jeo-code 0.9.15 → 0.9.17

package/CHANGELOG.md

@@ -1,6 +1,7 @@
 # Changelog
 
 All notable changes to **jeo-code** are documented here.
+
 The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
@@ -8,6 +9,20 @@ The README mirrors the latest 5 entries — regenerate with `bun run changelog:s
 
 ## [Unreleased]
 
+## [0.9.17] - 2026-08-18
+_Live model discovery already existed (`listProviderModels`/`discoverModels`), but only `jeo auth login openai` ever called it — logging into Anthropic or Antigravity left the account pinned to the maintained static catalog snapshot until the next unrelated live-discovery call happened to run._
+
+### Added
+- **`jeo auth login anthropic` / `jeo auth login antigravity` now fetch and cache the account's live model list immediately after login**, matching the existing OpenAI/Codex behavior (`src/commands/auth.ts`) — the prior `reportOpenAiCodexModels` helper was generalized into `reportLiveModels(provider)` (driven by a `LIVE_MODEL_REPORT_PROVIDERS` set covering `openai`/`anthropic`/`antigravity`) so a freshly logged-in Anthropic or Antigravity account immediately shows its real, current model ids (e.g. newly released Claude/Gemini models not yet in the bundled static catalog) instead of waiting for the next launch's background discovery. The result is persisted via the existing `writeModelCache` so the very next `jeo` launch rehydrates it before any network call. Verified live: `listProviderModels("anthropic", …)` and `listProviderModels("antigravity", …)` both returned real, larger-than-catalog model sets (10 and 11 ids respectively) against live logged-in accounts, and the merged cache file reflected all three providers correctly.
+
+
+## [0.9.16] - 2026-08-18
+_Routing looked broken while it was actually working: the idle status bar and `/compact`/`/handoff` never reflected which model `routePrompt` actually routed to._
+
+### Fixed
+- **`/route on`'s footer/status bar and `/compact`/`/handoff` model budget kept showing the pre-routing default model instead of what was actually routed** (`src/commands/launch.ts`) — the idle status bar (`statusBarLine`) and `/compact`'s/`/handoff`'s `activeModel` all recomputed `sessionModel || defaultModel` from scratch every render, which has no idea what `routePrompt` actually picked for the last turn. With routing on, a turn would genuinely route to (and reply from) e.g. `claude-sonnet-5`, but the footer kept showing `claude-opus-5` (the unrouted default) the whole time — indistinguishable from routing silently doing nothing. Fixed by tracking `lastActiveModel` (the model that actually served the most recently completed turn) and reading it first everywhere the idle/compaction UI previously assumed `sessionModel || defaultModel` was the truth.
+
+
 ## [0.9.15] - 2026-08-18
 _The 0.9.14 fix wired `promptInput` into the `$a $b …` one-shot skill-chain's `io.input`, but that chain runs before the interactive REPL's own `readline.Interface` exists — so a bundled workflow skill (`$deep-interview`, with or without `--tmux`) invoked directly from the command line still froze forever on its first question._
 

package/README.ja.md

@@ -287,10 +287,10 @@ CI は `.github/workflows/npm-publish.yml` で公開します — GitHub リリ
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
 - **[Unreleased]**
+- **[0.9.17]** (2026-08-18) — Live model discovery already existed (`listProviderModels`/`discoverModels`), but only `jeo auth login openai` ever called it — logging into Anthropic or Antigravity left the account pinned to the maintained static catalog snapshot until the next unrelated live-discovery call happened to run.
+- **[0.9.16]** (2026-08-18) — Routing looked broken while it was actually working: the idle status bar and `/compact`/`/handoff` never reflected which model `routePrompt` actually routed to.
 - **[0.9.15]** (2026-08-18) — The 0.9.14 fix wired `promptInput` into the `$a $b …` one-shot skill-chain's `io.input`, but that chain runs before the interactive REPL's own `readline.Interface` exists — so a bundled workflow skill (`$deep-interview`, with or without `--tmux`) invoked directly from the command line still froze forever on its first question.
 - **[0.9.14]** (2026-08-04) — A `$a $b … [intent]` skill-chain invocation of a bundled workflow skill (deep-interview/ralplan/team/ultragoal) hung forever with no error and no prompt.
-- **[0.9.13]** (2026-08-03) — The provider-login browser tab now actually goes away — by itself on macOS, by a working button everywhere else.
-- **[0.9.12]** (2026-08-02) — Fixes the boxed prompt's stolen keystrokes, drifting caret, and paste/attachment spacing under `jeo --tmux`.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.ko.md

@@ -288,10 +288,10 @@ CI는 `.github/workflows/npm-publish.yml`로 배포합니다 — GitHub 릴리
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
 - **[Unreleased]**
+- **[0.9.17]** (2026-08-18) — Live model discovery already existed (`listProviderModels`/`discoverModels`), but only `jeo auth login openai` ever called it — logging into Anthropic or Antigravity left the account pinned to the maintained static catalog snapshot until the next unrelated live-discovery call happened to run.
+- **[0.9.16]** (2026-08-18) — Routing looked broken while it was actually working: the idle status bar and `/compact`/`/handoff` never reflected which model `routePrompt` actually routed to.
 - **[0.9.15]** (2026-08-18) — The 0.9.14 fix wired `promptInput` into the `$a $b …` one-shot skill-chain's `io.input`, but that chain runs before the interactive REPL's own `readline.Interface` exists — so a bundled workflow skill (`$deep-interview`, with or without `--tmux`) invoked directly from the command line still froze forever on its first question.
 - **[0.9.14]** (2026-08-04) — A `$a $b … [intent]` skill-chain invocation of a bundled workflow skill (deep-interview/ralplan/team/ultragoal) hung forever with no error and no prompt.
-- **[0.9.13]** (2026-08-03) — The provider-login browser tab now actually goes away — by itself on macOS, by a working button everywhere else.
-- **[0.9.12]** (2026-08-02) — Fixes the boxed prompt's stolen keystrokes, drifting caret, and paste/attachment spacing under `jeo --tmux`.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.md

@@ -295,10 +295,10 @@ Huge thanks to [gajae-code](https://github.com/Yeachan-Heo/gajae-code) for the i
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
 - **[Unreleased]**
+- **[0.9.17]** (2026-08-18) — Live model discovery already existed (`listProviderModels`/`discoverModels`), but only `jeo auth login openai` ever called it — logging into Anthropic or Antigravity left the account pinned to the maintained static catalog snapshot until the next unrelated live-discovery call happened to run.
+- **[0.9.16]** (2026-08-18) — Routing looked broken while it was actually working: the idle status bar and `/compact`/`/handoff` never reflected which model `routePrompt` actually routed to.
 - **[0.9.15]** (2026-08-18) — The 0.9.14 fix wired `promptInput` into the `$a $b …` one-shot skill-chain's `io.input`, but that chain runs before the interactive REPL's own `readline.Interface` exists — so a bundled workflow skill (`$deep-interview`, with or without `--tmux`) invoked directly from the command line still froze forever on its first question.
 - **[0.9.14]** (2026-08-04) — A `$a $b … [intent]` skill-chain invocation of a bundled workflow skill (deep-interview/ralplan/team/ultragoal) hung forever with no error and no prompt.
-- **[0.9.13]** (2026-08-03) — The provider-login browser tab now actually goes away — by itself on macOS, by a working button everywhere else.
-- **[0.9.12]** (2026-08-02) — Fixes the boxed prompt's stolen keystrokes, drifting caret, and paste/attachment spacing under `jeo --tmux`.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/README.zh.md

@@ -287,10 +287,10 @@ CI 通过 `.github/workflows/npm-publish.yml` 发布 — GitHub 发布 release
 
 <!-- CHANGELOG:START (auto-generated from CHANGELOG.md — run `bun run changelog:sync`) -->
 - **[Unreleased]**
+- **[0.9.17]** (2026-08-18) — Live model discovery already existed (`listProviderModels`/`discoverModels`), but only `jeo auth login openai` ever called it — logging into Anthropic or Antigravity left the account pinned to the maintained static catalog snapshot until the next unrelated live-discovery call happened to run.
+- **[0.9.16]** (2026-08-18) — Routing looked broken while it was actually working: the idle status bar and `/compact`/`/handoff` never reflected which model `routePrompt` actually routed to.
 - **[0.9.15]** (2026-08-18) — The 0.9.14 fix wired `promptInput` into the `$a $b …` one-shot skill-chain's `io.input`, but that chain runs before the interactive REPL's own `readline.Interface` exists — so a bundled workflow skill (`$deep-interview`, with or without `--tmux`) invoked directly from the command line still froze forever on its first question.
 - **[0.9.14]** (2026-08-04) — A `$a $b … [intent]` skill-chain invocation of a bundled workflow skill (deep-interview/ralplan/team/ultragoal) hung forever with no error and no prompt.
-- **[0.9.13]** (2026-08-03) — The provider-login browser tab now actually goes away — by itself on macOS, by a working button everywhere else.
-- **[0.9.12]** (2026-08-02) — Fixes the boxed prompt's stolen keystrokes, drifting caret, and paste/attachment spacing under `jeo --tmux`.
 
 See [CHANGELOG.md](CHANGELOG.md) for the full history.
 <!-- CHANGELOG:END -->

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "jeo-code",
-  "version": "0.9.15",
+  "version": "0.9.17",
   "description": "Clean, highly optimized AI coding agent using spec-first loop",
   "type": "module",
   "main": "src/cli.ts",

package/src/commands/auth.ts

@@ -5,7 +5,8 @@ import { type StoredOAuth } from "../agent/state";
 import { createInterface } from "node:readline/promises";
 import { readGlobalConfig } from "../agent/state";
 import { jeoEnv } from "../util/env";
-import { listProviderModels, CODEX_MODELS_URL } from "../ai/model-discovery";
+import { listProviderModels } from "../ai/model-discovery";
+
 import { writeModelCache } from "../ai/model-cache";
 import {
   OAUTH_FLOWS,
@@ -123,7 +124,7 @@ async function runAuthLogin(rest: string[]): Promise<void> {
     rl.close();
     await loginOAuth(chosen, manualToken.trim());
     console.log(`[SUCCESS] Stored manual OAuth bearer for ${chosen} (no auto-refresh).`);
-    if (chosen === "openai") await reportOpenAiCodexModels();
+    if (LIVE_MODEL_REPORT_PROVIDERS.has(chosen)) await reportLiveModels(chosen);
     return;
   }
 
@@ -131,7 +132,7 @@ async function runAuthLogin(rest: string[]): Promise<void> {
     const { email } = await interactiveOAuthLogin(chosen, rl);
     console.log(`\n[SUCCESS] OAuth login complete for ${chosen}${email ? ` (${email})` : ""}.`);
     console.log("Stored access + refresh tokens in ~/.jeo/config.json; jeo will auto-refresh on expiry.");
-    if (chosen === "openai") await reportOpenAiCodexModels();
+    if (LIVE_MODEL_REPORT_PROVIDERS.has(chosen)) await reportLiveModels(chosen);
   } catch (err) {
     console.log(`\n[FAILED] ${(err as Error).message}`);
     console.log("Tip: paste the redirect URL when prompted, or use 'jeo auth login <provider> --token <bearer>'.");
@@ -142,30 +143,35 @@ async function runAuthLogin(rest: string[]): Promise<void> {
 }
 
 /**
- * Best-effort, timeout-bounded live Codex model check run immediately after a
- * successful `jeo auth login openai`. Uses `preferOAuth` so the report reflects
- * the account that was JUST logged in — never the catalog fallback, and never
- * silently swapped to an unrelated pre-existing `providers.openai` API key.
- * A failure here is NEVER fatal: login has already succeeded and is not rolled
- * back by this check. Prints only model ids/count — no tokens/secrets.
+ * Best-effort, timeout-bounded live model check run immediately after a
+ * successful `jeo auth login <provider>` for the providers whose model set
+ * changes fastest (openai/Codex, anthropic, antigravity/Cloud Code Assist).
+ * Uses `preferOAuth` so the report reflects the account that was JUST logged
+ * in — never the catalog fallback, and never silently swapped to an
+ * unrelated pre-existing API key. A failure here is NEVER fatal: login has
+ * already succeeded and is not rolled back by this check. Prints only model
+ * ids/count — no tokens/secrets.
  */
-async function reportOpenAiCodexModels(): Promise<void> {
+const LIVE_MODEL_REPORT_PROVIDERS = new Set<AuthProvider>(["openai", "anthropic", "antigravity"]);
+
+async function reportLiveModels(provider: AuthProvider): Promise<void> {
   try {
-    const result = await listProviderModels("openai", { preferOAuth: true });
+    const result = await listProviderModels(provider, { preferOAuth: true });
     if (result.ok) {
       // Persist immediately: the very next `jeo` launch then knows this account's
       // real model set (including ids newer than the maintained static snapshot)
       // before it makes any network call.
       await writeModelCache([result]);
-      console.log(`Live Codex models (${result.models.length}): ${result.models.join(", ")}`);
+      console.log(`Live ${provider} models (${result.models.length}): ${result.models.join(", ")}`);
     } else {
-      console.log(`[WARN] Could not verify live Codex models from ${CODEX_MODELS_URL}: ${result.error ?? "unknown error"} (non-fatal — login succeeded).`);
+      console.log(`[WARN] Could not verify live ${provider} models: ${result.error ?? "unknown error"} (non-fatal — login succeeded).`);
     }
   } catch (err) {
-    console.log(`[WARN] Could not verify live Codex models from ${CODEX_MODELS_URL}: ${(err as Error).message} (non-fatal — login succeeded).`);
+    console.log(`[WARN] Could not verify live ${provider} models: ${(err as Error).message} (non-fatal — login succeeded).`);
   }
 }
 
+
 async function runAuthImport(rest: string[]): Promise<void> {
   const provider = rest.find(a => a !== "--import") as AuthProvider | undefined;
   if (!provider) {

package/src/commands/launch.ts

@@ -687,6 +687,19 @@ export async function runLaunchCommand(args: string[]): Promise<void> {
   let sessionThinking: "low" | "medium" | "high" | "xhigh" | undefined = flags.thinking ?? cfg.thinkingLevel;
   // PromptRouter session override: undefined = follow config.routing.enabled, true/false = /route on|off wins this session.
   let sessionRouteOverride: boolean | undefined;
+  // Last model that actually served a completed turn (routed or pinned) — the idle
+  // status bar (`statusBarLine`) and `/compact`/`/handoff` read THIS instead of
+  // recomputing `sessionModel || defaultModel`, which ignored routing entirely and
+  // kept showing the pre-routing default even while routing was actively picking a
+  // different model turn-to-turn (e.g. "/route on" routing every prompt to
+  // claude-sonnet-5 while the footer kept showing claude-opus-5 — looked exactly
+  // like routing "wasn't working" even though it was). `undefined` until the first
+  // turn completes, so the initial idle render still falls back to `sessionModel ||
+  // defaultModel` unchanged.
+  let lastActiveModel: string | undefined;
+
+
+
   // Last routing decision (or the reason none applied) — /route why reads this.
   let lastRouteDecision: RouteDecision | { note: string } | null = null;
   // Bounded FIFO of this session's routing decisions — /route history reads this.
@@ -1929,7 +1942,9 @@ export async function runLaunchCommand(args: string[]): Promise<void> {
       }
     }
 
+    lastActiveModel = activeModel;
     return { done: result.done, steps: result.steps, reply, rendered: !!tui, usage };
+
   };
 
 
@@ -3017,7 +3032,8 @@ export async function runLaunchCommand(args: string[]): Promise<void> {
   // The gjc-layout status bar pinned directly ABOVE the input box: bg-gradient
   // identity block (model · thinking / branch / cwd) left, live ctx% right.
   const statusBarLine = (cols: number): string => {
-    const activeModel = sessionModel || defaultModel;
+    const activeModel = lastActiveModel ?? (sessionModel || defaultModel);
+
     const meta = catalogMetadata(activeModel);
     const used = historyTokens(history);
     const theme = uiTheme;
@@ -4321,10 +4337,16 @@ export async function runLaunchCommand(args: string[]): Promise<void> {
         continue;
       }
       if (input === "/compact") {
+
         // Session-start default snapshot (not the live global default) — keeps a
         // concurrent session's `/model` write from changing this session's context budget.
-        const activeModel = sessionModel || defaultModel;
+        // Compaction budget should reflect the model that actually served the last
+        // turn (routing may have picked something other than sessionModel/defaultModel),
+        // not the pre-routing session-start snapshot — otherwise a routed session could
+        // compact against the wrong context window.
+        const activeModel = lastActiveModel ?? (sessionModel || defaultModel);
         const contextTokens = catalogMetadata(activeModel)?.contextTokens;
+
         const res = await maybeCompact(history, { model: activeModel, force: true, contextTokens });
         if (res.error) {
           console.error(chalk.red(res.error));
@@ -4346,7 +4368,8 @@ export async function runLaunchCommand(args: string[]): Promise<void> {
         // has no ACP/SDK-broker managed-session boundary to hand off INTO, so this
         // NEVER mutates history (see buildHandoffDocument's doc comment).
         const parsed = parseHandoffCommand(input) ?? {};
-        const activeModel = sessionModel || defaultModel;
+        const activeModel = lastActiveModel ?? (sessionModel || defaultModel);
+
         const cfg = await readGlobalConfig();
         // `Config` (state.ts) predates config-schema.ts's `compaction` field and is
         // out of this change's scope — read the passthrough-preserved value via a