assignmentsignal

AssignmentSignal helps professors and TAs see how students worked through an assignment, not just the final files they submitted.

It is the academic sibling of InterviewSignal: one assignment code, many students, private GitHub repos, captured AI-assistant workflow, and a professor/TA dashboard for review.

v0 Scope

• Professors or TAs create assignments from the dashboard.
• Students start a session with /assignment <CODE>.
• The session captures prompts, tool calls, file changes, diffs, commits, and a tamper-evident event log.
• If GitHub OAuth is configured on the relay, AssignmentSignal creates a private repo named assignment-<CODE> for the student session.
• Professor/TA GitHub usernames entered during assignment setup are invited to the private repo.
• On /submit, the student sees submission confirmation only. Scores are instructor-only.
• Professors and TAs see the same dashboard from their own machines when connected to the same relay credentials.

Install

pip install assignmentsignal
assignment install

Professor / TA Flow

assignment dashboard

The dashboard handles setup, assignment creation, grading, review, comments, and exports.

For v0, the relay credentials are intentionally blank in local config and Docker defaults. A separate AssignmentSignal relay should be created later and configured with:

assignment configure-relay
assignment configure-api-key

Optional relay environment variables:

RELAY_API_KEY=
GRADING_API_KEY=
GITHUB_CLIENT_ID=
GITHUB_CLIENT_SECRET=
RELAY_BASE_URL=

Student Flow

assignment install
/assignment ASG-1234-AB

Students work normally in their assignment workspace. When done:

/submit

The student receives a confirmation. They do not see a score, rubric assessment, ranking, or grading summary.

Relay

Run a local relay for development:

docker compose up --build

Or directly:

python -m assignment.relay.server

The relay defaults to dev mode if RELAY_API_KEY is blank. Production deployments should set a relay key, GitHub OAuth credentials, and a persistent /data volume.

Notes

See IMPROVEMENTS.md for product and engineering follow-ups discovered during the v0 fork.