Troubleshooting
jq is requiredbd/ Dolt errors- An MCP server won’t start
- Ollama models are never used
- Dolt sync conflict (team)
- Wrong git host / token
- Container run fails
- Windows: something wants to install MinGW / a native build fails
pre-pushblocks a push- My project is on
master, and the branching rules never apply - A project’s hooks/scripts feel out of date, or a fix in a new aiflow release isn’t showing up
jq is required
Install jq — aiflow install-deps does. aiflow reads/writes .aiflow/config.json with it.
bd / Dolt errors
aiflow install-deps installs both Beads and Dolt. Check the server with bd dolt status. If the
embedded server won’t start, bd dolt start (or restart your shell) and retry.
An MCP server won’t start
Run aiflow doctor. Confirm the underlying tool is installed:
- cocoindex-code →
ccc(uv tool install 'cocoindex-code[full]') - graphify →
graphify(uv tool install graphifyy && graphify install) - git-host MCP → Podman or Docker running (GitHub MCP) and the token env in
.envmatchesremote.tokenEnv.
Ollama models are never used
Enable router and run aiflow shell --router. Confirm .aiflow/router-config.json lists your
models and ollama list has them locally (aiflow ollama pull).
Dolt sync conflict (team)
bd dolt pull to merge, resolve, then bd dolt push. Never force-push. See
Team collaboration.
Wrong git host / token
Re-run aiflow change-settings; ensure .env has the variable named in remote.tokenEnv;
aiflow doctor prints the resolved remote + host MCP.
Container run fails
Ensure Podman or Docker is installed and its daemon/machine is running. Force one with
AIFLOW_CONTAINER=podman|docker docker/run.sh "<task>".
Windows: something wants to install MinGW / a native build fails
You’re missing the Windows prerequisites. aiflow itself runs in PowerShell + Git Bash, but anything
that compiles native code (C/C++, node-gyp, Python C-extensions, uv-built tools) belongs in
WSL, not MinGW/MSYS2 — two parallel toolchains on one machine means ABI mismatches, PATH
collisions between the two sh.exes, and builds that differ from CI.
Fix, in order — the full walkthrough is in Installation → Windows prerequisites:
- Enable Intel VT-x / AMD SVM Mode in the BIOS/UEFI. Check with
Get-ComputerInfo -Property HyperVRequirementVirtualizationFirmwareEnabled. - In an admin PowerShell:
wsl --install, then reboot. wsl --install -d Ubuntu, start it once, confirmwsl -l -vshows VERSION 2.- Inside WSL:
sudo apt update && sudo apt install -y build-essential(pluscmake,python3-dev, … as your stack needs).
aiflow doctor reports all four of these on Windows.
pre-push blocks a push
That’s the branching model. Use a proper branch/PR. See Workflows.
My project is on master, and the branching rules never apply
aiflow governs the mainline by name: main. branching.json, the pre-push hook,
aiflow release and aiflow hotfix all reference main, so a repo sitting on master is simply
ungoverned. Since 0.6.x, aiflow init forces main at git init time, and aiflow apply renames
an existing master → main and prints the remote-migration commands:
git push -u origin main
# switch the default branch to 'main' in your host's settings, then:
git push origin --delete master
If both master and main exist, aiflow refuses to touch either — merge or delete master
yourself, then re-run aiflow apply. To suppress the rename entirely: AIFLOW_NO_BRANCH_RENAME=1.
A project’s hooks/scripts feel out of date, or a fix in a new aiflow release isn’t showing up
Two separate steps: aiflow update brings the installed CLI (AIFLOW_HOME) up to the
latest release; aiflow project-update then refreshes this project’s mechanical scripts
(.aiflow/*, .claude/hooks/*, docker/run.*, .github/scripts/*) from those templates and re-applies
config. Updating the CLI alone doesn’t touch existing projects — that’s deliberate, so a
project never changes underneath you without asking. Compare the project’s stamped
version (meta.aiflowVersion in .aiflow/config.json) against aiflow version
(the installed CLI) to check whether it’s behind.
Still stuck? Open an issue with repro steps, your OS,
and the relevant aiflow doctor output.