Skip to main content
When this applies: your organization already bought ByteSpike, your admin already created your lark account + topped up balance. You got this doc and in ~5 minutes you’ll be calling every Claude / GPT / Gemini / DeepSeek / Doubao model the admin has allowed.

​
1. Install DOSIA

Grab the latest build from the NAS shared drive:
If the first launch reports “file is damaged” or “cannot be opened because it is from an unidentified developer”, run this in Terminal:
Then relaunch DOSIA. That’s macOS Gatekeeper’s default block on un-notarized dmgs — nothing wrong with the app itself.

​
2. Switch to enterprise mode

1

Open Settings

After DOSIA launches, top menu: DOSIA → Settings (shortcut ⌘,).
2

Switch Deploy Mode

In the left nav click Deploy Mode, switch the mode from “Personal” to “Enterprise”.
3

Fill in the enterprise service URL

Paste the URL your admin gave you — the default is:
(If your admin provided an internal-network address or a self-hosted domain, use what the admin said.)
4

Click Save

The blue “Save config” button at the bottom of the page. DOSIA will prompt you to restart — go ahead and restart.
That URL is the DOSIA enterprise web backend (apps/web deployment on Tencent Cloud), separate from the ByteSpike LLM Gateway (https://llm.bytespike.ai) — two services, different hosts, different ports. The backend handles ByteSpike config sync, lark OAuth proxy, and enterprise workspace management; the LLM Gateway handles the actual model calls. Employees don’t need to care about the distinction — admin gives you the URL, you fill it in.If the service URL changes and admin sends a new one, employees just update Settings → Deploy Mode → save. No reinstall needed.

​
3. Enterprise SSO login

1

Confirm admin has granted access

Ask the admin to add your lark employee ID / email to the ByteSpike allowlist. The admin does this in console.bytespike.ai or in the lark group admin backend — you don’t have to do anything.
2

DOSIA triggers lark OAuth automatically

After restart, DOSIA detects enterprise mode and pops the lark authorization page automatically. The first time it opens the browser; once authorized you come back to DOSIA.
3

Logged in

The top-right corner of DOSIA should now show your lark avatar + email.
No lark authorization page appeared / stuck on “initializing”? Jump to Section 6 FAQ to investigate.

​
4. Auto-pick up the allowed models

Once logged in, DOSIA calls ByteSpike automatically to fetch your allowed model list — no manual API key config needed.
1

Open the AI Models tab

Settings → AI Models (left nav).
2

Confirm the 'ByteSpike Shared' provider shows up

The AI Models list should contain a provider called “ByteSpike Shared” (pre-configured by admin under enterprise mode).
3

Confirm the model list

Expand “ByteSpike Shared” — you should see every model the admin has granted you:
  • Claude family: Opus 4.8 / Sonnet 4.6 / Haiku 4.5
  • OpenAI family: GPT-5-4 / GPT-5-4-mini / GPT-5-5, etc.
  • Gemini family: Gemini Pro / Gemini 3.5 Flash
  • DeepSeek: DeepSeek v4 Pro / v4 Flash
  • ByteDance / Doubao: Doubao Seed 2.0 Pro, etc.
  • Image / video: Nano Banana, Veo 3.1, etc. (if admin has granted multimodal capability)
The exact list depends on your group’s permission config — if it doesn’t match, talk to admin.

​
5. Start using it

Send a text message

Pick a model at the top of the main chat box → type message → hit enter. Same interaction across vendors — the ByteSpike gateway routes automatically.

Vision (image input)

Drag an image into the chat box (or paste it). DOSIA calls the image-tools MCP’s analyze_image tool, which invokes a vision model to recognize the image.

Image generation

Ask for an image in the chat (“draw a cyberpunk cat”) — the agent calls the generate_image MCP tool, routing through Nano Banana / GPT Image, etc.

Video generation

Same idea as images — ask for “generate a video”, the agent calls generate_video MCP, routing through Veo 3.1, etc.
Billing is transparent: each model call’s USD cost shows in the “credits” hint below the chat box. Failed calls (5xx / network errors) don’t bill.

​
6. FAQ

Likely causes:
  1. Enterprise service URL is wrong → go back to Settings → Deploy Mode, double-check the URL, save again
  2. Lark grant hasn’t taken effect → ask admin to confirm your employee ID is on the ByteSpike allowlist
  3. SSO authorization expired → DOSIA Settings → My Account, click sign out then sign in again
The org wallet ran out. Screenshot + ask admin — once admin tops up at console.bytespike.ai/billing, it recovers immediately.
Admin hasn’t granted your group access to that model. Ask admin to go to llm.bytespike.ai/admin → Users → your account → Allowed Models, and tick the matching model group.
Multimodal capability is granted separately (independent of text models). Once admin adds image / video capability to your group in console, DOSIA picks up the tool after a restart.
Clear DOSIA’s local state:
Then relaunch DOSIA. If still broken, screenshot + DM the admin on lark.

​
What’s next

  • Want to call the ByteSpike API directly with your own key (not via DOSIA)? See Configure client — base URL + key config per SDK.
  • Want to know how to call each endpoint? See API Reference — protocol / shape / 4xx error codes.
  • Want the full model catalog + pricing? See bytespike.ai/pricing.
Any questions, find your admin or @ByteSpike Ops in the lark group.