IChat

IChat Extension Guide

Language: English | 简体中文

What IChat Does

IChat is a Chrome side panel extension for context-aware AI chat. Its core idea is intelligent context capture with a seamless Q&A flow, while staying fully local-first so users do not need to route their prompts or API keys through a project-owned backend.

More concretely, IChat watches the region you hover over or the text you select, intelligently builds a context object called FlowContext, opens the side panel, and turns the whole context-to-chat handoff into a much smoother experience.

Example scenario:

On ordinary web pages, the shortcut supports two capture paths:

Current Capabilities

Right-click Selected Text

Select text on a web page or in Chrome’s PDF viewer, right-click the selection, and choose Capture selected text with IChat. IChat opens the side panel and uses the text Chrome supplies as the new FlowContext. Auto-send on sends it through the configured provider flow; auto-send off leaves a draft ready for your question. This path uses plain text, requires no vision model or OCR, and does not touch the clipboard. It captures the selected text and available source metadata only, not surrounding paragraphs. The menu appears only when Chrome provides a text selection and follows IChat’s UI language setting.

Online PDFs

In Chrome’s native PDF viewer, use the extension action or capture shortcut to attach the currently visible tab area as an image. A vision-capable model is required. Zoom and scroll to the content you want before capturing; the image can include viewer controls and portions of multiple pages. This captures neither the full document nor the exact text selection. Auto-send follows your existing setting; turn it off in Settings to review the attachment before sending. Local file:// PDFs and PDFs embedded within ordinary web pages are not covered by this fallback.

How To Use IChat

1. Load / Install The Extension

From GitHub Releases:

  1. open the Releases page
  2. download the latest ichat-chrome-mv3-prod-<version>.zip
  3. extract the zip to a local folder
  4. open chrome://extensions/
  5. enable Developer mode
  6. choose Load unpacked
  7. select the extracted extension folder

For local development:

  1. run npm install
  2. run npm run build
  3. open chrome://extensions/
  4. enable Developer mode
  5. choose Load unpacked
  6. select build/chrome-mv3-prod

2. Open The Side Panel

You can open IChat by:

3. Settings

General

LLM Providers

Open Settings and add one of the following:

FlowContext

Voice Input And Composer

The composer combines the text area and its controls in one panel. The arrow sends the draft; during generation it becomes a square that stops the response. Enter sends, Shift+Enter adds a line break, and confirming text with an input method does not send.

Choose a speech provider in Settings → STT Provider, independently of the chat model:

Chrome and Fun-ASR follow the interface language (English or Simplified Chinese); MOSS detects the spoken language automatically. Text is inserted at the cursor or replaces the selected text, preserving the rest of the draft. Editing is paused during capture/transcription. While recording, clicking Send stops dictation, waits for transcription, and sends the merged draft once. Clicking the microphone to stop only fills the draft. Failed, cancelled, or timed-out recognition does not send the draft; leaving the chat or opening Settings cancels a pending send. Existing installs keep Chrome as the default; changing the speech provider preserves chat settings and keys.

If microphone access is denied or the side panel cannot display a permission prompt, click Allow microphone in the error message. Grant access in the extension tab, return to the chat, and click the microphone again. Unsupported browsers, device failures, and unavailable speech services show an error while keeping the draft available for typing.

Microphone permission belongs to IChat’s extension origin, not each website beside the side panel. If no speech is recognized, click Test microphone in the error message. The extension tab can run a local test for up to ten seconds, showing the default input device and a sound level meter. It does not record or upload audio. A moving meter confirms input signal, not necessarily intelligible speech. If it stays still, check the default microphone at chrome://settings/content/microphone, the device mute switch, and the Windows input volume. Stopping the test or leaving the tab releases the microphone.

The selected speech provider processes audio independently of the chat model: the browser’s speech service for Chrome, Alibaba Cloud for Fun-ASR-Flash, or MOSS for MOSS transcription. Fun-ASR and MOSS recordings stay only in memory until processing ends or is cancelled; they are sent directly as WAV audio over HTTPS. IChat does not persist recordings. Closing or leaving the chat stops capture and cancels transcription; listening never restarts automatically. You can revoke microphone access in Chrome or system settings. To remove a speech provider key, select that provider and clear its field and click Save API Key.

Local Storage And Data Flow

In the current implementation:

For more detail, see the Privacy Policy.

Permissions Summary

IChat currently needs Chrome extension capabilities related to:

How Capture Works

When you trigger IChat on a normal http or https page, the extension tries to gather relevant context from the current page.

Capture may include:

If an image target cannot be resolved directly, IChat may use a screenshot-based fallback for the visible target area.