aai-ws

The public Answer.AI workspace baseline: the repos and starting configuration for our aai-coding harness, managed by fastws. Each project has its own Git checkout, sharing one uv environment with editable Python installs.

Install uv and Git, and configure GitHub SSH access. The starter builds exhash and rgapi from source, so Rust and native build tools are also required (Xcode Command Line Tools on macOS).

uvx --from 'fastws-cli>=0.0.14' ws-setup AnswerDotAI/aai-ws ~/aai-ws
source ~/aai-ws/.venv/bin/activate
cd ~/aai-ws

ws-setup clones this repo, creates a Python 3.13 environment, installs fastws, and runs ws-sync. Choose any new destination directory; existing paths are refused rather than overwritten. It prints the activation command without changing your shell startup files. Then open Claude Code or codex in the workspace and ask it to follow aai-coding/SETUP.md.

The same command accepts your own workspace repo. Answer.AI team members with access to the private workspace use:

uvx --from 'fastws-cli>=0.0.14' ws-setup AnswerDotAI/private-ws ~/aai-ws

• repos.txt: the baseline checkouts. It includes fastws, aai-coding, llmdojo, ipykernel-helper, exhash, rgapi, safecmd, and slopometer. Other Python dependencies install as packages.
• pyproject.tmpl: starting Python and workspace settings, copied only when the root pyproject.toml is missing.
• .gitignore: keeps checkouts, personal additions, and generated environment metadata out of this repo.

Your personal additions go in gitignored repos-local.txt, using the same format as the shared list:

owner/project
owner/private ~/private

The root pyproject.toml, pyrightconfig.json, and uv.lock are local state. fastws creates and maintains the workspace metadata; it does not overwrite existing settings from the template. Project checkouts retain their own tracked metadata.

Activate the workspace environment, then work in a project checkout.

ws-status
ws-sync
ws-add owner/project

ws-sync first pulls this workspace repo, clones missing entries from the shared and local lists, pulls project updates, and syncs the environment. It upgrades dependencies at most once per day. ws-add writes only to the local list. A checkout removed from the shared baseline is never deleted; existing root checkouts become personal additions. ws-remove handles personal repos and refuses shared baseline members.

To add a personal package without a source checkout, run uv add package-name from the workspace root. It is recorded in your local pyproject.toml and retained by later syncs.

Edit and commit repos.txt to change the shared baseline. Use pyproject.tmpl for defaults for new workspaces, not for overwriting existing personal configuration. Downstream workspaces can bring in public baseline changes through normal Git merges.

This public repo starts with fresh history. The previous private AnswerDotAI/aai-ws was renamed to AnswerDotAI/private-ws. Existing clones of that private repo must update their remote to the new name before fetching or pushing; the public repo is a different repository, not a visibility change.