inbox-process

Inbox Processing Pipeline

Process all notes in the vault's Inbox directory using AI classification, frontmatter injection, and PARA routing.

Process

Step 1: Check prerequisites

Verify AI provider is configured:

of doctor --vault <name>

Check the AI section of the output. If AI provider is not configured, the user needs to set up [ai] in vault.toml and add API keys to ~/.obsidian-forge/.env.

Step 2: Count inbox items

of doctor --vault <name>

Note the inbox item count from the output.

Step 3: Process inbox

of process-all --vault <name>

This reads all files from the Inbox directory, calls the configured AI provider for classification, injects frontmatter, and moves each file to the appropriate PARA folder.

Step 4: Verify and report

of doctor --vault <name>

Check that inbox count is now 0 (or reduced).

Report in Korean:

1. 처리 전 인박스 — number of items before processing
2. 분류 결과 — how many went to each PARA folder
3. 처리 후 인박스 — remaining items (failures or partial)
4. 실패 항목 — any notes that could not be classified (with reasons)

Anti-Rationalization

Excuse	Rebuttal	What to do instead
"AI provider should be fine"	Config drift, key rotation happen silently	Verify with of doctor before processing
"100% success is guaranteed"	API limits, malformed notes can cause failures	Check before/after inbox counts
"Just re-run if some fail"	Repeated failures indicate systemic issue	Investigate failures before re-running

Evidence Required

• AI provider status from of doctor
• Before/after inbox item counts
• Specific file paths for any failures
• PARA folder destination for each processed note

No evidence = not done.

Red Flags

• AI provider error → check ~/.obsidian-forge/.env and [ai] config in vault.toml
• All notes fail → API key may be invalid or provider endpoint is down
• Notes routed to wrong folders → AI model may need prompt tuning
• Inbox is empty before processing → nothing to do, inform user