mushi-debug

Mushi Mushi — Debug Guide

Diagnostic Checklist (run in order)

1. Check SDK connectivity

# Run the doctor command — checks connectivity and key validity
mushi doctor

Expected output: all checks green. If MUSHI_API_KEY fails, re-run mushi login.

2. Check Sentry for edge function errors

Use the Sentry MCP:

get_sentry_issues project=mushi-be

Look for errors in story-mapper, test-gen-from-story, pdca-runner, inventory-propose.

3. Story mapping failures

Symptom: "Map from live app" shows failed status

Diagnose:

1. Open the console → Inventory → Discovery → Recent crawls
2. Expand the failed run to see error_message

Common causes:

• No Firecrawl API key: go to Settings → API Keys → Add a Firecrawl key
• URL is behind auth: use Browserbase provider + configure session cookies
• Claude quota exhausted: add a backup Anthropic key with mushi keys add

4. TDD test generation failures

Symptom: mushi tdd gen <storyId> returns error

Diagnose:

# Check that a story exists in the accepted inventory
mushi tdd pending  # lists qa stories pending review

Common causes:

• Story id doesn't exist in accepted inventory → accept the inventory proposal first
• All LLM keys exhausted → mushi keys list then add a backup key

5. QA stories not running

Symptom: Tests never execute on schedule

Diagnose:

mushi tdd pending  # check if stories are stuck in pending_review

• If approval_status = pending_review: approve them with mushi tdd approve <id>
• If enabled = false: toggle enabled in console → QA Coverage → story detail
• If automation_mode = approve: the story requires manual enable

6. API key quota / rate limit

Symptom: Fix attempts fail with "All LLM keys exhausted"

Diagnose:

mushi keys list

Look for status=quota_exhausted with a cooldown time.

Fix:

# Add a backup key
mushi keys add --provider anthropic --key sk-ant-... --label "backup2" --priority 200

# Or add an OpenAI key as fallback
mushi keys add --provider openai --key sk-... --label "openai-backup" --priority 300

7. Inventory not showing up

Symptom: Inventory page shows no active inventory after accepting a proposal

Check in DB (Supabase MCP):

SELECT id, status, source, created_at 
FROM inventory_proposals 
WHERE project_id = '<your-project-id>'
ORDER BY created_at DESC LIMIT 5;

SELECT id, status, created_at
FROM inventories
WHERE project_id = '<your-project-id>'
ORDER BY created_at DESC LIMIT 3;

If inventories is empty but inventory_proposals has accepted rows, the accept flow failed — check Sentry.

8. PDCA auto-improve not triggering

Symptom: Failing tests never get rewritten

Check:

1. Cron is registered: run SELECT jobname, schedule FROM cron.job; in Supabase SQL editor
2. The stories have source=test_gen_from_story and automation_mode in ['auto', 'review']
3. Trigger manually: mushi tdd improve

9. Edge function logs

Use the Supabase MCP:

get_logs service=api
get_logs service=postgres

Or via CLI:

supabase functions logs story-mapper
supabase functions logs test-gen-from-story
supabase functions logs pdca-runner

Common Error Messages

Error	Cause	Fix
No Firecrawl API key configured	Missing BYOK key	Add key in Settings → API Keys
All LLM keys exhausted	All Anthropic/OpenAI keys hit quota	Add backup key with mushi keys add
Story not found in test-gen	Story id not in accepted inventory	Accept the inventory proposal first
byok_keys_provider_slug_check	Invalid provider slug	Use: anthropic, openai, firecrawl, browserbase, cursor
relation story_map_runs does not exist	Migration not applied	Run pending migrations on remote

Applying Pending Migrations

If you get "relation does not exist" errors, migrations may not be deployed:

cd packages/server
supabase db push --db-url postgresql://...

Or use the Supabase MCP apply_migration for each file in order:

1. 20260602000000_byok_multikey_pool.sql
2. 20260602000001_story_map_runs.sql
3. 20260602000002_qa_stories_tdd_columns.sql
4. 20260602000003_pdca_qa_improve_cron.sql