What causes "Please install nltk: `pip install nltk`" in run-llama/llama_index?

pip install nltk (and rake_nltk — both are required, see the sibling error) or install the extras bundle that includes them.

run-llama/llama_index · error · ImportError

Please install nltk: `pip install nltk`

Error message

Please install nltk: `pip install nltk`

What it means

rake_extract_keywords() lazily imports nltk at call time and converts ImportError into a user-facing ImportError telling you to pip install nltk. nltk itself is an optional dependency: it is only needed when extracting keywords with the RAKE strategy, so the package imports fine without it and the failure appears late, at index build time.

Source

Thrown at llama-index-core/llama_index/core/indices/keyword_table/utils.py:33

    tokens = [t.strip().lower() for t in re.findall(r"\w+", text_chunk)]
    if filter_stopwords:
        tokens = [t for t in tokens if t not in globals_helper.stopwords]

    token_counts = Counter(tokens)
    keywords = [keyword for keyword, count in token_counts.most_common(max_keywords)]
    return set(keywords)


def rake_extract_keywords(
    text_chunk: str,
    max_keywords: Optional[int] = None,
    expand_with_subtokens: bool = True,
) -> Set[str]:
    """Extract keywords with RAKE."""
    try:
        import nltk
    except ImportError:
        raise ImportError("Please install nltk: `pip install nltk`")
    try:
        from rake_nltk import Rake
    except ImportError:
        raise ImportError("Please install rake_nltk: `pip install rake_nltk`")

    r = Rake(
        sentence_tokenizer=nltk.tokenize.sent_tokenize,
        word_tokenizer=nltk.tokenize.wordpunct_tokenize,
    )
    r.extract_keywords_from_text(text_chunk)
    keywords = r.get_ranked_phrases()[:max_keywords]
    if expand_with_subtokens:
        return set(expand_tokens_with_subtokens(keywords))
    else:
        return set(keywords)


def extract_keywords_given_response(

View on GitHub (pinned to afd0fef371)

Solutions

pip install nltk (and rake_nltk — both are required, see the sibling error) or install the extras bundle that includes them.
Verify the import early in your startup: python -c "import nltk; import rake_nltk" before ingestion runs.
If you cannot add deps, use SimpleKeywordTableIndex (regex-based extraction) which needs neither package.

Example fix

# before
index = RAKEKeywordTableIndex.from_documents(docs)  # ImportError: install nltk

# after
pip install nltk rake_nltk
index = RAKEKeywordTableIndex.from_documents(docs)

Defensive patterns

Strategy: validation

Validate before calling

def rake_available() -> bool:
    try:
        import nltk  # noqa: F401
        import rake_nltk  # noqa: F401
        return True
    except ImportError:
        return False

assert rake_available(), "RAKE keyword extraction requires nltk and rake_nltk"

Try / catch

try:
    index = RAKEKeywordTableIndex.from_documents(docs)
except ImportError as e:
    if "nltk" in str(e):
        index = SimpleKeywordTableIndex.from_documents(docs)  # no extra deps
    else:
        raise

Prevention

Install optional NLP deps (nltk, rake_nltk) in every environment that builds RAKE indices.
Exercise the RAKE path in CI so missing deps fail at build time, not in production.
Prefer SimpleKeywordTableIndex when you want zero extra dependencies.

When it happens

Trigger: Building a KeywordTableIndex (or calling as_retriever with KeywordTableRetrieverMode.RAKE) on a base llama-index install without the extras; SimpleKeywordTableIndex vs RAKEKeywordTableIndex — using RAKEKeywordTableIndex.from_documents triggers rake_extract_keywords.

Common situations: Deploying to slim Docker images that only pip install llama-index-core; swapping SimpleKeywordTableIndex for RAKEKeywordTableIndex without adding deps; CI passing (no RAKE path exercised) while production ingestion fails.

Related errors

AI-assisted analysis of run-llama/llama_index@afd0fef371 (2026-08-15). Data as JSON: /api/errors/170cc618ef3ae861. Report an issue: GitHub.