete list of available parameters, please refer to the
+[LlamaIndex LLM documentation](https://docs.llamaindex.ai/en/stable/module_guides/models/llms/).
+
 ### Add more LLM models
 
 You can add more LLM models for AutoRAG.

{autorag-0.3.8 → autorag-0.3.9}/docs/source/migration.md

@@ -2,9 +2,9 @@
 
 1. [v0.3 migration guide](#v03-migration-guide)
 
-# v0.3 migration guide
+## v0.3 migration guide
 
-## Data Creation
+### Data Creation
 
 From the v0.3 version, the previous data creation library goes into the `legacy` package.
 Instead of legacy data creation, the `beta` package is introduced.
@@ -38,7 +38,7 @@ from autorag.data.qa.sample import random_single_hop
 from autorag.data.qa.schema import Raw
 ```
 
-# v0.3.7 migration guide
+## v0.3.7 migration guide
 
 At v0.3.6, there are changes of the vectordb.
 You have to specify what vectordb you want to use at the config YAML file.