history-context

File-based history search for Claude Code. Find which sessions edited specific files and retrieve conversation context.

Why This Plugin?

Unlike semantic search (episodic-memory), this plugin uses Claude Code's native file-history tracking to know exactly which sessions edited which files. No git commits needed, works immediately.

Features

• File-to-session mapping: Find all sessions that edited a specific file
• Instant results: Uses a cached index of file-history, not full-text search on every query
• Full context retrieval: Read conversation content with key decisions
• Session IDs: Returns complete UUIDs for reference

Installation

From GitHub

# Add the marketplace
/plugins marketplace add gentritbiba/history-context

# Install the plugin
/plugins install history-context

That's It!

On first session start, the plugin automatically:

• Copies the CLI tool to ~/.claude/commands/
• Builds the file-history index

To manually rebuild the index later:

~/.claude/commands/claude-history.py index --rebuild

Usage

Via Skill (Automatic)

The file-history skill triggers automatically when you:

• Ask about past work on a specific file
• Want to understand why code was written
• Reference "what we did before" on a feature

Via Slash Command

/file-history src/components/AuthForm.tsx

Via Agent

Use Task tool with subagent_type="history-context"
Prompt: "Find context for [file/feature]"

Via CLI

# Find sessions that edited a file
~/.claude/commands/claude-history.py file "AuthForm.tsx"

# Search conversations for a topic
~/.claude/commands/claude-history.py search "authentication"

# Get session details
~/.claude/commands/claude-history.py session <id> --read

# Rebuild index
~/.claude/commands/claude-history.py index --rebuild --stats

How It Works

• Indexes file-history: Scans ~/.claude/file-history/ to build file→session mapping (cached for speed)
• Maps sessions to projects: Cross-references with ~/.claude/projects/
• Reads conversations: Extracts messages from session .jsonl files
• Returns structured context: Summary, decisions, session IDs, paths

Why "Instant Results"?

• First run/rebuild: Scans all file-history and builds an index at ~/.claude/file-session-index.json
• Subsequent queries: Reads from cached index (~5ms) instead of scanning all files (~500ms+)
• Auto-refresh: Cache rebuilds if older than 1 hour

vs episodic-memory

Aspect	history-context	episodic-memory
Query type	File-specific	Semantic/conceptual
Accuracy	Exact (file tracking)	Fuzzy (embedding search)
Speed	Indexed	Full search
Best for	"What sessions touched this file?"	"How did we handle auth patterns?"

Use history-context when you have a specific file or feature in mind. Use episodic-memory for broad conceptual searches across all projects.

License

MIT