Nicolas Cantu 088eab84b7 Platform docs, services, ia_dev submodule, smart_ide project config

- Add ia_dev submodule (projects/smart_ide on forge 4nk)
- Document APIs, orchestrator, gateway, local-office, rollout
- Add systemd/scripts layout; relocate setup scripts
- Remove obsolete nginx/enso-only docs from this repo scope

2026-04-03 16:07:58 +02:00

2.1 KiB

Raw Blame History

agent-regex-search-api

Local HTTP API on 127.0.0.1 for regex search over files using ripgrep (rg). Results are returned as structured JSON.

This is not the closed-source “instant grep” index described in Cursor’s article (Recherche regex rapide); it is a local, open approach (ripgrep) with the same high-level goal: fast agent-oriented code search. For monorepos at extreme scale, consider adding Zoekt or another indexed backend later (see feature doc).

Prerequisites

rg available in PATH (e.g. sudo apt install ripgrep on Debian/Ubuntu).

Environment

Variable	Required	Description
`REGEX_SEARCH_TOKEN`	yes	`Authorization: Bearer <token>` on every request except `GET /health`.
`REGEX_SEARCH_ROOT`	no	Absolute base directory searches are confined to (default `/home/ncantu/code`).
`REGEX_SEARCH_HOST`	no	Bind address (default `127.0.0.1`).
`REGEX_SEARCH_PORT`	no	Port (default `37143`).

Endpoints

GET /health — liveness; includes configured root path.
POST /search — JSON body:
- pattern (string, required): Rust regex passed to ripgrep.
- subpath (string, optional): path relative to REGEX_SEARCH_ROOT (no .., no absolute paths).
- maxMatches (number, optional): cap on matches (default 500, max 50000).
- timeoutMs (number, optional): kill rg after this many ms (default 60000, max 300000).

Response: { root, target, matches: [{ path, lineNumber, line }], truncated, exitCode }.

Ripgrep exit code 1 means “no matches” and is still returned as 200 with an empty matches array when no other error occurred.

Run

npm install
npm run build
export REGEX_SEARCH_TOKEN='…'
npm start

Risks

ReDoS: pathological regexes can burn CPU until timeoutMs. Keep timeouts conservative for shared hosts.
Scope: all readable files under target that ripgrep traverses may be searched; align REGEX_SEARCH_ROOT with policy.

License

MIT.

2.1 KiB Raw Blame History Unescape Escape