feder-cr/Jobs_Applier_AI_Agent_AIHawk · error · ValueError

Vectorstore not initialized. Run extract_job_description fir

Error message

Vectorstore not initialized. Run extract_job_description first.

What it means

LLMJobParser first ingests the job posting into a vector store (extract_job_description) and then answers extraction queries via similarity search in _retrieve_context. If extraction is attempted before ingestion, self.vectorstore is falsy and the guard raises.

Source

Thrown at src/libs/resume_and_cover_builder/llm/llm_job_parser.py:101

        # Create the vectorstore using FAISS
        try:
            self.vectorstore = FAISS.from_documents(documents=all_splits, embedding=self.llm_embeddings)
            logger.debug("Vectorstore successfully initialized.")
        except Exception as e:
            logger.error(f"Error during vectorstore creation: {e}")
            raise

    def _retrieve_context(self, query: str, top_k: int = 3) -> str:
        """
        Retrieves the most relevant text fragments using the retriever.
        Args:
            query (str): The search query.
            top_k (int): Number of fragments to retrieve.
        Returns:
            str: Concatenated text fragments.
        """
        if not self.vectorstore:
            raise ValueError("Vectorstore not initialized. Run extract_job_description first.")
        
        retriever = self.vectorstore.as_retriever()
        retrieved_docs = retriever.get_relevant_documents(query)[:top_k]
        context = "\n\n".join(doc.page_content for doc in retrieved_docs)
        logger.debug(f"Context retrieved for query '{query}': {context[:200]}...")  # Log the first 200 characters
        return context
    
    def _extract_information(self, question: str, retrieval_query: str) -> str:
        """
        Generic method to extract specific information using the retriever and LLM.
        Args:
            question (str): The question to ask the LLM for extraction.
            retrieval_query (str): The query to use for retrieving relevant context.
        Returns:
            str: The extracted information.
        """
        context = self._retrieve_context(retrieval_query)
        

View on GitHub (pinned to 79155b52fa)

Solutions

1. Call extract_job_description(job_text) on the parser before any information-extraction call.
2. Check parser.vectorstore is truthy before extracting; re-run ingestion if not.
3. If ingestion can fail, propagate that error so extraction is never attempted on an un-ingested parser.

Example fix

# before
info = parser._extract_information('salary')
# after
parser.extract_job_description(job_text)  # populates vectorstore
info = parser._extract_information('salary')

if not getattr(parser, 'vectorstore', None):
    parser.extract_job_description(job_text)
info = parser._extract_information('salary')

def parser_ready(parser) -> bool:
    return getattr(parser, 'vectorstore', None) is not None

try:
    info = parser._extract_information(query)
except ValueError as e:
    if 'Vectorstore not initialized' in str(e):
        parser.extract_job_description(job_text)
        info = parser._extract_information(query)
    else:
        raise

When it happens

Trigger: Calling _extract_information (or a public method that uses it) before calling extract_job_description on the same LLMJobParser instance, or when ingestion failed and left vectorstore unset/None.

Common situations: Reordering the parse pipeline, reusing a parser instance across jobs without re-ingesting, or ingestion errors (embedding API failure) that were caught upstream leaving the store empty.

Related errors

• Job application file path is not set. Please create the appl
• dir path cannot be None
• Unsupported model type: {llm_model_type}
• Could not extract section name from the response.
• Section '{section_name}' not found in either resume or job_a

AI-assisted analysis of feder-cr/Jobs_Applier_AI_Agent_AIHawk@79155b52fa (2026-08-28). Data as JSON: /api/errors/a821faa274eb928c. Report an issue: GitHub.