Troubleshooting

Sync issues

”config not found — run devrecall config init first”

You’re running a command before initial setup. Either run the wizard:

devrecall setup

or just initialize the config + database:

devrecall setup --quick

“Slack sync failed — token may have expired”

Slack tokens occasionally need re-authorization (e.g., after a workspace policy change). Re-run:

devrecall auth slack

Sync is slow on first run

First sync pulls sync.initial_lookback_days (default 90 days) of history per source. After that, syncs are incremental and fast.

For Git, large monorepos with deep history can be slow on the first walk — narrow your scope with explicit git.repos instead of broad scan_paths.

LLM issues

”Ollama not running”

ollama serve   # start Ollama
ollama list    # confirm your model is pulled

If you don’t want to use Ollama, switch to BYOK:

devrecall config set llm.provider openai
devrecall auth openai

“OpenAI API quota exhausted”

DevRecall falls back through your configured chain. Check what happened:

devrecall status

To unblock immediately, configure a fallback (local Ollama is free):

{
  "llm": {
    "provider": "openai",
    "fallback": [{ "provider": "ollama", "model": "gemma4" }]
  }
}

Standups read like a list, not natural language

You either don’t have an LLM configured, or all configured providers fell through to the template-based fallback. Confirm with devrecall status. To always force natural language, configure at least one working LLM (local Ollama is free).

Port conflicts

Port 3725 already in use

Another DevRecall instance is probably running. Check:

curl -s http://127.0.0.1:3725/api/status

If something else is on the port:

devrecall config set server.port 9876
devrecall serve --port 9876

The desktop app reads server.port from the config and will follow the new port automatically.

Database issues

”database is locked”

SQLite uses WAL mode, but a hung process can still hold a lock. Find it:

lsof ~/.devrecall/devrecall.db

Kill the process and retry. If devrecall serve is the holder, use devrecall daemon instead of multiple serve instances.

Database is huge

Periodic activity accumulates. Prune old raw activities while keeping summaries:

devrecall prune --older-than 1y --keep-summaries

Storage is roughly 50 MB / year before pruning, much less after.

Update issues

”devrecall is out of date and cannot run”

A critical security update has shipped and the relay’s kill switch fired. The fix is the message — run:

devrecall update

This is the only command exempt from the kill switch.

Auto-update fails on macOS Homebrew install

Homebrew-installed binaries are managed by Homebrew. Update via:

brew update && brew upgrade devrecall-cli
# or for the cask:
brew upgrade --cask devrecall

Privacy / data residency

”Did anything get sent to a DevRecall server?”

Two short answers:

• OAuth callback for Slack, Google, Jira, Linear — yes, the auth code passes through the relay (deleted within 60s, see Privacy model).
• Version kill-switch check — once an hour, the CLI asks the relay whether the running build has been flagged unsafe. No identifying data is sent.
• Telemetry — none. DevRecall does not collect telemetry.

LLM calls go directly to the provider (BYOK) or localhost:11434 (Ollama). DevRecall never proxies them.

Where to file bugs

github.com/pavelpilyak/devrecall/issues

Include:

• devrecall --version
• devrecall status output
• The exact command and the full error