import json from typing import Dict, Any, List def parse_postman_collection(content_str: str) -> Dict[str, Any]: """ Parses a Postman Collection (V2.1) JSON string. Returns a structure compatible with the LLM planner and executor. """ try: data = json.loads(content_str) except json.JSONDecodeError as e: raise ValueError(f"Failed to parse Postman collection as JSON: {e}") # Check if it's a Postman collection info = data.get("info", {}) schema = info.get("schema", "") if "postman.com/json/collection" not in schema: raise ValueError("Not a valid Postman collection (missing schema identifier)") title = info.get("name", "Postman Collection") version = "2.1" # Standard for most modern exports all_requests = [] def extract_items(items): for item in items: if "item" in item: # This is a folder, recurse extract_items(item["item"]) elif "request" in item: # This is a request request = item["request"] name = item.get("name", "") # Extract path and method method = request.get("method", "GET").upper() url_obj = request.get("url", {}) path = "" raw_url = "" if isinstance(url_obj, str): raw_url = url_obj # Try to get path from string path = url_obj.split("?")[0] elif isinstance(url_obj, dict): raw_url = url_obj.get("raw", "") path_parts = url_obj.get("path", []) if isinstance(path_parts, list): path = "/" + "/".join([str(p) for p in path_parts]) else: path = str(path_parts) # Extract headers headers = {} for h in request.get("header", []): if not h.get("disabled"): headers[h.get("key")] = h.get("value") # Extract body body_info = request.get("body", {}) body = None if body_info.get("mode") == "raw": try: body = json.loads(body_info.get("raw", "{}")) except: body = body_info.get("raw") elif body_info.get("mode") == "urlencoded": body = {p.get("key"): p.get("value") for p in body_info.get("urlencoded", []) if not p.get("disabled")} elif body_info.get("mode") == "formdata": body = {p.get("key"): p.get("value") for p in body_info.get("formdata", []) if not p.get("disabled")} # Extract description description = request.get("description", "") if isinstance(description, dict): description = description.get("content", "") all_requests.append({ "name": name, "method": method, "path": path, "raw_url": raw_url, "headers": headers, "body": body, "description": description }) extract_items(data.get("item", [])) # Build simplified endpoints for the LLM planner simplified_endpoints = [] for req in all_requests: simplified_endpoints.append({ "path": req["path"], "method": str(req.get("method", "GET")), "summary": str(req.get("name", "Unnamed Request"))[:100] }) # Prepare the "paths" object for the executor, mimicking OpenAPI structure # The executor uses this to find schemas, but for Postman we just provide the request details directly # since there's no formal schema in the same way. paths = {} for req in all_requests: p = req["path"] m = str(req.get("method", "get")).lower() if p not in paths: paths[p] = {} # Structure this to be somewhat compatible with what api_executor expects paths[p][m] = { "summary": req["name"], "description": req["description"], "parameters": [], # Postman doesn't always have typed params "requestBody": { "content": { "application/json": { "example": req["body"] } } } if req["body"] else None } return { "title": title, "version": version, "endpoints": simplified_endpoints, "paths": paths, "components": {}, # No global components in Postman "raw_requests": all_requests # Keep extra detail just in case }