Problem

At NovaFlow, a data pipeline script emits log lines for each processing step. Implement a function that scans these lines and returns a summary of exception activity by counting failed records, collecting unique error types, and preserving the order of the first occurrence of each error.

Given a list of strings logs, where each string is one log line, return a dictionary with:

1. failed_count: total number of lines that contain an exception
2. error_types: a list of unique exception names in first-seen order
3. latest_message: the message text from the last exception line, or "" if none exist

A line is considered an exception line if it matches this format exactly: ERROR:<ExceptionName>:<message>

<ExceptionName> contains only letters and ends with Error or Exception. The <message> part may contain spaces and punctuation.

Formal Specification

• Input: logs — list of strings
• Output: dictionary with keys failed_count, error_types, and latest_message

Examples

Example 1 Input: logs = ["INFO:start", "ERROR:ValueError:bad row", "ERROR:TypeError:wrong type"] Output: {"failed_count": 2, "error_types": ["ValueError", "TypeError"], "latest_message": "wrong type"} Explanation: Two exception lines appear, both unique, and the last message is from the final error line.

Example 2 Input: logs = ["INFO:ok", "WARN:retry", "INFO:done"] Output: {"failed_count": 0, "error_types": [], "latest_message": ""} Explanation: No line matches the exception format.

Constraints

• 1 <= len(logs) <= 10^4
• 0 <= len(logs[i]) <= 200
• Each log line is an arbitrary string