This panel provides a central place to review what aclarai has done — or is scheduled to do — across all tiers. It shows claim metadata, automation state, file-level job status, and lets the user pause/resume processing.

It does not replace Obsidian’s Markdown views — it overlays structure and status without interfering with editing workflows.

graph TD
    A[📁 File / Block Index] --> B[🔎 Claim Detail View]
    D[🧭 Claim Explorer] --> B
    C[⚙️ Automation Status + Controls]

This is a panel-based layout: select a file/block → inspect claim metadata → see or control automation status.

Shows recent files, extracted blocks, and processing state.

At the top of the interface:

[ 📁 File View ] [ 🔍 Claim Explorer ]

The default view shows the file/block index. Clicking Claim Explorer activates is (section 4).

User can click any row to inspect its content or metadata.

When a block is selected, show:

• The original Markdown block

• Metadata:

  • entailed_score, coverage_score, decontextualization_score
  • aclarai:id and ver=
  • Linked concepts

• Link status:

  • SUPPORTS_CONCEPT → [[Concept]]
  • Graph edge exists: ✅ / ❌

Actions (if allowed):

• [Trigger reprocessing]
• [Unlink concept]
• [View in Obsidian]

Shows whether aclarai is running in live mode, paused, or selectively disabled.

[ ⏸️ Pause Automation ]

• Sets .aclarai_pause flag in vault root
• Disables all job scheduling except manual actions
• Button toggles to:

[ ▶️ Resume Automation ]

Each scheduled job shows last run time and next scheduled run:

• Pulls state from .aclarai_pause, settings/aclarai.config.yaml, and runtime job registry
• Should support job-specific pause overrides (e.g., scheduler.claim_linking.enabled: false)
• Displays Vault-relative paths for links to Markdown blocks

You're absolutely right — thanks for calling that out.

Given the actual structure of design_review_panel.md (as provided), here's the properly integrated update, including:

• ✅ An updated mermaid diagram
• ✅ Correct insertion as section 4
• ✅ Sequential and structural consistency

Replace the existing diagram at the top with:

graph TD
    A[1. 📁 File / Block Index] --> B[2. 🔎 Claim Detail View]
    D[4. 🧭 Claim Explorer] --> B
    C[3. ⚙️ Automation Status + Controls]

This mode offers a searchable, cross-file view of all extracted claims — useful for identifying failed evaluations, low-quality scores, or topics of interest.

At the top of the interface:

[ 📁 File View ] [ 🔍 Claim Explorer ]

The default view shows the file/block index (Section 1). Clicking Claim Explorer activates the global view.

A simple text box lets users filter the claim list by:

• Text content (e.g., "GPT-4")
• Score values (e.g., "null", "0.67")
• Planned keywords:
  • failed: show claims with any null score
  • low: show claims with low geomean
  • concept:<term> (future)

Example:

[ 🔍 filter claims... _______ ] 

Filtered claims appear as a scrollable list, each line showing:

"OpenAI released GPT-4 in 2023."
  entailed: 0.92 | coverage: 0.84 | decontextualized: 0.87

"It increased rapidly."
  entailed: null | coverage: 0.67 | decontextualized: 0.42

Clicking any row navigates to its full detail view (same as selecting a block).