RTFD (Read The F*****g Docs)

Instructions

Implementation Reference

• src/RTFD/providers/dockerhub.py:407-435 (handler)

  The MCP tool handler for 'fetch_dockerfile'. It invokes the private _fetch_dockerfile method and serializes the response using serialize_response_with_meta.
  async def fetch_dockerfile(image: str) -> CallToolResult: """ Fetch the actual Dockerfile used to build a Docker image. USE THIS WHEN: You need to see exactly how an image is built (base image, installed packages, configuration). BEST FOR: Understanding image composition, security analysis, or learning how to build similar images. Attempts to find Dockerfile link in DockerHub description and fetches from source (usually GitHub). Useful for: - Seeing what base image is used - Identifying installed packages and dependencies - Understanding build process and optimizations - Security auditing (what's included in the image) - Learning Dockerfile best practices from official images Note: Not all images have publicly accessible Dockerfiles. Many official images do. Args: image: Docker image name (e.g., "nginx", "python", "postgres") Returns: JSON with Dockerfile content, source URL, and metadata (or error if not found) Example: fetch_dockerfile("nginx") → Returns Dockerfile from nginx GitHub repository """ result = await self._fetch_dockerfile(image) return serialize_response_with_meta(result)
• src/RTFD/providers/dockerhub.py:224-318 (helper)

  Core helper function implementing the logic to fetch Dockerfile: retrieves image metadata, parses full_description with regex for GitHub blob links to Dockerfile, converts to raw URL, downloads content.
  async def _fetch_dockerfile(self, image: str) -> dict[str, Any]: """Fetch Dockerfile for an image by parsing its description for GitHub links. Args: image: Image name (e.g., 'nginx', 'python') Returns: Dict with Dockerfile content or error """ import re try: # 1. Get image metadata to find the description metadata = await self._fetch_image_metadata(image) if "error" in metadata: return { "image": image, "error": metadata["error"], "source": None, } # 2. Get the full description (README) # Note: _fetch_image_metadata might not return full_description if it uses the summary endpoint, # but let's check if we need to make a separate call or if we can rely on what we have. # The previous investigation showed we need to hit the repository endpoint which _fetch_image_metadata does. # However, the 'readme' field in _fetch_image_metadata comes from 'readme' key in JSON. # Let's verify if 'full_description' is available in the payload of _fetch_image_metadata. # Looking at _fetch_image_metadata implementation: # url = f"{self.DOCKERHUB_API_URL}/repositories/{repo_path}/" # This endpoint returns 'full_description' usually. # But _fetch_image_metadata returns a dict with specific keys. # We need to access the raw description. # Let's re-fetch to be sure we get the full description text to parse if "/" not in image: repo_path = f"library/{image}" else: repo_path = image url = f"{self.DOCKERHUB_API_URL}/repositories/{repo_path}/" async with await self._http_client() as client: resp = await client.get(url) resp.raise_for_status() data = resp.json() full_desc = data.get("full_description", "") # 3. Find GitHub Dockerfile links # Pattern: https://github.com/[owner]/[repo]/blob/[ref]/[path/to/]Dockerfile # We want to capture the whole URL github_pattern = r"https://github\.com/[a-zA-Z0-9_.-]+/[a-zA-Z0-9_.-]+/blob/[a-zA-Z0-9_.-]+(?:/[a-zA-Z0-9_.-]+)*/Dockerfile" matches = re.findall(github_pattern, full_desc) if not matches: return { "image": image, "error": "No GitHub Dockerfile link found in image description", "source": "dockerhub_description", } # Use the first match (often the 'latest' or most prominent one) # Ideally we'd match against a specific tag if provided, but for now we take the first one. dockerfile_url = matches[0] # 4. Convert to raw GitHub URL # From: https://github.com/user/repo/blob/ref/path/Dockerfile # To: https://raw.githubusercontent.com/user/repo/ref/path/Dockerfile raw_url = dockerfile_url.replace("github.com", "raw.githubusercontent.com").replace( "/blob/", "/" ) # 5. Fetch the Dockerfile async with await self._http_client() as client: resp = await client.get(raw_url) resp.raise_for_status() content = resp.text return { "image": image, "content": content, "size_bytes": len(content.encode("utf-8")), "source": raw_url, "found_in_description": True, } except Exception as exc: return { "image": image, "error": f"Failed to fetch Dockerfile: {exc!s}", "source": None, }
• src/RTFD/providers/dockerhub.py:20-34 (registration)

  Provider metadata method registers 'fetch_dockerfile' in tool_names list conditionally if fetching is enabled.
  def get_metadata(self) -> ProviderMetadata: tool_names = ["search_docker_images", "docker_image_metadata"] if is_fetch_enabled(): tool_names.append("fetch_docker_image_docs") tool_names.append("fetch_dockerfile") return ProviderMetadata( name="dockerhub", description="DockerHub Docker image search and metadata", expose_as_tool=True, tool_names=tool_names, supports_library_search=False, # DockerHub search is image-centric, not lib-doc required_env_vars=[], optional_env_vars=[], )
• src/RTFD/providers/dockerhub.py:436-444 (registration)

  In get_tools method, the fetch_dockerfile handler function is assigned to tools dict conditionally and returned for MCP registration.
  tools = { "search_docker_images": search_docker_images, "docker_image_metadata": docker_image_metadata, } if is_fetch_enabled(): tools["fetch_docker_image_docs"] = fetch_docker_image_docs tools["fetch_dockerfile"] = fetch_dockerfile return tools