Walk Tutor
An open-source, fully offline voice-based DSA (Data Structures and Algorithms) coaching tool. It enables users to solve algorithm problems through spoken responses without keyboard input, supporting local deployment and voice interaction throughout the practice session.
Timeline (last 90 days)
Walk Tutor is a fully offline voice-based data structures and algorithms (DSA) coach that reads questions aloud to the user, has the user verbally explain their solution approach, and then provides hints or follow-up questions—all without requiring a keyboard.
Walk Tutor's tech stack includes faster-whisper (STT), Gemma 3 1B running locally via Ollama (the coach's brain), pyttsx3 (TTS), and a lightweight Python loop for voice activity detection.
The entire Walk Tutor system runs on the author's Windows laptop with no cloud dependencies whatsoever.
In the first version, the author had Gemma 3 1B directly judge whether the user's answer was correct, and found that the small model was "too agreeable"—when a user claimed to solve a bracket-matching problem using a reversed linked list, the model still praised it as a good approach.
The author adopted a division of labor where "code acts as judge and the model acts as communicator": core solution keywords are predefined for each question, code logic checks whether the user's answer hits the keywords and makes the judgment, while Gemma is only responsible for composing the feedback text based on that judgment.
The keyword-matching judgment method has limitations: a user who happens to mention the correct keywords within an incorrect solution may still be marked as passing.
During testing, the author fixed the problem of "long answers being cut off mid-way" by increasing the pause detection time and the maximum answer length.
All Facts (7)
Walk Tutor is a fully offline voice-based data structures and algorithms (DSA) coach that reads questions aloud to the user, has the user verbally explain their solution approach, and then provides hints or follow-up questions—all without requiring a keyboard.
50%UnverifiedWalk Tutor's tech stack includes faster-whisper (STT), Gemma 3 1B running locally via Ollama (the coach's brain), pyttsx3 (TTS), and a lightweight Python loop for voice activity detection.
50%UnverifiedThe entire Walk Tutor system runs on the author's Windows laptop with no cloud dependencies whatsoever.
50%UnverifiedIn the first version, the author had Gemma 3 1B directly judge whether the user's answer was correct, and found that the small model was "too agreeable"—when a user claimed to solve a bracket-matching problem using a reversed linked list, the model still praised it as a good approach.
50%UnverifiedThe author adopted a division of labor where "code acts as judge and the model acts as communicator": core solution keywords are predefined for each question, code logic checks whether the user's answer hits the keywords and makes the judgment, while Gemma is only responsible for composing the feedback text based on that judgment.
50%UnverifiedThe keyword-matching judgment method has limitations: a user who happens to mention the correct keywords within an incorrect solution may still be marked as passing.
50%UnverifiedDuring testing, the author fixed the problem of "long answers being cut off mid-way" by increasing the pause detection time and the maximum answer length.
50%