Chat completions, FIM code completion, model listing and account balance via the DeepSeek API.
DeepSeek ships in the w6w first-party pack. It declares 4 actions, 2 health checks, and the host runs its code in a sandbox that never sees the credential.
io.w6w.deepseekDeepSeek runs chat completions from a workflow against an OpenAI-Chat-Completions-shaped endpoint, with two DeepSeek-specific controls layered on top: a thinking switch (its reasoning models default to thinking mode ON, unlike OpenAI’s opt-in reasoning_effort) and a reasoningEffort knob (low/high/max) that shapes how much chain-of-thought the model spends before answering. The reply carries the final content and, in thinking mode, a separate reasoning_content field — the model’s visible chain-of-thought — so a workflow can log or discard it independently of the answer.
A Beta FIM (Fill-In-the-Middle) completion action covers code-completion-shaped prompts (a prefix and, optionally, a suffix) against a base model — the same request shape editor/IDE integrations use, exposed here as a workflow step rather than requiring a full chat-formatted prompt. Models can be listed to keep a workflow’s model choice current: DeepSeek has renamed its model line before, and the API’s own /models endpoint is the only source that will not go stale. Account balance can be read directly, so a workflow can check headroom before a batch of calls rather than discovering a 402 mid-run.
A fit for adding an LLM step to a workflow — especially one that wants visibility into the model’s reasoning process, or a lower-cost alternative to a frontier chat model for summarizing, classifying or drafting code.
Three routes to the same 4 actions. The Workflow tab is generated from DeepSeek's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
fim-complete Fill-in-the-middle text completion for a prefix (and optional suffix) — the shape code editors use, without a chat-formatted prompt.
get-balance Read the account's current balance (granted + topped-up), by currency.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying DeepSeek's real ids.
{
"manifestVersion": "2",
"name": "deepseek-example",
"steps": [
{
"id": "chat-complete",
"uses": {
"app": "io.w6w.deepseek",
"action": "chat-complete",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"model": "<model>",
"messages": "<messages>"
}
}
]
}chat-complete fim-complete get-balance list-models Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for DeepSeek, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to DeepSeek. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "chat-complete",
payload: {
model: "<value>",
messages: "<value>",
// thinking: "<value>",
// reasoningEffort: "<value>",
// maxTokens: "<value>",
// responseFormat: "<value>",
// stop: "<value>",
// temperature: "<value>",
// topP: "<value>",
// tools: "<value>",
// toolChoice: "<value>",
// logprobs: "<value>",
// topLogprobs: "<value>",
// userId: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value); npm install -g @w6w/cli w6w run conn_YOUR_CONNECTION_ID --action chat-complete --payload '{"model":"<value>","messages":"<value>"}' Give an AI agent DeepSeek — without giving it DeepSeek's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.deepseek#chat-complete",
"input": {
"model": "<model>",
"messages": "<messages>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's DeepSeek connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
DeepSeek's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. DeepSeek itself is MIT, and the runtime that executes it is source-available (FSL).
DeepSeek declares its own checks, so its health is a property of the app rather than something the host guesses at.
Atlassian Statuspage rollup for deepseek.statuspage.io, with per-component detail for the API and web-chat services. Unauthenticated and unsigned.
The account's current balance by currency, and whether DeepSeek itself reports it as sufficient for API calls.