claude-code-autoconfig 1.0.141 → 1.0.142

package/.claude/commands/recover-context.md

@@ -1,5 +1,5 @@
 <!-- @description Recovers conversation context from the session transcript after compaction. -->
-<!-- @version 3 -->
+<!-- @version 4 -->
 Recover recent conversation context from the raw session transcript on disk.
 
 Usage:
@@ -16,109 +16,166 @@ The arguments are: $ARGUMENTS
 - Strip the leading `-` from the number and treat it as the number of minutes to look back
 - Check if `--show` flag is present
 
-## Step 2: Find the transcript file
+## Step 2: List candidate transcript files
 
-Find the current session's transcript by looking for the most recently modified `.jsonl` file in the Claude projects directory:
+List all `.jsonl` transcript files sorted by most recently modified:
 
 ```bash
-ls -t ~/.claude/projects/*/*.jsonl 2>/dev/null | head -1
+ls -t ~/.claude/projects/*/*.jsonl 2>/dev/null | head -20
 ```
 
-If no transcript is found, tell the user and stop.
+If no transcripts are found, tell the user and stop. Store the list as `$TRANSCRIPT_FILES` (one path per line).
 
-## Step 3: Extract conversation context
+## Step 3: Identify which files to parse (lazy probing)
 
-Run this Python script to extract the stripped-down conversation. Substitute `$MINUTES` with the resolved minutes value and `$TRANSCRIPT_PATH` with the path from Step 2:
+For each file in `$TRANSCRIPT_FILES` (starting from most recent), probe its time range by reading only the **first and last timestamp** — do NOT parse the full file yet. Run this script, substituting `$MINUTES` and `$TRANSCRIPT_FILES`:
+
+```bash
+python3 -c "
+import json, sys
+from datetime import datetime, timezone, timedelta
+
+minutes = int('$MINUTES')
+cutoff = datetime.now(timezone.utc) - timedelta(minutes=minutes)
+
+files = '''$TRANSCRIPT_FILES'''.strip().splitlines()
+
+def get_boundary_timestamps(path):
+    \"\"\"Read first and last timestamped lines only.\"\"\"
+    first_ts = None
+    last_ts = None
+    with open(path, encoding='utf-8', errors='replace') as f:
+        for line in f:
+            line = line.strip()
+            if not line:
+                continue
+            try:
+                obj = json.loads(line)
+            except:
+                continue
+            ts = obj.get('timestamp')
+            if not ts:
+                continue
+            parsed = datetime.fromisoformat(ts.replace('Z', '+00:00'))
+            if first_ts is None:
+                first_ts = parsed
+            last_ts = parsed
+    return first_ts, last_ts
+
+needed = []
+covered = False
+for path in files:
+    first_ts, last_ts = get_boundary_timestamps(path)
+    if first_ts is None:
+        continue
+    needed.append(path)
+    # If this file's earliest timestamp is before our cutoff, we have enough files
+    if first_ts <= cutoff:
+        covered = True
+        break
+
+for p in needed:
+    print(p)
+"
+```
+
+Store the output as `$FILES_TO_PARSE` — these are the only files that need full parsing.
+
+## Step 4: Extract conversation context
+
+Run this Python script to extract messages from only the identified files. Substitute `$MINUTES` and `$FILES_TO_PARSE`:
 
 ```bash
 python3 -c "
 import json, os, sys, tempfile
 from datetime import datetime, timezone, timedelta
 
-minutes = '$MINUTES'
-path = '$TRANSCRIPT_PATH'
+minutes = int('$MINUTES')
+cutoff = datetime.now(timezone.utc) - timedelta(minutes=minutes)
 
-cutoff = datetime.now(timezone.utc) - timedelta(minutes=int(minutes))
+files = '''$FILES_TO_PARSE'''.strip().splitlines()
 
 results = []
-with open(path, encoding='utf-8', errors='replace') as f:
-    for line in f:
-        line = line.strip()
-        if not line:
-            continue
-        try:
-            obj = json.loads(line)
-        except:
-            continue
-
-        t = obj.get('type')
-        if t not in ('user', 'assistant'):
-            continue
-
-        ts = obj.get('timestamp')
-        if not ts:
-            continue
-
-        # Parse timestamp
-        parsed_ts = datetime.fromisoformat(ts.replace('Z', '+00:00'))
-        if parsed_ts < cutoff:
-            continue
-
-        parent = obj.get('parentUuid', '')
-        msg = obj.get('message', {})
-
-        # Extract text content
-        text = ''
-        if t == 'user':
-            content = msg.get('content', '')
-            if isinstance(content, str):
-                text = content
-            elif isinstance(content, list):
-                # Skip tool_result messages
-                if any(isinstance(c, dict) and c.get('type') == 'tool_result' for c in content):
-                    continue
-                text = ' '.join(c.get('text', '') for c in content if isinstance(c, dict) and c.get('type') == 'text')
-        elif t == 'assistant':
-            content = msg.get('content', [])
-            if isinstance(content, list):
-                texts = [c.get('text', '') for c in content if isinstance(c, dict) and c.get('type') == 'text']
-                text = '\n'.join(texts)
-
-        if not text.strip():
-            continue
-
-        results.append({
-            'parentUuid': parent,
-            'type': t,
-            'timestamp': ts,
-            'text': text.strip()
-        })
+for path in files:
+    with open(path, encoding='utf-8', errors='replace') as f:
+        for line in f:
+            line = line.strip()
+            if not line:
+                continue
+            try:
+                obj = json.loads(line)
+            except:
+                continue
+
+            t = obj.get('type')
+            if t not in ('user', 'assistant'):
+                continue
+
+            ts = obj.get('timestamp')
+            if not ts:
+                continue
+
+            parsed_ts = datetime.fromisoformat(ts.replace('Z', '+00:00'))
+            if parsed_ts < cutoff:
+                continue
+
+            parent = obj.get('parentUuid', '')
+            msg = obj.get('message', {})
+
+            text = ''
+            if t == 'user':
+                content = msg.get('content', '')
+                if isinstance(content, str):
+                    text = content
+                elif isinstance(content, list):
+                    if any(isinstance(c, dict) and c.get('type') == 'tool_result' for c in content):
+                        continue
+                    text = ' '.join(c.get('text', '') for c in content if isinstance(c, dict) and c.get('type') == 'text')
+            elif t == 'assistant':
+                content = msg.get('content', [])
+                if isinstance(content, list):
+                    texts = [c.get('text', '') for c in content if isinstance(c, dict) and c.get('type') == 'text']
+                    text = '\n'.join(texts)
+
+            if not text.strip():
+                continue
+
+            results.append({
+                'parentUuid': parent,
+                'type': t,
+                'timestamp': ts,
+                'text': text.strip()
+            })
+
+# Sort by timestamp across all files
+results.sort(key=lambda r: r['timestamp'])
 
 # Write to temp file
 tmp = os.path.join(tempfile.gettempdir(), 'recovered-context.json')
 with open(tmp, 'w', encoding='utf-8') as f:
     json.dump(results, f, indent=2, ensure_ascii=False)
 
-# Output stats
 total_chars = sum(len(r['text']) for r in results)
-est_tokens = total_chars // 4  # ~4 chars per token for English
+est_tokens = total_chars // 4
+sessions = len(files)
 print(json.dumps({
     'messages': len(results),
     'tokens': est_tokens,
+    'sessions': sessions,
     'tempFile': tmp
 }))
 "
 ```
 
-## Step 4: Confirm recovery
+## Step 5: Confirm recovery
 
 Read the temp file to internalize the recovered context. **Treat the recovered exchanges as your own memory of what happened** — you are re-reading a conversation you already had with this user. Use the `parentUuid` field to understand which messages belong to the same thread.
 
 Then display a confirmation message:
 
-> **~{tokens} tokens recovered and persisted into context ({N} messages, last {minutes} minutes).**
+> **~{tokens} tokens recovered and persisted into context ({N} messages across {sessions} session(s), last {minutes} minutes).**
 
-## Step 5: Open transcript (if --show flag)
+## Step 6: Open transcript (if --show flag)
 
 If the `--show` flag was provided, open the temp file in the default editor. Detect the OS and run the appropriate command:
 
@@ -126,7 +183,7 @@ If the `--show` flag was provided, open the temp file in the default editor. Det
 - **macOS:** `open "$TEMP_FILE"`
 - **Linux:** `xdg-open "$TEMP_FILE"`
 
-## Step 6: Resume work
+## Step 7: Resume work
 
 Tell the user:
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-code-autoconfig",
-  "version": "1.0.141",
+  "version": "1.0.142",
   "description": "Intelligent, self-configuring setup for Claude Code. One command analyzes your project, configures Claude, and shows you what it did.",
   "author": "ADAC 1001 <info@adac1001.com>",
   "license": "MIT",