hansjone/dsh-im-ops
1
0
Fork 0
mirror of https://github.com/hansjone/dsh-im-ops.git synced 2026-10-09 03:03:24 +08:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: support image prompts across IM channels

Browse source
This commit is contained in:
xmanrui 2026-08-19 02:56:14 +08:00
parent 99a877c640
commit 65c842c045
40 changed files with 3806 additions and 274 deletions
Download patch file Download diff file Expand all files Collapse all files

8
README.en.md
View file

@ -38,6 +38,8 @@ Connect IM bots to DeepSeek Harness by scanning a QR code, using an App Manifest
Other IM platforms can be added through the same channel-adapter structure.
All nine built-in channels can send JPEG, PNG, and WebP images, plus GIFs sent as image files, with optional captions to Harness. Each image is limited to 5 MB, and images in one message are limited to 20 MB in total.
## Installation
```sh
@ -56,19 +58,19 @@ Feishu, QQ, DingTalk, and WeCom each provide two entry points. The blue **QR acc
Telegram and Discord do not provide an official QR flow for creating bots, so their pages expose only the key-marked **Manual access** action and request a Bot Token. Generate the Telegram token with BotFather; an existing webhook must be removed by its current service before Bot API long polling can receive updates. Generate the Discord token on the Developer Portal's Bot page, invite the bot to the target server, and grant View Channel, Send Messages, and Read Message History. The plugin reads DMs and server messages that explicitly mention the bot, so it does not request the privileged Message Content intent.
Slack provides Manifest-assisted creation with dual-Token access. Choose **Start setup**, copy the bundled App Manifest, open Slack's create page, and select **From a manifest**. Under **Basic Information → App-Level Tokens**, generate an App Token with `connections:write`; then install the app to the workspace under **OAuth & Permissions** to obtain the Bot Token. The plugin validates both Tokens before opening Socket Mode. Slack has no official QR-based bot-creation flow. Both Tokens are sent only to the local Harness Host and stored through its protected credential provider; status responses and bot lists never return them.
Slack provides Manifest-assisted creation with dual-Token access. Choose **Start setup**, copy the bundled App Manifest, open Slack's create page, and select **From a manifest**. Under **Basic Information → App-Level Tokens**, generate an App Token with `connections:write`; then install the app to the workspace under **OAuth & Permissions** to obtain the Bot Token. The plugin validates both Tokens before opening Socket Mode. Slack has no official QR-based bot-creation flow. Image access uses the Manifest's `files:read` scope; Slack Apps installed before this upgrade must be reinstalled or reauthorized to receive it. Both Tokens are sent only to the local Harness Host and stored through its protected credential provider; status responses and bot lists never return them.
WhatsApp exposes only **QR access**. On the phone, open **WhatsApp → Settings → Linked devices → Link a device**, then scan the QR code shown by Harness. No Meta console, Cloud API, Webhook, Phone Number ID, or Access Token is required. Linked-device state stays under `~/.dsh/integrations/dsh-whatsapp/auth`; the browser receives only the one-time QR code and redacted account status. Personal accounts can use WhatsApp's **Message yourself** chat directly; the plugin suppresses only its own exact reply message IDs to prevent reply loops.
Use a dedicated WhatsApp number for the bot when possible. Linking a personal account makes DMs sent to that account eligible Harness input; group messages trigger only when they mention or reply to the linked account. Limit the number to trusted contacts, and remove the device from both Harness and the phone's **Linked devices** list when it is no longer used.
For DingTalk QR binding, scan with an account that belongs to an enterprise or organization and can create bots, then choose **Create a new bot** on the authorization page. If DingTalk reports that the account has not joined an organization, create one or switch to an account that has, then scan again. There is no second local sender-approval flow: the bot's DingTalk visibility is its inbound access scope, so restrict it to trusted organizations, groups, or members.
For DingTalk QR binding, scan with an account that belongs to an enterprise or organization and can create bots, then choose **Create a new bot** on the authorization page. If DingTalk reports that the account has not joined an organization, create one or switch to an account that has, then scan again. There is no second local sender-approval flow: the bot's DingTalk visibility is its inbound access scope, so restrict it to trusted organizations, groups, or members. Image downloads add no separate scope, but they depend on the bot's existing **企业内机器人发送消息权限** permission. A manually bound app without that permission can receive an image callback but cannot exchange it for a temporary download URL.
For WeCom QR binding, scan with an account that belongs to an enterprise and can create or manage bots, then confirm creation of the intelligent bot in the mobile app. This creates a WeCom intelligent bot; it does not sign the plugin into a personal WeChat account. For both QR and credential binding, restrict the bot's WeCom visibility to trusted enterprise members and group chats.
QQ QR binding uses Tencent's official QQBot v2 flow. Tencent's default authorization page labels the integration as a third-party bot. Scanning creates a QQ Open Platform bot; it does not give the plugin direct control of a personal QQ account. QR binding accepts only the scanner's messages. Manual credentials cannot identify a scanner, so the bot's QQ Open Platform visibility becomes its inbound access scope.
Feishu QR binding records the scanner as an allowed user. Manual credentials cannot identify a scanner, so the Feishu application's visibility becomes its inbound access scope. Restrict the application to trusted tenants, groups, or members.
Feishu QR binding records the scanner as an allowed user. Manual credentials cannot identify a scanner, so the Feishu application's visibility becomes its inbound access scope. Restrict the application to trusted tenants, groups, or members. Reading user-sent images requires the tenant scope `im:message:readonly`. Newly QR-created apps request it automatically; apps created before this upgrade must add the scope in the Feishu developer console, publish a version, and complete any required administrator approval.
Each bot maintains an independent Harness workspace. A newly connected bot records the Harness Host process's current working directory (`process.cwd()`) as its default; the path is persisted and does not change when the Host is later restarted from another directory. Every bot card shows the current path and lets it be edited.