Skip to main content

RAGatouille

RAGatouille makes it as simple as can be to use ColBERT!

ColBERT is a fast and accurate retrieval model, enabling scalable BERT-based search over large text collections in tens of milliseconds.

See the ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction paper.

We can use this as a retriever. It will show functionality specific to this integration. After going through, it may be useful to explore relevant use-case pages to learn how to use this vector store as part of a larger chain.

This page covers how to use RAGatouille as a retriever in a LangChain chain.

Setupโ€‹

The integration lives in the ragatouille package.

pip install -U ragatouille

Usageโ€‹

This example is taken from their documentation

from ragatouille import RAGPretrainedModel

RAG = RAGPretrainedModel.from_pretrained("colbert-ir/colbertv2.0")
import requests


def get_wikipedia_page(title: str):
    """
    Retrieve the full text content of a Wikipedia page.

    :param title: str - Title of the Wikipedia page.
    :return: str - Full text content of the page as raw string.
    """
    # Wikipedia API endpoint
    URL = "https://en.wikipedia.org/w/api.php"

    # Parameters for the API request
    params = {
        "action": "query",
        "format": "json",
        "titles": title,
        "prop": "extracts",
        "explaintext": True,
    }

    # Custom User-Agent header to comply with Wikipedia's best practices
    headers = {"User-Agent": "RAGatouille_tutorial/0.0.1 (ben@clavie.eu)"}

    response = requests.get(URL, params=params, headers=headers)
    data = response.json()

    # Extracting page content
    page = next(iter(data["query"]["pages"].values()))
    return page["extract"] if "extract" in page else None
full_document = get_wikipedia_page("Hayao_Miyazaki")
RAG.index(
    collection=[full_document],
    index_name="Miyazaki-123",
    max_document_length=180,
    split_documents=True,
)


[Jan 07, 10:38:18] #> Creating directory .ragatouille/colbert/indexes/Miyazaki-123 


#> Starting...
[Jan 07, 10:38:23] Loading segmented_maxsim_cpp extension (set COLBERT_LOAD_TORCH_EXTENSION_VERBOSE=True for more info)...
``````output
/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/cuda/amp/grad_scaler.py:125: UserWarning: torch.cuda.amp.GradScaler is enabled, but CUDA is not available.  Disabling.
  warnings.warn(

  0%|          | 0/2 [00:00<?, ?it/s]/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/amp/autocast_mode.py:250: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
  warnings.warn(
``````output
[Jan 07, 10:38:24] [0] 		 #> Encoding 81 passages..
``````output
 50%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ     | 1/2 [00:02<00:02,  2.85s/it]/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/amp/autocast_mode.py:250: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
  warnings.warn(
100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 2/2 [00:03<00:00,  1.74s/it]
WARNING clustering 10001 points to 1024 centroids: please provide at least 39936 training points
``````output
[Jan 07, 10:38:27] [0] 		 avg_doclen_est = 129.9629669189453 	 len(local_sample) = 81
[Jan 07, 10:38:27] [0] 		 Creating 1,024 partitions.
[Jan 07, 10:38:27] [0] 		 *Estimated* 10,527 embeddings.
[Jan 07, 10:38:27] [0] 		 #> Saving the indexing plan to .ragatouille/colbert/indexes/Miyazaki-123/plan.json ..
Clustering 10001 points in 128D to 1024 clusters, redo 1 times, 20 iterations
  Preprocessing in 0.00 s
  Iteration 0 (0.02 s, search 0.02 s): objective=3772.41 imbalance=1.562 nsplit=0       
  Iteration 1 (0.02 s, search 0.02 s): objective=2408.99 imbalance=1.470 nsplit=1       
  Iteration 2 (0.03 s, search 0.03 s): objective=2243.87 imbalance=1.450 nsplit=0       
  Iteration 3 (0.04 s, search 0.04 s): objective=2168.48 imbalance=1.444 nsplit=0       
  Iteration 4 (0.05 s, search 0.05 s): objective=2134.26 imbalance=1.449 nsplit=0       
  Iteration 5 (0.06 s, search 0.05 s): objective=2117.18 imbalance=1.449 nsplit=0       
  Iteration 6 (0.06 s, search 0.06 s): objective=2108.43 imbalance=1.449 nsplit=0       
  Iteration 7 (0.07 s, search 0.07 s): objective=2102.62 imbalance=1.450 nsplit=0       
  Iteration 8 (0.08 s, search 0.08 s): objective=2100.68 imbalance=1.451 nsplit=0       
  Iteration 9 (0.09 s, search 0.08 s): objective=2099.66 imbalance=1.451 nsplit=0       
  Iteration 10 (0.10 s, search 0.09 s): objective=2099.03 imbalance=1.451 nsplit=0       
  Iteration 11 (0.10 s, search 0.10 s): objective=2098.67 imbalance=1.453 nsplit=0       
  Iteration 12 (0.11 s, search 0.11 s): objective=2097.78 imbalance=1.455 nsplit=0       
  Iteration 13 (0.12 s, search 0.12 s): objective=2097.31 imbalance=1.455 nsplit=0       
  Iteration 14 (0.13 s, search 0.12 s): objective=2097.13 imbalance=1.455 nsplit=0       
  Iteration 15 (0.14 s, search 0.13 s): objective=2097.09 imbalance=1.455 nsplit=0       
  Iteration 16 (0.14 s, search 0.14 s): objective=2097.09 imbalance=1.455 nsplit=0       
  Iteration 17 (0.15 s, search 0.15 s): objective=2097.09 imbalance=1.455 nsplit=0       
  Iteration 18 (0.16 s, search 0.15 s): objective=2097.09 imbalance=1.455 nsplit=0       
  Iteration 19 (0.17 s, search 0.16 s): objective=2097.09 imbalance=1.455 nsplit=0       
[0.037, 0.038, 0.041, 0.036, 0.035, 0.036, 0.034, 0.036, 0.034, 0.034, 0.036, 0.037, 0.032, 0.039, 0.035, 0.039, 0.033, 0.035, 0.035, 0.037, 0.037, 0.037, 0.037, 0.037, 0.038, 0.034, 0.037, 0.035, 0.036, 0.037, 0.036, 0.04, 0.037, 0.037, 0.036, 0.034, 0.037, 0.035, 0.038, 0.039, 0.037, 0.039, 0.035, 0.036, 0.036, 0.035, 0.035, 0.038, 0.037, 0.033, 0.036, 0.032, 0.034, 0.035, 0.037, 0.037, 0.041, 0.037, 0.039, 0.033, 0.035, 0.033, 0.036, 0.036, 0.038, 0.036, 0.037, 0.038, 0.035, 0.035, 0.033, 0.034, 0.032, 0.038, 0.037, 0.037, 0.036, 0.04, 0.033, 0.037, 0.035, 0.04, 0.036, 0.04, 0.032, 0.037, 0.036, 0.037, 0.034, 0.042, 0.037, 0.035, 0.035, 0.038, 0.034, 0.036, 0.041, 0.035, 0.036, 0.037, 0.041, 0.04, 0.036, 0.036, 0.035, 0.036, 0.034, 0.033, 0.036, 0.033, 0.037, 0.038, 0.036, 0.033, 0.038, 0.037, 0.038, 0.037, 0.039, 0.04, 0.034, 0.034, 0.036, 0.039, 0.033, 0.037, 0.035, 0.037]
[Jan 07, 10:38:27] [0] 		 #> Encoding 81 passages..
``````output
0it [00:00, ?it/s]
  0%|          | 0/2 [00:00<?, ?it/s]
 50%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ     | 1/2 [00:02<00:02,  2.53s/it]
100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 2/2 [00:03<00:00,  1.56s/it]
1it [00:03,  3.16s/it]
100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 1/1 [00:00<00:00, 4017.53it/s]
100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 1024/1024 [00:00<00:00, 306105.57it/s]
``````output
[Jan 07, 10:38:30] #> Optimizing IVF to store map from centroids to list of pids..
[Jan 07, 10:38:30] #> Building the emb2pid mapping..
[Jan 07, 10:38:30] len(emb2pid) = 10527
[Jan 07, 10:38:30] #> Saved optimized IVF to .ragatouille/colbert/indexes/Miyazaki-123/ivf.pid.pt

#> Joined...
Done indexing!
results = RAG.search(query="What animation studio did Miyazaki found?", k=3)
Loading searcher for index Miyazaki-123 for the first time... This may take a few seconds
[Jan 07, 10:38:34] Loading segmented_maxsim_cpp extension (set COLBERT_LOAD_TORCH_EXTENSION_VERBOSE=True for more info)...
[Jan 07, 10:38:35] #> Loading codec...
[Jan 07, 10:38:35] #> Loading IVF...
[Jan 07, 10:38:35] Loading segmented_lookup_cpp extension (set COLBERT_LOAD_TORCH_EXTENSION_VERBOSE=True for more info)...
``````output
/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/cuda/amp/grad_scaler.py:125: UserWarning: torch.cuda.amp.GradScaler is enabled, but CUDA is not available.  Disabling.
  warnings.warn(
``````output
[Jan 07, 10:38:35] #> Loading doclens...
``````output
100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 1/1 [00:00<00:00, 3872.86it/s]
``````output
[Jan 07, 10:38:35] #> Loading codes and residuals...
``````output

100%|โ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆโ–ˆ| 1/1 [00:00<00:00, 604.89it/s]
``````output
[Jan 07, 10:38:35] Loading filter_pids_cpp extension (set COLBERT_LOAD_TORCH_EXTENSION_VERBOSE=True for more info)...
``````output
[Jan 07, 10:38:35] Loading decompress_residuals_cpp extension (set COLBERT_LOAD_TORCH_EXTENSION_VERBOSE=True for more info)...
Searcher loaded!

#> QueryTokenizer.tensorize(batch_text[0], batch_background[0], bsize) ==
#> Input: . What animation studio did Miyazaki found?, 		 True, 		 None
#> Output IDs: torch.Size([32]), tensor([  101,     1,  2054,  7284,  2996,  2106,  2771,  3148, 18637,  2179,
         1029,   102,   103,   103,   103,   103,   103,   103,   103,   103,
          103,   103,   103,   103,   103,   103,   103,   103,   103,   103,
          103,   103])
#> Output Mask: torch.Size([32]), tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0])
``````output
/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/amp/autocast_mode.py:250: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
  warnings.warn(
results
[{'content': 'In April 1984, Miyazaki opened his own office in Suginami Ward, naming it Nibariki.\n\n\n=== Studio Ghibli ===\n\n\n==== Early films (1985โ€“1996) ====\nIn June 1985, Miyazaki, Takahata, Tokuma and Suzuki founded the animation production company Studio Ghibli, with funding from Tokuma Shoten. Studio Ghibli\'s first film, Laputa: Castle in the Sky (1986), employed the same production crew of Nausicaรค. Miyazaki\'s designs for the film\'s setting were inspired by Greek architecture and "European urbanistic templates".',
  'score': 25.90749740600586,
  'rank': 1},
 {'content': 'Hayao Miyazaki (ๅฎฎๅดŽ ้งฟ or ๅฎฎ๏จ‘ ้งฟ, Miyazaki Hayao, [mija๊œœzaki hajao]; born January 5, 1941) is a Japanese animator, filmmaker, and manga artist. A co-founder of Studio Ghibli, he has attained international acclaim as a masterful storyteller and creator of Japanese animated feature films, and is widely regarded as one of the most accomplished filmmakers in the history of animation.\nBorn in Tokyo City in the Empire of Japan, Miyazaki expressed interest in manga and animation from an early age, and he joined Toei Animation in 1963. During his early years at Toei Animation he worked as an in-between artist and later collaborated with director Isao Takahata.',
  'score': 25.4748477935791,
  'rank': 2},
 {'content': 'Glen Keane said Miyazaki is a "huge influence" on Walt Disney Animation Studios and has been "part of our heritage" ever since The Rescuers Down Under (1990). The Disney Renaissance era was also prompted by competition with the development of Miyazaki\'s films. Artists from Pixar and Aardman Studios signed a tribute stating, "You\'re our inspiration, Miyazaki-san!"',
  'score': 24.84897232055664,
  'rank': 3}]

We can then convert easily to a LangChain retriever! We can pass in any kwargs we want when creating (like k)

retriever = RAG.as_langchain_retriever(k=3)
retriever.invoke("What animation studio did Miyazaki found?")
/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/amp/autocast_mode.py:250: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
  warnings.warn(
[Document(page_content='In April 1984, Miyazaki opened his own office in Suginami Ward, naming it Nibariki.\n\n\n=== Studio Ghibli ===\n\n\n==== Early films (1985โ€“1996) ====\nIn June 1985, Miyazaki, Takahata, Tokuma and Suzuki founded the animation production company Studio Ghibli, with funding from Tokuma Shoten. Studio Ghibli\'s first film, Laputa: Castle in the Sky (1986), employed the same production crew of Nausicaรค. Miyazaki\'s designs for the film\'s setting were inspired by Greek architecture and "European urbanistic templates".'),
 Document(page_content='Hayao Miyazaki (ๅฎฎๅดŽ ้งฟ or ๅฎฎ๏จ‘ ้งฟ, Miyazaki Hayao, [mija๊œœzaki hajao]; born January 5, 1941) is a Japanese animator, filmmaker, and manga artist. A co-founder of Studio Ghibli, he has attained international acclaim as a masterful storyteller and creator of Japanese animated feature films, and is widely regarded as one of the most accomplished filmmakers in the history of animation.\nBorn in Tokyo City in the Empire of Japan, Miyazaki expressed interest in manga and animation from an early age, and he joined Toei Animation in 1963. During his early years at Toei Animation he worked as an in-between artist and later collaborated with director Isao Takahata.'),
 Document(page_content='Glen Keane said Miyazaki is a "huge influence" on Walt Disney Animation Studios and has been "part of our heritage" ever since The Rescuers Down Under (1990). The Disney Renaissance era was also prompted by competition with the development of Miyazaki\'s films. Artists from Pixar and Aardman Studios signed a tribute stating, "You\'re our inspiration, Miyazaki-san!"')]

Chainingโ€‹

We can easily combine this retriever in to a chain.

from langchain.chains import create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI

prompt = ChatPromptTemplate.from_template(
    """Answer the following question based only on the provided context:

<context>
{context}
</context>

Question: {input}"""
)

llm = ChatOpenAI()

document_chain = create_stuff_documents_chain(llm, prompt)
retrieval_chain = create_retrieval_chain(retriever, document_chain)
retrieval_chain.invoke({"input": "What animation studio did Miyazaki found?"})
/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/amp/autocast_mode.py:250: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
  warnings.warn(
{'input': 'What animation studio did Miyazaki found?',
 'context': [Document(page_content='In April 1984, Miyazaki opened his own office in Suginami Ward, naming it Nibariki.\n\n\n=== Studio Ghibli ===\n\n\n==== Early films (1985โ€“1996) ====\nIn June 1985, Miyazaki, Takahata, Tokuma and Suzuki founded the animation production company Studio Ghibli, with funding from Tokuma Shoten. Studio Ghibli\'s first film, Laputa: Castle in the Sky (1986), employed the same production crew of Nausicaรค. Miyazaki\'s designs for the film\'s setting were inspired by Greek architecture and "European urbanistic templates".'),
  Document(page_content='Hayao Miyazaki (ๅฎฎๅดŽ ้งฟ or ๅฎฎ๏จ‘ ้งฟ, Miyazaki Hayao, [mija๊œœzaki hajao]; born January 5, 1941) is a Japanese animator, filmmaker, and manga artist. A co-founder of Studio Ghibli, he has attained international acclaim as a masterful storyteller and creator of Japanese animated feature films, and is widely regarded as one of the most accomplished filmmakers in the history of animation.\nBorn in Tokyo City in the Empire of Japan, Miyazaki expressed interest in manga and animation from an early age, and he joined Toei Animation in 1963. During his early years at Toei Animation he worked as an in-between artist and later collaborated with director Isao Takahata.'),
  Document(page_content='Glen Keane said Miyazaki is a "huge influence" on Walt Disney Animation Studios and has been "part of our heritage" ever since The Rescuers Down Under (1990). The Disney Renaissance era was also prompted by competition with the development of Miyazaki\'s films. Artists from Pixar and Aardman Studios signed a tribute stating, "You\'re our inspiration, Miyazaki-san!"')],
 'answer': 'Miyazaki founded Studio Ghibli.'}
for s in retrieval_chain.stream({"input": "What animation studio did Miyazaki found?"}):
    print(s.get("answer", ""), end="")
/Users/harrisonchase/.pyenv/versions/3.10.1/envs/langchain/lib/python3.10/site-packages/torch/amp/autocast_mode.py:250: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
  warnings.warn(
``````output
Miyazaki founded Studio Ghibli.

Was this page helpful?


You can also leave detailed feedback on GitHub.