Full-Text Search Guide

Last Updated: October 2026 | Reading Time: 10 minutes

Master Memvault's full-text search capabilities powered by SQLite FTS5. Learn search syntax, optimization techniques, and advanced querying patterns to instantly find conversations across your entire archive.

Quick Start
Search Syntax
Advanced Queries
Performance Optimization
Troubleshooting

Quick Start

Basic Search

bash

# Search all conversations
memvault search "folder picker"

# Search with agent filter
memvault search "implementation" --agent claude

# Search by date range
memvault search "bug fix" --from 2026-10-01 --to 2026-10-15

Search from Web UI

Open http://localhost:27961
Click search icon or press /
Type your query
Results appear instantly

How It Works

Memvault uses SQLite FTS5 (Full-Text Search, version 5), which builds an inverted index mapping every word to the documents containing it.

Architecture

Query: "folder picker"
    ↓
Tokenizer splits: ["folder", "picker"]
    ↓
Inverted Index lookup:
  "folder" → [session_abc, session_def, ...]
  "picker" → [session_abc, session_xyz, ...]
    ↓
Intersection: [session_abc]
    ↓
BM25 Ranking
    ↓
Results (sorted by relevance)

Real-world performance: 10-100ms on archives with 10,000+ conversations. According to FTS5 benchmarks, this is 99.6% faster than traditional LIKE queries.

Search Syntax

Single Terms

bash

# Find "implementation"
memvault search implementation

# Case-insensitive by default
memvault search IMPLEMENTATION  # Same results

Multiple Terms

bash

# AND (all terms must match)
memvault search "folder picker"

# Results contain both "folder" AND "picker"

Phrase Search

bash

# Exact phrase
memvault search '"native dialog picker"'

# Must appear in that exact order

Boolean Operators

bash

# OR (either term)
memvault search "folder OR directory"

# AND (both terms, explicit)
memvault search "picker AND dialog"

# NOT (exclude term)
memvault search "picker NOT windows"

# Combine operators
memvault search "folder OR directory AND picker NOT test"

Prefix Search

bash

# Words starting with "impl"
memvault search "impl*"

# Matches: implement, implementation, implicit, etc.

Wildcard Search

bash

# Middle wildcard (slower)
memvault search "*plementation"

# Better: Use prefix when possible
memvault search "impl*"

Column-Specific Search

bash

# Search in specific fields
memvault search "role:user picker"

# Search session titles only
memvault search "title:implementation"

# Search message content
memvault search "content:folder"

Proximity Search

bash

# Terms within N words of each other
memvault search "folder NEAR/5 picker"

# "folder" and "picker" within 5 words

Advanced Queries

Chinese and English Mixed

FTS5's unicode61 tokenizer handles multi-language content:

bash

# Chinese search
memvault search "文件夹选择器"

# Mixed language
memvault search "implement 文件夹 picker"

Filtering by Metadata

bash

# By agent
memvault search "bug fix" --agent claude

# By date range
memvault search "implementation" \
  --from 2026-09-01 \
  --to 2026-09-30

# By session type
memvault search "review" --type code-review

# Combine filters
memvault search "security" \
  --agent claude \
  --from 2026-10-01 \
  --tag important

Context Window

bash

# Show 3 messages before/after match
memvault search "error" --context 3

# Show more context for debugging
memvault search "exception" --context 10

Result Limiting

bash

# Top 10 results
memvault search "picker" --limit 10

# Pagination
memvault search "implementation" --limit 20 --offset 20

Sorting

bash

# By relevance (default, BM25 score)
memvault search "folder picker"

# By date (newest first)
memvault search "picker" --sort date

# By date (oldest first)
memvault search "picker" --sort date --order asc

Output Formats

bash

# Default: human-readable
memvault search "picker"

# JSON output
memvault search "picker" --format json

# JSON with full message content
memvault search "picker" --format json --full

# CSV for analysis
memvault search "bug" --format csv > bugs.csv

Search Strategies

bash

# 1. Search for core concept
memvault search "authentication"

# 2. Note session IDs from results

# 3. Find linked sessions
memvault sessions links abc123

# 4. Read full context
memvault sessions read abc123

Debugging Past Issues

bash

# Find error patterns
memvault search "error OR exception OR failed"

# Filter by time period when bug appeared
memvault search "authentication failed" \
  --from 2026-10-10 \
  --to 2026-10-12

# Check what changed before the error
memvault sessions read <session-before-bug>

Tracking Decisions

bash

# Find architectural discussions
memvault search "architecture OR design OR pattern"

# Find specific technology choices
memvault search "chose Go OR picked TypeScript OR decided"

# Export for documentation
memvault search "decision" --format json | \
  jq '.results[] | {date, session, decision: .snippet}' > decisions.json

Learning from History

bash

# How did I solve similar problems?
memvault search "solved OR fixed OR resolved"

# What have I learned about X?
memvault search "learned OR discovered OR found out"

# Review past code reviews
memvault search "code review" --tag review

Performance Optimization

Index Maintenance

The search index is automatically updated as conversations are captured. For optimal performance:

bash

# Rebuild index after bulk import
memvault index rebuild

# Optimize index (defragment)
memvault db optimize

# Check index health
memvault index stats

Query Performance Tips

Fast Queries (< 10ms):

bash

# Single term
memvault search implementation

# Prefix search
memvault search impl*

# Two terms AND
memvault search "folder picker"

Slower Queries (10-100ms):

bash

# Many OR terms
memvault search "word1 OR word2 OR word3 OR ..."

# Wildcard (not prefix)
memvault search "*mentation"

# Complex proximity
memvault search "word1 NEAR/2 word2 AND word3 NEAR/5 word4"

Avoid (can be slow on large archives):

bash

# Leading wildcard
memvault search "*word"

# Very broad OR with no filters
memvault search "a OR the OR is OR and"

Index Statistics

bash

$ memvault index stats

Index Statistics:
  Total documents: 5,432
  Total tokens: 1,245,678
  Index size: 45.2 MB
  Last rebuild: 2026-10-15 10:30:00
  
Performance:
  Avg query time: 15ms
  95th percentile: 45ms
  99th percentile: 120ms

Large Archive Optimization

For archives with 100,000+ conversations:

bash

# 1. Split by date
memvault search "recent query" --from 2026-10-01

# 2. Use specific filters
memvault search "query" --agent claude --tag project-a

# 3. Archive old sessions to separate vault
memvault sessions archive --older-than 730d \
  --to ~/memvault-archive-2024/

Troubleshooting

No Results Found

Check query syntax:

bash

# Wrong: Unmatched quotes
memvault search "picker

# Right: Proper quotes
memvault search "picker"

Verify index:

bash

# Check if index is healthy
memvault index stats

# Rebuild if needed
memvault index rebuild

Check filters:

bash

# Remove filters to broaden search
memvault search "picker"  # No filters

# vs
memvault search "picker" --agent codex  # Might be too narrow

Slow Searches

Profile query:

bash

# Enable timing
memvault search "query" --verbose

Output:
  Parse: 0.2ms
  Index lookup: 45ms
  Load messages: 12ms
  Ranking: 3ms
  Total: 60.2ms

Optimize:

bash

# Rebuild index
memvault index rebuild

# Optimize database
memvault db optimize

# Check disk speed
time find ~/memvault/sessions/ -type f | wc -l

Index Corruption

Symptoms:

Search returns errors
Results are incomplete
Queries time out

Fix:

bash

# Backup current index
cp -r ~/.memvault/db/ ~/.memvault/db-backup/

# Rebuild from scratch
memvault index rebuild --force

# Verify
memvault search "test query"

Missing Recent Conversations

Cause: Index update lag

Solution:

bash

# Force index update
memvault index update

# Or rebuild
memvault index rebuild

Search API

Programmatic Search

bash

# JSON output for scripting
curl -s http://localhost:27961/api/v1/search \
  -d '{"query":"picker","limit":10}' | jq .

json

{
  "results": [
    {
      "session_id": "abc123",
      "message_id": "0042",
      "score": 0.95,
      "snippet": "...implement folder **picker** with native...",
      "timestamp": "2026-10-15T08:23:45Z"
    }
  ],
  "total": 42,
  "query_time_ms": 15
}

Python Integration

python

import subprocess
import json

def search_memvault(query, limit=10):
    result = subprocess.run(
        ["memvault", "search", query, "--format", "json", "--limit", str(limit)],
        capture_output=True,
        text=True
    )
    return json.loads(result.stdout)

# Use it
results = search_memvault("folder picker")
for hit in results["results"]:
    print(f"{hit['timestamp']}: {hit['snippet']}")

Next Steps

Agent Memory Guide - Let agents learn from search results
Archiving Guide - Optimize what gets indexed
CLI Reference - Complete search command documentation
Configuration - Advanced search settings

Sources

Last reviewed: October 2026 by the Memvault team. Contribute improvements.

Full-Text Search Guide ​

Table of Contents ​

Quick Start ​

Basic Search ​

Search from Web UI ​

How It Works ​

Architecture ​

Search Syntax ​

Single Terms ​

Multiple Terms ​

Phrase Search ​

Boolean Operators ​

Prefix Search ​

Wildcard Search ​

Column-Specific Search ​

Proximity Search ​

Advanced Queries ​

Chinese and English Mixed ​

Filtering by Metadata ​

Context Window ​

Result Limiting ​

Sorting ​

Output Formats ​

Search Strategies ​

Finding Related Conversations ​

Debugging Past Issues ​

Tracking Decisions ​

Learning from History ​

Performance Optimization ​

Index Maintenance ​

Query Performance Tips ​

Index Statistics ​

Large Archive Optimization ​

Troubleshooting ​

No Results Found ​

Slow Searches ​

Index Corruption ​

Missing Recent Conversations ​

Search API ​

Programmatic Search ​

Python Integration ​

Next Steps ​

Sources ​

Full-Text Search Guide

Table of Contents

Quick Start

Basic Search

Search from Web UI

How It Works

Architecture

Search Syntax

Single Terms

Multiple Terms

Phrase Search

Boolean Operators

Prefix Search

Wildcard Search

Column-Specific Search

Proximity Search

Advanced Queries

Chinese and English Mixed

Filtering by Metadata

Context Window

Result Limiting

Sorting

Output Formats

Search Strategies

Finding Related Conversations

Debugging Past Issues

Tracking Decisions

Learning from History

Performance Optimization

Index Maintenance

Query Performance Tips

Index Statistics

Large Archive Optimization

Troubleshooting

No Results Found

Slow Searches

Index Corruption

Missing Recent Conversations

Search API

Programmatic Search

Python Integration

Next Steps

Sources