ask_vectara
Query Vectara's RAG system to retrieve search results and generate contextual responses using specified corpus keys and API parameters for accurate information extraction.
Instructions
Run a RAG query using Vectara, returning search results with a generated response.
Args:
query: str, The user query to run - required.
corpus_keys: list[str], List of Vectara corpus keys to use for the search - required. Please ask the user to provide one or more corpus keys.
api_key: str, The Vectara API key - required.
n_sentences_before: int, Number of sentences before the answer to include in the context - optional, default is 2.
n_sentences_after: int, Number of sentences after the answer to include in the context - optional, default is 2.
lexical_interpolation: float, The amount of lexical interpolation to use - optional, default is 0.005.
max_used_search_results: int, The maximum number of search results to use - optional, default is 10.
generation_preset_name: str, The name of the generation preset to use - optional, default is "vectara-summary-table-md-query-ext-jan-2025-gpt-4o".
response_language: str, The language of the response - optional, default is "eng".
Returns:
The response from Vectara, including the generated answer and the search results.Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| api_key | No | ||
| corpus_keys | No | ||
| generation_preset_name | No | vectara-summary-table-md-query-ext-jan-2025-gpt-4o | |
| lexical_interpolation | No | ||
| max_used_search_results | No | ||
| n_sentences_after | No | ||
| n_sentences_before | No | ||
| query | Yes | ||
| response_language | No | eng |
Implementation Reference
- vectara_mcp/server.py:483-575 (handler)The primary handler function for the 'ask_vectara' tool. Registered via @mcp.tool() decorator. Includes type-hinted parameters serving as input schema, comprehensive docstring describing args/returns, full execution logic including validation, API payload construction, Vectara query execution, response processing with citations extraction, and error handling.
@mcp.tool() async def ask_vectara( query: str, ctx: Context, corpus_keys: list[str], n_sentences_before: int = 2, n_sentences_after: int = 2, lexical_interpolation: float = 0.005, max_used_search_results: int = 10, generation_preset_name: str = "vectara-summary-table-md-query-ext-jan-2025-gpt-4o", response_language: str = "eng", ) -> dict: """ Run a RAG query using Vectara, returning search results with a generated response. Args: query: str, The user query to run - required. corpus_keys: list[str], List of Vectara corpus keys to use for the search - required. Please ask the user to provide one or more corpus keys. n_sentences_before: int, Number of sentences before the answer to include in the context - optional, default is 2. n_sentences_after: int, Number of sentences after the answer to include in the context - optional, default is 2. lexical_interpolation: float, The amount of lexical interpolation to use - optional, default is 0.005. max_used_search_results: int, The maximum number of search results to use - optional, default is 10. generation_preset_name: str, The name of the generation preset to use - optional, default is "vectara-summary-table-md-query-ext-jan-2025-gpt-4o". response_language: str, The language of the response - optional, default is "eng". Note: API key must be configured first using 'setup_vectara_api_key' tool Returns: dict: Structured response containing: - "summary": Generated AI summary with markdown citations - "citations": List of citation objects with score, text, and metadata - "factual_consistency_score": Score indicating factual consistency (if available) On error, returns dict with "error" key containing error message. """ # Validate parameters validation_error = _validate_common_parameters(query, corpus_keys) if validation_error: return {"error": validation_error} if ctx: ctx.info(f"Running Vectara RAG query: {query}") try: payload = _build_query_payload( query=query, corpus_keys=corpus_keys, n_sentences_before=n_sentences_before, n_sentences_after=n_sentences_after, lexical_interpolation=lexical_interpolation, max_used_search_results=max_used_search_results, generation_preset_name=generation_preset_name, response_language=response_language, enable_generation=True ) result = await _call_vectara_query(payload, ctx) # Extract the generated summary from the response summary_text = "" if "summary" in result: summary_text = result["summary"] elif "answer" in result: summary_text = result["answer"] else: return {"error": f"Unexpected response format: {json.dumps(result, indent=2)}"} # Build citations list citations = [] if "search_results" in result and result["search_results"]: for i, search_result in enumerate(result["search_results"], 1): citation = { "id": i, "score": search_result.get("score", 0.0), "text": search_result.get("text", ""), "document_metadata": search_result.get("document_metadata", {}) } citations.append(citation) # Build response dict response = { "summary": summary_text, "citations": citations } # Add factual consistency score if available if "factual_consistency_score" in result: response["factual_consistency_score"] = result["factual_consistency_score"] return response except Exception as e: return {"error": _format_error("Vectara RAG query", e)} - vectara_mcp/server.py:285-297 (helper)Helper function that makes the actual API request to Vectara's /query endpoint, used by ask_vectara.
async def _call_vectara_query( payload: dict, ctx: Context = None, api_key_override: str = None ) -> dict: """Make API call to Vectara query endpoint""" return await _make_api_request( f"{VECTARA_BASE_URL}/query", payload, ctx, api_key_override, "query" ) - vectara_mcp/server.py:234-283 (helper)Helper function that constructs the detailed JSON payload for the Vectara query API, incorporating all parameters from ask_vectara.
def _build_query_payload( query: str, corpus_keys: list[str], n_sentences_before: int = 2, n_sentences_after: int = 2, lexical_interpolation: float = 0.005, max_used_search_results: int = 10, generation_preset_name: str = "vectara-summary-table-md-query-ext-jan-2025-gpt-4o", response_language: str = "eng", enable_generation: bool = True ) -> dict: """Build the query payload for Vectara API""" payload = { "query": query, "search": { "limit": 100, "corpora": [ { "corpus_key": corpus_key, "lexical_interpolation": lexical_interpolation } for corpus_key in corpus_keys ], "context_configuration": { "sentences_before": n_sentences_before, "sentences_after": n_sentences_after }, "reranker": { "type": "customer_reranker", "reranker_name": "Rerank_Multilingual_v1", "limit": 100, "cutoff": 0.2 } }, "save_history": True, } if enable_generation: payload["generation"] = { "generation_preset_name": generation_preset_name, "max_used_search_results": max_used_search_results, "response_language": response_language, "citations": { "style": "markdown", "url_pattern": "{doc.url}", "text_pattern": "{doc.title}" }, "enable_factual_consistency_score": True } return payload - vectara_mcp/server.py:98-114 (helper)Helper function for parameter validation (query, corpus_keys, API key), called at the start of ask_vectara.
def _validate_common_parameters(query: str = "", corpus_keys: list[str] = None) -> str | None: """Validate common parameters used across Vectara tools. Returns: str: Error message if validation fails, None if valid """ if not query: return "Query is required." if not corpus_keys: return "Corpus keys are required. Please ask the user to provide one or more corpus keys." # Check API key availability api_key = _get_api_key() if not api_key: return API_KEY_ERROR_MESSAGE return None