recent-history

Steps to retrieving the context of recent chat history

Step 1: Prepare the history file

If there is in your system prompt, just use it, then go to Step 2. Otherwise, run this command which should run in current project root dir.

bunx repomix --header-text "file path parent dir: .claude/sessions/" --quiet --no-dot-ignore --no-gitignore --no-git-sort-by-changes --no-file-summary --no-directory-structure .claude/sessions/ -o .claude/sessions/history.xml >/dev/null 2>&1 && echo "history exist" || echo "no history file"

If the command output is "no history file", just quit, because there are no session history to look up. Otherwise, please go to step 2.

Step 2: Query history (sessions are ordered old→new in the file)

If user intended meant for last session and our context is sufficient, just focus on

Quick Overview: List all sessions with line numbers

# List all session titles with line numbers (shows where each session starts)
rg -n '<file path="' .claude/sessions/history.xml | head -20

# Cleaner output: line number + filename only
rg -no '<file path="[^"]+' .claude/sessions/history.xml | sed 's/<file path="/  /'

# List all session titles (no line numbers)
rg -o '<file path="[^"]+' .claude/sessions/history.xml | sed 's/<file path="//'

Get Recent Sessions (recommended for "what have we done")

# Get last 3 session titles with line numbers
rg -no '<file path="[^"]+' .claude/sessions/history.xml | tail -3

# Get last N lines of history (recent content, adjust N as needed)
tail -500 .claude/sessions/history.xml

# Get content of last session only
tac .claude/sessions/history.xml | sed -n '1,/<file path=/p' | tac

Extract Specific Line Range (after finding line numbers)

# Extract lines 1500-1800 from history file
sed -n '1500,1800p' .claude/sessions/history.xml

# Extract from line 1500 to end
sed -n '1500,$p' .claude/sessions/history.xml

# Extract 100 lines starting from line 1500
sed -n '1500,1600p' .claude/sessions/history.xml

Keyword Search in Recent History (with line numbers and context)

# Search keyword in last 1000 lines with line numbers and 3 lines context
tail -1000 .claude/sessions/history.xml | rg -n -C 3 "keyword"

# Search with more context (5 lines before/after)
tail -1000 .claude/sessions/history.xml | rg -n -C 5 "keyword"

# Case-insensitive search with context
tail -1000 .claude/sessions/history.xml | rg -n -i -C 3 "keyword"

# Show only 2 lines after match (useful for user/agent messages)
tail -1000 .claude/sessions/history.xml | rg -n -A 2 "keyword"

Full File Keyword Search (with line numbers and context)

# Basic keyword search with line numbers
rg -n "keyword" .claude/sessions/history.xml

# With context lines (3 before/after) and line numbers
rg -n -C 3 "keyword" .claude/sessions/history.xml

# Case-insensitive with context
rg -n -i -C 3 "keyword" .claude/sessions/history.xml

# Search multiple keywords (OR) with line numbers
rg -n -e "keyword1" -e "keyword2" .claude/sessions/history.xml

# Search with both keywords (AND) with line numbers
rg -n "keyword1" .claude/sessions/history.xml | rg "keyword2"

# Show only matching parts with line numbers (useful for long lines)
rg -n -o ".{0,50}keyword.{0,50}" .claude/sessions/history.xml

# Count matches per file
rg -c "keyword" .claude/sessions/history.xml

# Search with regex pattern and line numbers
rg -n "error|warning|failed" .claude/sessions/history.xml

# Show 5 lines after each match (good for reading full responses)
rg -n -A 5 "keyword" .claude/sessions/history.xml

# Show 2 lines before and 5 after
rg -n -B 2 -A 5 "keyword" .claude/sessions/history.xml

Extract Specific Session by Date/Topic

# Get current date in session filename format (YYYYMMDD)
date +%Y%m%d

# Find today's sessions
rg -A 100 "<file path=\"$(date +%Y%m%d)" .claude/sessions/history.xml

# Find session by specific date pattern (e.g., 20251212)
rg -A 100 '<file path="20251212' .claude/sessions/history.xml | head -150

# Find sessions from this week
date +%Y%m%d  # Today
date -v-1d +%Y%m%d  # Yesterday (macOS)
date -d "yesterday" +%Y%m%d  # Yesterday (Linux)

# Find session by topic keyword in filename
rg -o '<file path="[^"]*litellm[^"]*' .claude/sessions/history.xml

Recommended Workflow

1. "What have we done?" → Use tail -500 or list recent session titles with line numbers
2. "Find specific topic" → Use keyword search with rg -n -C 3
3. "Continue previous work" → Search for keywords in recent history (tail -1000 | rg -n -C 3)
4. "Extract full session" → Get line numbers, then use sed -n 'START,ENDp'

Practical Example: Complete Workflow

# Step 1: Find where "litellm" was discussed (with line numbers)
rg -n -C 2 "litellm" .claude/sessions/history.xml | head -20
# Output: 162:<file path="20251212-215454-session-nixhmlitellm...
#         165:@nix/hm/litellm/ we have a litellm running well...

# Step 2: Get all session boundaries with line numbers
rg -no '<file path="[^"]+' .claude/sessions/history.xml | tail -5
# Output: 5175:<file path="20251219-101237-summary.txt
#         5179:<file path="20251219-104149-session...
#         5245:<file path="20251219-104733-session...
#         5334:<file path="20251219-112255-session...

# Step 3: Extract specific session (lines 5245-5333)
sed -n '5245,5333p' .claude/sessions/history.xml

# Step 4: Search within that extracted session
sed -n '5245,5333p' .claude/sessions/history.xml | rg -n -C 2 "keyword"