inq-review 0.1.2

Terminal-native web UI for AI-assisted code and document review

inq

A lightweight, terminal-native web UI for reviewing and annotating code, PDFs, and markdown — together with AI.

inq is a self-hosted server. You run it on a remote machine, forward the port locally, and open it in a browser. It gives you a file explorer, a scrollable viewer for code / PDFs / markdown, and an annotation surface that humans and AI agents both write into. Threads stay anchored to the line (or page) you asked about, and persist per source.

┌─────────┬────────────────────────┬────────────────┐
│ FILES   │ src/inq/sandbox.py     │ THREADS        │
│  inq/   │  56  class Gitignore…  │  sandbox.py:56 │
│   ai.py │  57    def __init__(   │  › what does…  │
│   ▸ …   │  58      …             │  · It walks …  │
└─────────┴────────────────────────┴────────────────┘

Press @ on any line (or with a range selected) to ask the AI. Press # to leave a note — same anchor, no model roundtrip. Responses stream in, render as markdown, and stay threaded per anchor.

Quickstart

pip install inq-review                    # PyPI distribution; CLI is `inq`
inq init                                  # pick provider, paste API key, choose model
inq --port 9090 --root .                  # serve the current directory

On your local machine:

ssh -L 9090:localhost:9090 user@remote

Open http://localhost:9090.

Features

• Anchored threads. Every question or note pins to a line range (code/markdown) or page (PDF). Threads persist per source at ~/.config/inq/threads/.
• Two annotation modes. @ asks the AI (streaming, threaded follow-ups). # saves a comment-only note (no AI call). Both render side-by-side in the panel.
• Multi-format viewers. CodeMirror for source, pdf.js for PDFs (cmaps + standard fonts bundled so math/ligatures render correctly), react-markdown for .md files.
• Multi-source. Browse the local filesystem by default; click "Open Drive folder" in the UI to pick a Google Drive folder via OAuth.
• Three providers. Anthropic, Google (Gemini), OpenAI — all with prompt-cached file context so follow-ups are cheap.
• Markdown export. "Copy md" buttons in the UI, or inq notes --root <dir> for the same output from the CLI.

Providers

Provider	Models	Where to mint a key
Anthropic	claude-sonnet-4-6, claude-opus-4-7, …	https://console.anthropic.com/settings/keys
Google	gemini-2.5-pro, gemini-2.5-flash, …	https://aistudio.google.com/apikey
OpenAI	gpt-4o, gpt-4o-mini, gpt-4.1, …	https://platform.openai.com/api-keys

Credential resolution order: --api-key-file flag → <PROVIDER>_API_KEY env var → ~/.config/inq/config.toml (mode 0600) → macOS keychain (service inq).

Building from source

If you want to hack on the frontend, you'll need Node 20+:

git clone https://github.com/sutra-lab/inq && cd inq
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"

cd web
npm install
npm run build           # produces src/inq/web/ — FastAPI mounts it at /
# or `npm run dev` for hot-reload on :5173 (backend stays on :9090)

Why

Reading code, specs, and documents deeply means asking questions inline — but most workflows force you into a chat window, losing context. Cursor and VS Code remote are heavy editor environments; Claude Code lives in the terminal, not in a scrollable viewer. inq is the small, focused tool that fits between them: zero-install on the client, browser-native, and built around asking questions in place.

See project-spec.md for the full design.

License

MIT — see LICENSE.