python_get_documentation
Search Python library documentation using natural language queries to find specific functions, classes, and usage examples from consolidated sources.
Instructions
Primary Python documentation lookup tool. Use this for every Python documentation-related query.
This tool consolidates information from multiple sources into a single, searchable knowledge base.
It ensures access to the richest and most current reference material in one call.
Args:
query: A natural language question (e.g., "How do I define a Deployment?").
library: Python library to search documentation for.
version: Optional Library version (e.g., "4.46.1"). Defaults to detected library version if not specified.
top_k: Optional number of top matching documents to return. Defaults to 10.
Returns:
A list of dictionaries, each containing document path and corresponding content.
Example Usage:
# Search Python docs for Transformers
python_get_documentation(query="what is a transformers mlm token", library="transformers", version="4.46.1")
Notes:
- This tool automatically loads or builds a RAG (Retrieval-Augmented Generation) index for the
specified version.
- If an index is not found locally, the tool will fetch and index the documentation before responding.
- You should call this function for any question that needs project documentation context.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | A natural language question (e.g., "How do I use transformers?"). | |
| library | Yes | Python library to search documentation for. | |
| version | No | Optional Library version (e.g., "4.46.1"). Defaults to detected library version if not specified. | |
| top_k | No | Optional number of top matching documents to return. Defaults to 10. |
Implementation Reference
- src/tools/python-documentation.ts:43-81 (handler)The async execute method implements the core logic of the tool: logs the query, posts to the /api/python/documentation endpoint using groundDocsClient, validates the response, parses each document result into {type: "text", text: ...}, and returns the content array. Errors are logged and rethrown.async execute({ query, library, version, top_k = 10 }: z.infer<typeof this.schema>) { try { console.log(`Fetching Python docs for library: "${library}", query: "${query}" with top_k=${top_k}`); const { data }: { data: { results: string[] } } = await groundDocsClient.post( "/api/python/documentation", { query, library, version, top_k } ); if (!data || !data.results || !Array.isArray(data.results)) { console.error("Invalid response format:", data); throw new Error("Invalid response format"); } return { content: data.results.map((doc: any) => { try { const parsed = typeof doc === "string" ? JSON.parse(doc) : doc; return { type: "text" as const, text: parsed.text, }; } catch (parseError) { console.error("Error parsing document:", parseError); throw parseError; } }), }; } catch (error) { console.error(`Error fetching Python documentation for ${library}:`, error); throw error; } }
- Zod schema defining the input parameters for the tool: required query (string) and library (string), optional version (string), and top_k (number, default 10). Includes descriptions for each field.schema = z.object({ query: z.string().describe("A natural language question (e.g., \"How do I use transformers?\")."), library: z.string().describe("Python library to search documentation for."), version: z.string().optional().describe("Optional Library version (e.g., \"4.46.1\"). Defaults to detected library version if not specified."), top_k: z.number().optional().default(10).describe("Optional number of top matching documents to return. Defaults to 10."), });
- src/index.ts:20-20 (registration)Registers the Python documentation tool by instantiating GetPythonDocumentationTool and calling its register method on the MCP server instance. The BaseTool's register method uses the tool's name, description, schema, and execute function.new GetPythonDocumentationTool().register(server);