Jamailar/beav-deepseek-harness

Beav Creator brings Beav's social-media AI creation workflows into DeepSeek Harness. It exposes Beav workspaces, projects, durable content tasks, and verified article, image, audio, and video artifacts through Cordis services, Harness tools, background jobs, slash commands, session events, and Web client contributions. It does not use MCP.

Install

Requirements:

• Beav 2.7.4 or later
• DeepSeek Harness 0.1.0-rc.6
• Node.js 22 or later

Install into a Harness profile:

dsh plugin --profile web add beav-creator-dsh

Start that profile, open Settings → Plugins → Beav Creator, and select Connect Beav. Harness opens beav://connect/authorize; Beav starts if needed and shows the requested permissions. After you approve in Beav, the connector completes a five-minute, single-use PKCE exchange and stores the resulting credential under BEAV_CREATOR_TOKEN. No token is copied through the clipboard, URL, browser client, chat, session event, or tool result.

If Beav is not running, an operation that needs it opens beav://open once and performs a bounded health retry. You can override the local endpoint or credential reference in the profile patch:

- update:
    id: beav
    config:
      endpoint: http://127.0.0.1:41700
      tokenRef: BEAV_CREATOR_TOKEN
      autoLaunch: on-demand
      requestTimeoutMs: 30000

Only loopback HTTP endpoints are accepted.

Use

• Ask naturally: “Use Beav to turn the latest knowledge in my launch workspace into an article, cover images, and a 60-second video.”
• Type @beav to select a stable workspace or project reference.
• Use /beav connect, /beav status, /beav open, /beav workspaces, /beav new, /beav import, or /beav save for direct actions without a model turn.
• Long-running work appears as a native Harness job and a replayable Beav task card.
• Completion is accepted only after Beav reports durable artifact read-back verification and each returned artifact can be read back through the Creator Gateway.

Beav retains ownership of AI orchestration, knowledge access, media generation, approvals, billing, persistence, and artifact verification. The connector cannot auto-approve paid or destructive actions.

Trust and privacy

This package runs as trusted local Host code. Pairing exposes only a local approval/status surface before authorization; creator operations still require a revocable client credential. Deep links contain a random request ID and PKCE challenge, never a token or verifier. The connector connects only to Beav on loopback, requests bounded summaries, and never transfers raw media bytes through model tool parameters. The connector source is public so the community can audit the integration. The Beav desktop application, creator runtime, knowledge engine, media pipeline, and commercial services remain proprietary and are not included in this repository or npm package.

This is a community plugin and is not an official DeepSeek product or endorsement.

Development

Clone the official Harness checkout at the pinned compatibility revision beside this repository:

git clone https://github.com/deepseek-ai/deepseek-harness.git ../deepseek-harness
git -C ../deepseek-harness checkout 47f943859bef60e4160492346772ded9b24f765a
pnpm -C ../deepseek-harness install
pnpm -C ../deepseek-harness run build:lib
pnpm install
pnpm check

The published tarball contains prebuilt lib/ files and treats every Harness runtime package as a peer dependency. Users do not run a build or prepare script during plugin installation.

License

MIT. This license applies to the connector only, not to the Beav application or services.