[codex] add persistent graph query server

[jayhack]

Summary

• Add graph-sitter query for foreground JSONL sessions that parse once and answer multiple graph requests.
• Add graph-sitter query-server for coding agents: start, run, status, reload, and stop keep the parsed graph in a local background HTTP server.
• Make server queries reload-aware with --reload-if-stale by default, so agent edit/query loops refresh only after repo changes.
• Document the persistent session and background server workflow in the CLI reference.

Validation

• uv run ruff check src/graph_sitter/cli/commands/query/main.py src/graph_sitter/cli/cli.py tests/unit/cli/commands/graph/test_graph_commands.py
• uv run ty check --output-format=concise src/graph_sitter/cli/commands/query/main.py src/graph_sitter/cli/cli.py tests/unit/cli/commands/graph/test_graph_commands.py
• uv run pytest tests/unit/cli/commands/graph/test_graph_commands.py
• uv run --frozen pre-commit run --show-diff-on-failure --color=always --files site/app/cli/page.tsx src/graph_sitter/cli/cli.py src/graph_sitter/cli/commands/query/main.py tests/unit/cli/commands/graph/test_graph_commands.py
• PATH="$HOME/.nvm/versions/node/v22.19.0/bin:$PATH" npm --prefix site run build

Manual Smoke

• Started a background server for /Users/jayhack/CS/CLICKUP/sd with --language typescript --backend rust --fallback error; initial parse reported 12,045 files, 74,566 symbols, and 13.99s parse time.
• Queried runInference without repeating parser flags; the existing server was discovered by repo path and returned results without reparsing.
• Ran a callgraph query for packages/ai-agent-harness/src/inference.ts.runInference; it resolved runInference -> streamInference from the in-memory server.
• Stopped the smoke-test server after verification.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
graph-sitter	Ready	Preview, Comment	Jun 24, 2026 5:09pm

[kopekC]

Lgtm