claude-multi 0.11.0 → 0.12.0

package/CHANGELOG.md

@@ -5,6 +5,22 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.12.0] - 2026-08-26
+
+### Changed
+- **GLM template updated**: sonnet-tier now maps to `glm-5.3-flash[1m]` instead of mirroring `glm-5.3[1m]`. Three-tier split restored: opus → `glm-5.3[1m]`, sonnet → `glm-5.3-flash[1m]`, haiku → `glm-5-turbo`. Existing instances are migrated automatically on next launch.
+
+## [0.11.1] - 2026-08-22
+
+### Fixed
+- Existing instances can now pick up provider template changes. `claude-multi doctor check` spots model and structural environment values that differ from the current template, even when the stored migration version is current. Run `claude-multi doctor fix` to back up the instance settings and apply the update.
+- The GLM 5.3 mapping now reaches existing GLM instances. Model slots update to `glm-5.3[1m]`; a legacy `MAX_OUTPUT_TOKENS=64000` becomes `128000`; obsolete GLM auto-compaction overrides are removed. MiniMax instances still using the old 64K output limit move to 512K.
+- Provider detection now accepts endpoints with a trailing slash, so a valid edited GLM URL does not cause template sync to be skipped.
+- The TUI migration option and health warning no longer appear for a version mismatch with no work to do.
+
+### Changed
+- Migrations and the instance-details "Sync template" action now use the same provider environment sync code. Model and structural settings follow the current template. API keys and user-tuned values stay in place, except for documented legacy defaults that need an upgrade.
+
 ## [0.11.0] - 2026-08-14
 
 ### Changed
@@ -491,6 +507,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Support for custom config and binary paths
 - Command-line interface built with Commander.js
 
+[0.12.0]: https://github.com/hmziqrs/claude-multi/compare/v0.11.1...v0.12.0
 [0.11.0]: https://github.com/hmziqrs/claude-multi/compare/v0.10.0...v0.11.0
 [0.8.2]: https://github.com/hmziqrs/claude-multi/compare/v0.8.1...v0.8.2
 [0.8.1]: https://github.com/hmziqrs/claude-multi/compare/v0.8.0...v0.8.1

package/README.md

@@ -55,6 +55,7 @@ Each instance is isolated. All of them run the normal, unmodified Claude Code bi
 * Plugin management per instance: enable, disable, install, copy, remove.
 * MCP management: list and copy server configs between instances.
 * Optional auto-sync that symlinks plugins and skills from your default `~/.claude`.
+* `doctor` reports provider template updates for existing instances and can apply them after making a backup.
 * No fork. It wraps your existing `claude` binary instead of patching it.
 
 ---
@@ -207,6 +208,17 @@ That's a full Claude Code session on the provider you picked, with its own isola
 
 If the menu shows a `!` health hint, press `!` to see what's wrong and fix it.
 
+### Update an existing provider instance
+
+Provider templates change as providers ship new models or revise their settings. After updating claude-multi, check existing instances:
+
+```bash
+claude-multi doctor check
+claude-multi doctor fix
+```
+
+The TUI also shows `Run instance migrations` when an update is available. The sync updates model names and other template settings, while keeping API keys and values you changed yourself. It creates a backup before writing `settings.json`.
+
 ---
 
 ## Providers
@@ -222,7 +234,7 @@ If the menu shows a `!` health hint, press `!` to see what's wrong and fix it.
 | Alibaba Qwen | `dashscope-intl.aliyuncs.com` | `qwen` |
 | Alibaba Qwen Coding Plan | `coding-intl.dashscope.aliyuncs.com` | `qwen-coding` |
 
-The templates write `ANTHROPIC_BASE_URL`, `ANTHROPIC_MODEL`, and a few related env vars into `~/.claude-<name>/settings.json`. Edit that file if you want to tweak things further.
+The templates write `ANTHROPIC_BASE_URL`, `ANTHROPIC_MODEL`, and related env vars into `~/.claude-<name>/settings.json`. If you edit a setting, provider sync keeps your custom tunables but restores model slots to the current template.
 
 ---