@textcortex/zenocode 0.1.14 → 0.1.16

package/README.md

@@ -16,6 +16,7 @@ If you already use pay-as-you-go and just ran out of wallet credits, top up unde
 - Connects the runtime to your TextCortex account
 - Discovers the models available to your TextCortex workspace
 - Configures the underlying coding agent automatically so you can start coding instead of hand-wiring providers and credentials
+- Runs a pinned, source-branded Zenocode runtime; it does not modify OpenCode binaries in npm or pnpm caches
 
 ## Install
 
@@ -37,9 +38,43 @@ The first `zenocode login` launches Zenocode automatically after browser authent
 
 If you already have an API key, you can also start Zenocode by setting `TEXTCORTEX_API_KEY` or `TEXTCORTEX_API_TOKEN`.
 
+## Update
+
+Zenocode releases its launcher and managed runtime together. Update both with:
+
+```bash
+npm install -g @textcortex/zenocode@latest
+```
+
+The terminal runtime never checks for, offers, or installs upstream OpenCode updates: the startup update check, the in-app "Update Available" prompt, and the runtime's `/global/upgrade` API are all removed at build time. Running `zenocode upgrade` prints the managed update command above.
+
+Runtime package versions include a Zenocode revision suffix (for example, `1.18.10-zenocode.2`). Increment that revision when the Zenocode source overlay changes without an upstream version bump, so an immutable npm runtime release can always be replaced deliberately.
+
 ## Models
 
-Zenocode uses **GPT-5.6 Luna** by default. You can switch to any other model available to your TextCortex workspace from the model picker inside the agent. Available models are discovered automatically from your account, so the default falls back to the next best coding model (for example Kimi K2.7 Code) when Luna is not available to you.
+Zenocode uses **GPT-6 Luna** by default. You can switch to any other model available to your TextCortex workspace from the model picker inside the agent. Available models are discovered from the canonical backend catalog on every launch. A newly added model therefore appears as soon as the backend grants the account access; no Zenocode package release or separate model-list update is required. The default falls back to GPT-5.6 Luna, then the next best coding model (for example Kimi K3), when GPT-6 Luna is not available to you.
+
+Alongside the Kimi, GLM, and MiniMax coding models, the catalog includes **Kimi K3**, **Claude Sonnet 5**, and **Claude Opus 5** when your workspace has access to them.
+
+### Thinking
+
+Most models express thinking as a pair of catalog entries — pick `…-thinking` or `…-non-thinking` in the model picker.
+
+Claude Sonnet 5 and Claude Opus 5 additionally support a **thinking level**. Inside the agent, press `ctrl+t` to cycle `low → medium → high → default`; the choice is remembered per model across sessions.
+
+For non-interactive runs, pass the level explicitly:
+
+```bash
+zenocode run --variant high "refactor the auth module"
+```
+
+Models without selectable levels ignore the flag and run at their deployment default.
+
+## Usage Attribution
+
+Zenocode generation is recorded in the existing `model_usage_events` metering pipeline with `source_type = 'direct_generation'` and `client_medium = 'zenocode'`. Super admins can filter or group the Model Cost Analytics view by **Client medium → zenocode**. The same dimension is available from the `/internal/v2/model-cost-analytics/query` backend endpoint.
+
+Attribution starts when the backend change is deployed. Older direct-generation events did not carry a reliable Zenocode marker and are not backfilled.
 
 ## Built For Security And Compliance
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@textcortex/zenocode",
-  "version": "0.1.14",
+  "version": "0.1.16",
   "description": "Secure, TextCortex-managed coding agent for customers that runs in your terminal, edits files, runs scripts, and more.",
   "keywords": [
     "ai",