scaledthought 0.3.1 → 0.3.2

package/README.md

@@ -41,7 +41,7 @@ ends, so the link would reach the human only after the payment it is waiting for
 
 ```bash
 npx -y scaledthought@0.3 models --live                 # models serving in your region now
-npx -y scaledthought@0.3 chat -m auto "Reply with OK"  # one real completion
+npx -y scaledthought@0.3 chat -m <id> "Reply with OK"  # one real completion (an id from models --live)
 npx -y scaledthought@0.3 mcp                           # `claude mcp add …` command + JSON config
 npx -y scaledthought@0.3 account                       # plan, region, spend
 npx -y scaledthought@0.3 account ad-measurement off    # opt out of ad measurement

package/dist/index.js

@@ -553,7 +553,7 @@ import { join } from "node:path";
 // package.json
 var package_default = {
   name: "scaledthought",
-  version: "0.3.1",
+  version: "0.3.2",
   description: "Headless open-model inference, run by your agent. Prepaid, pay per token. US or Canada.",
   license: "MIT",
   keywords: [
@@ -677,10 +677,11 @@ var HINTS = {
   video_not_ready: "Poll `video get <id>` until status is completed, or pass --wait.",
   video_expired: "Videos are deleted 24 hours after they complete. Generate it again.",
   video_failed: "Your credits were returned. Retry, or try another model or size.",
-  model_not_found: "Use an id from `models --live`, or model auto.",
+  model_not_found: "Use an id from `models --live`.",
+  model_required: "Pick a live chat model from `models --live` and pass its id with -m.",
   insufficient_scope: "This key lacks the scope. Use the account's default key or create one with the scope.",
   account_suspended: "Stop and tell your human. Contact hello@scaledthought.com.",
-  upstream_error: "Transient model-server error. Retry with backoff, or use model auto to fail over.",
+  upstream_error: "Transient model-server error. Retry with backoff, or try another model from `models --live`.",
   service_unavailable: "Transient. Retry in a few seconds.",
   invalid_body: "Fix the field named in the message. Retrying unchanged will fail again.",
   invalid_request: "Fix the field named in the message. Retrying unchanged will fail again.",
@@ -9809,7 +9810,6 @@ function nextSteps(model, needsCredits) {
   return [
     ...needsCredits ? CREDIT_STEPS : [],
     model ? `Verify with one real completion: ${NPX} chat -m ${model} "Reply with OK" (or POST $ST_BASE_URL/chat/completions with that model)` : `No chat model is live in this region right now. Check again with: ${NPX} models --live`,
-    `Or let the router pick: ${NPX} chat -m auto "Reply with OK"`,
     "Use it from code: any OpenAI SDK with base_url=$ST_BASE_URL and api_key=$ST_API_KEY (both are in .env)",
     `Optional, connect MCP: ${NPX} mcp`,
     ...needsCredits ? [] : [`More credits when needed: ${TOPUP_COMMAND} returns an approval_url for your human (inference is prepaid)`]
@@ -9831,7 +9831,7 @@ async function liveSummary(ctx, region, balanceUsd) {
   return {
     balance_usd: balanceUsd ?? 0,
     needs_credits: needsCredits,
-    next_command: needsCredits ? TOPUP_COMMAND : `${NPX} chat -m ${suggested ?? "auto"} "Reply with OK"`,
+    next_command: needsCredits ? TOPUP_COMMAND : suggested ? `${NPX} chat -m ${suggested} "Reply with OK"` : `${NPX} models --live`,
     newest_models: live.slice(0, 5),
     suggested_model: suggested,
     next_steps: nextSteps(suggested, needsCredits)
@@ -10051,12 +10051,15 @@ var keysRotate = async (ctx, id, o = {}) => api(ctx, "POST", `/v1/keys/${encodeU
 var keysRevokeSelf = (ctx) => api(ctx, "DELETE", "/v1/keys/self");
 var usage = (ctx, o) => api(ctx, "GET", `/v1/usage?days=${o.days}&group_by=${encodeURIComponent(o.by)}`);
 async function chat(ctx, o) {
+  const requested = o.model?.trim();
+  if (!requested)
+    throw new CliError("model_required", 'chat needs a model id: chat -m <id> "your message".');
   const { key, region } = resolveKey(ctx);
   const res = await ctx.fetch(`${inferenceUrl(region, ctx.env)}/chat/completions`, {
     method: "POST",
     headers: { authorization: `Bearer ${key}`, "content-type": "application/json", "user-agent": USER_AGENT },
     body: JSON.stringify({
-      model: o.model,
+      model: requested,
       messages: [{ role: "user", content: o.prompt }],
       stream: true,
       stream_options: { include_usage: true }
@@ -10066,7 +10069,7 @@ async function chat(ctx, o) {
     throw await responseError(res);
   let text = "";
   let usageOut = null;
-  let model = o.model;
+  let model = requested;
   const decoder = new TextDecoder;
   let buf = "";
   for await (const chunk of res.body) {
@@ -10097,7 +10100,6 @@ async function chat(ctx, o) {
   }
   return {
     model,
-    routed_to: res.headers.get("x-st-routed-to"),
     region: res.headers.get("x-st-region"),
     request_id: res.headers.get("x-st-request-id"),
     text,
@@ -11317,7 +11319,9 @@ async function connect(ctx, tool, o = {}) {
   const { region } = resolveKey(ctx);
   const baseUrl = inferenceUrl(region, ctx.env);
   const models = await liveChatModels(ctx);
-  const model = o.model ?? models[0]?.id ?? "auto";
+  const model = o.model ?? models[0]?.id;
+  if (!model)
+    throw new CliError("no_live_model", "No chat model is live in this region right now, so there is nothing to configure. Nothing was changed.", "Check `models --live` again later, or pass --model <id>.");
   if (!/^[A-Za-z0-9][A-Za-z0-9._:/-]{0,127}$/.test(model))
     throw new CliError("bad_argument", "--model isn't a model id.", "Pick one from `models --live`.");
   const host = machineName();
@@ -11345,7 +11349,7 @@ async function connect(ctx, tool, o = {}) {
       daily_budget_usd: k.daily_budget_usd,
       configured: ".env (OPENAI_API_BASE, OPENAI_API_KEY, AIDER_MODEL), mode 0600, git-ignored",
       model: `openai/${model}`,
-      models: ["auto", ...models.map((m) => m.id)],
+      models: models.map((m) => m.id),
       next: "Run `aider` in this directory. Aider may warn that it doesn't know the model's context window; that's harmless."
     };
   }
@@ -11355,7 +11359,7 @@ async function connect(ctx, tool, o = {}) {
   const config = readJsonObject(configPath, "opencode.json");
   const auth = readJsonObject(authPath, authPath);
   const providers = config.provider && typeof config.provider === "object" ? config.provider : {};
-  const modelEntries = { auto: { name: "auto (ScaledThought picks per request)" } };
+  const modelEntries = {};
   for (const m of models)
     modelEntries[m.id] = {
       name: m.id,
@@ -11396,11 +11400,11 @@ async function connect(ctx, tool, o = {}) {
     model: `${PROVIDER_ID}/${model}`,
     models: Object.keys(modelEntries),
     restart_needed: true,
-    next: `Restart OpenCode in this directory and pick a ${PROVIDER_ID}/… model (or ${PROVIDER_ID}/auto).`
+    next: `Restart OpenCode in this directory and pick a ${PROVIDER_ID}/… model.`
   };
 }
 function instructions(tool, baseUrl, model, models, name) {
-  const ids = ["auto", ...models.map((m) => m.id)];
+  const ids = models.some((m) => m.id === model) ? models.map((m) => m.id) : [model, ...models.map((m) => m.id)];
   if (tool === "cursor")
     return {
       tool,
@@ -11956,10 +11960,10 @@ var usage3 = defineCommand({
   run: ({ args }) => run2(() => usage(makeCtx(), { days: num("days", args.days) ?? 7, by: args.by }), args)
 });
 var chat2 = defineCommand({
-  meta: { name: "chat", description: "Send one message. Use -m auto to let the router pick" },
+  meta: { name: "chat", description: 'Send one message to a model: chat -m <id> "…" (ids from `models --live`)' },
   args: {
     ...jsonFlag,
-    model: { type: "string", alias: "m", default: "auto", description: "Model id or auto" },
+    model: { type: "string", alias: "m", description: "Model id, required (from `models --live`)" },
     prompt: { type: "positional", required: true, description: "The message" }
   },
   run: async ({ args }) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "scaledthought",
-  "version": "0.3.1",
+  "version": "0.3.2",
   "description": "Headless open-model inference, run by your agent. Prepaid, pay per token. US or Canada.",
   "license": "MIT",
   "keywords": [