fix(falkordb): escape group_ids in RediSearch fulltext queries

[Milofax]

Summary

• Group IDs like "main" are reserved words in RediSearch and cause syntax errors when used unquoted
• Group IDs with hyphens (e.g., "project-name") also need escaping
• This fix wraps all group_ids in double quotes in the fulltext query

Before: (@group_id:main|project-name)
After: (@group_id:"main"|"project-name")

Problem

Intermittent errors like:

RediSearch: Syntax error at offset 20 near main

Solution

Simple 4-line fix in graphiti_core/driver/falkordb_driver.py:

escaped_group_ids = [f'"{gid}"' for gid in group_ids]
group_values = '|'.join(escaped_group_ids)

Test plan

• Tested with group_id "main" - no more syntax errors
• Tested with group_id "Milofax-infrastructure" - works correctly
• Tested with multiple group_ids - proper OR query generated

🤖 Generated with Claude Code

[Milofax]

Hi team! 👋

Just wanted to follow up on my open PRs. I have 7 PRs waiting for review:

Bug Fixes (high priority):

• fix(falkordb): escape group_ids in RediSearch fulltext queries #1175 - RediSearch group_id escaping (fixes intermittent syntax errors with reserved words like "main")
• fix(mcp): prevent queue worker task garbage collection #1176 - Queue worker task GC fix (prevents worker tasks from being garbage collected mid-execution)
• fix(mcp): pass base_url to Anthropic LLM and OpenAI Embedder clients #1159 - Pass base_url to Anthropic/OpenAI clients (enables custom API endpoints)
• fix(docker): install anthropic package for Claude support #1160 - Install anthropic package in Docker (required for Claude support)

Features:

• feat(mcp): add configurable small_model support for OpenAI provider #1156 - Configurable small_model for lightweight tasks
• feat(mcp): add transport security configuration for remote access #1157 - Transport security config (allowed_hosts, allowed_origins)
• feat(search): add summary_embedding semantic search for nodes #1163 - Summary embedding semantic search

The CI workflows are waiting for approval to run. Would really appreciate if someone could approve them when you get a chance!

Happy to address any feedback or make changes as needed. Thanks! 🙏