min-agent 0.6.2 → 0.6.4

package/docs/API.md

@@ -1,6 +1,6 @@
 # min-agent HTTP API
 
-`min-agent serve` 启动本地 HTTP 服务，以编程方式调用代理能力。未配置服务商时使用内置 OpenRouter 模型，不必先运行 `min-agent setup`。
+`min-agent serve` 启动本地 HTTP 服务，以编程方式调用代理能力。未配置服务商时使用内置 OpenRouter 模型（`openrouter/free`，预设锁定、不可修改），不必先运行 `min-agent setup`。
 
 ## 启动
 
@@ -118,6 +118,8 @@ CLI 升级请使用 `min-agent update`（执行 `npm install -g min-agent`）。
 { "default_model": "gpt-4o", "models": ["gpt-4o", "gpt-4o-mini"] }
 ```
 
+当内置 OpenRouter 预设生效时，`models` 固定为 `["openrouter/free"]`（不发网络请求，也不允许其它模型）。
+
 ## `GET /v1/context`
 
 ```json
@@ -183,7 +185,7 @@ CLI 升级请使用 `min-agent update`（执行 `npm install -g min-agent`）。
 | `thinking` | string | 本轮思考强度：`off` / `low` / `medium` / `high` / `max`。不传则用配置，未配置时为 `medium` |
 | `memory` | string | 本轮是否使用记忆：`on` / `off`。不传则用配置，未配置时为 `off` |
 
-`auto_approve: false` 且未开流式时返回 `400 auto_approve_requires_stream`。确认/提问最多等待 5 分钟，超时视为拒绝或跳过。客户端断开连接时未完成的确认视为拒绝。
+`auto_approve: false` 且未开流式时返回 `400 auto_approve_requires_stream`。确认/提问最多等待 5 分钟，超时视为拒绝或跳过。客户端断开连接时未完成的确认视为拒绝。`auto_approve: false` 时所有需要确认的操作都会回调，包括文件写入/编辑（不再受服务端 `permission` 的 `accept-edits` 影响）；同一步的多个操作会各自发出一条 `confirm` 事件，分别用 `POST /v1/approvals` 回复。
 
 `max_steps_reached` 为 `true` 表示本轮已用尽自动续跑次数（上下文压缩/空回复等），或显式设置了 `agent.maxSteps` 且已到达该上限。默认不按固定步数切断一轮生成。`incomplete` 为 `true` 表示模型在调用工具后没有给出完整回复，或连续多次空回复。`stopped` 为 `true` 表示因同一操作反复执行而主动结束本轮。`empty_response` 为 `true` 表示 provider 连续返回空流（HTTP 200 但没有任何内容，多为上游超时/限流），`empty_attempts` 为已尝试次数；这类空回复会以原请求重发重试（首次立即，之后指数退避），可用 `agent.maxEmptyAttempts`（默认 4）与 `agent.emptyRetryDelayMs`（默认 1000）调整。发送下一条消息即可继续。`continues` 为本轮实际自动续跑次数。`task_state` 为当前任务目标与待办列表，会写入会话并在压缩后保留；纯问候不会记为任务目标。网页检索连续过久、或检索后尚未写出用户要求的结果时，代理会改为基于已有资料产出结果；本轮内再调用搜索/抓取会收到停止检索并直接产出的提示，而不是新的检索结果（即使 `autoContinue` 为 `false`）。
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "min-agent",
-  "version": "0.6.2",
+  "version": "0.6.4",
   "type": "module",
   "description": "Minimal AI coding agent with tool use, MCP, and skills support",
   "license": "MIT",

package/skills/self-config/SKILL.md

@@ -80,16 +80,16 @@ Instructions...
 
 - Rules: edit the files in the table above. `/reload` reloads rules only, not MCP.
 - Memory: **off** by default. Config `memory` is `on` | `off` (`true`/`enable` → on, `false`/`disable` → off). Project overlays global; writing `memory` globally also removes a project `memory` overlay. When off, memories are not injected and `memory_save` / `memory_search` / `memory_delete` are not registered (takes effect next turn after you change the file). CLI `min-agent memory on|off` / `--memory` follow the same write-vs-this-run rule as permission. `/memory on|off` writes **global** by default (and clears a project overlay); `/memory --project` writes only the project file. `/memory` lists the switch plus stored entries; `/memory <text>` still writes `memory.json` even when off. Default save scope is project when `.min-agent/` exists. Shape: `{ "content": string, "tags": string[], "created": ISO string }` array. When on, memories are injected each turn as background reference, not as work to resume. Do not store secrets or API keys. HTTP: `GET/POST /v1/memory/mode`, chat body `memory`, and `GET /v1/memory` include `memory` / `source` / `label`.
-- Permission: `accept-edits` (default) | `ask` | `allow-all` in config `permission`. CLI `min-agent permission …` / `--permission` without a message writes config and exits; with a message it is this session only. `--yes` is this session only and does not write config.
+- Permission: `accept-edits` (default) | `ask` | `allow-all` in config `permission`. CLI `min-agent permission …` / `--permission` without a message writes config and exits; with a message it is this session only. `--yes` is this session only and does not write config. `min-agent exec` (one-shot) uses a permissive default when neither `--permission` nor `--yes` is given: file edits and ordinary commands run without asking, and only genuinely destructive commands (deleting files/dirs, `drop`/`truncate`/`delete from`, `mkfs`/`dd`/raw-device writes, `shred`/`wipefs`, `git reset --hard`/`git clean -f`, `docker rm`/`prune`, `find … -delete`/`-exec rm`) still prompt. An explicit `--permission` (or `--yes`) overrides this and behaves exactly as in interactive sessions (`allow-all` runs even destructive commands silently). Confirmation prompts are coalesced: when several actions in one step need approval, a single overlay lists them all and one allow/deny decision applies to the whole batch instead of stacking separate prompts.
 - Thinking intensity: `off` | `low` | `medium` | `high` | `max` in config `thinking`. Omit the field to use **medium**, then clamp to levels the current model actually exposes (from the model catalog). `/think` opens an interactive picker of those English levels and writes **global** config; that also clears a project `thinking` overlay so a leftover project `off` cannot hide the new value after restart. `/think --project` writes only the project file. Typed `/think off|low|medium|high|max` still works when the model supports that value. `min-agent think …` / `--think` follow the same write-vs-this-run rule as permission. `xhigh` / `extra-high` / `none` are accepted as aliases (`xhigh` → `max`, `none` → `off`). Each request maps that intensity onto the current model and gateway (MiniMax uses `adaptive`/`disabled`; official OpenAI omits `thinking` and maps `max` to `xhigh`; an `ollama` provider uses native chat with `think` plus the configured window). Learned mappings are cached in `~/.min-agent/thinking-wire-cache.json` for 7 days. Per-model context window and supported effort values are loaded from the model catalog (`https://models.lonae.com`, override with `MIN_AGENT_MODELS_API_URL`) and cached in `~/.min-agent/model-catalog-cache.json` for 7 days after a successful lookup. When the provider host is unknown, effort levels are the majority across offerings that actually list efforts (toggle-only rows do not vote), so a single gateway cannot invent `medium`. An `ollama` provider skips that catalog and uses native `POST /api/show` instead (whether the model advertises thinking, plus the loaded window: Modelfile `num_ctx` if set, otherwise the architecture length capped at 32768). Chat completions are sent to native `POST /api/chat` so `options.num_ctx` and `think` actually apply — the OpenAI-compatible `/v1` endpoint ignores both. `/ctx` opens a picker of `2k` / `4k` / `8k` / `12k` / `16k` / `32k` / `64k` / `128k` / `256k` / `auto` and writes the active provider's `contextWindow` (global only; providers are not in project config). Typed `/ctx 8k` and CLI `min-agent ctx 8k` do the same; `auto` clears the override. The menu item is hidden unless the active provider is ollama. Results are cached in `~/.min-agent/ollama-model-cache.json` for 7 days.
 - Sandbox default is **off**. `mode`: `off` | `workspace` | `strict`. Optional `network`, `extraWriteRoots`, `extraReadRoots`. Isolation is independent of permission. `--sandbox` / `--network` without a message write config and exit.
 
 ## Providers
 
-Only in global `config.json`: `providers[]` with `name`, `type` (`openai-compatible` | `openai` | `ollama`), `baseURL`, `apiKey`, optional `defaultModel`, `contextWindow`. `activeProvider` selects one. Project config may set `activeProvider` and `defaultModel` (no keys). If `providers` is missing or empty, a built-in OpenRouter provider (`openrouter`, model `openrouter/free`) is used at runtime and is not written to disk. User-configured providers always win. `MIN_AGENT_NO_BUILTIN_PROVIDER=1` disables the fallback.
+Only in global `config.json`: `providers[]` with `name`, `type` (`openai-compatible` | `openai` | `ollama`), `baseURL`, `apiKey`, optional `defaultModel`, `contextWindow`. `activeProvider` selects one. Project config may set `activeProvider` and `defaultModel` (no keys). If `providers` is missing or empty, a built-in OpenRouter provider (`openrouter`, model `openrouter/free`) is used at runtime and is not written to disk. The preset is **locked**: while it is the active provider, `min-agent setup` rejects adding, editing, switching, or removing providers, the model list is fixed to `openrouter/free` (never fetched), and only that model can be selected (`/model`, `--model`, chat `model`). `MIN_AGENT_NO_BUILTIN_PROVIDER=1` disables the fallback — with it set (or once any user provider exists) the lock no longer applies and setup works normally.
 
 `/provider` opens an interactive picker and writes **global** `activeProvider`; that also clears a project `activeProvider` overlay so the choice is what later turns use. Typed `/provider <name>` still works. Adding, editing, or removing providers is `min-agent setup` (not the TUI picker).
 
-CLI: `min-agent setup` (interactive wizard or flags). Flags: `--type openai-compatible|openai|ollama`, `--url`, `--api-key`, `--name`, `--default-model`, `--switch <name>`, `--remove <name>`. `--yes` skips overwrite/delete confirmation during setup; it does not change permission. Keys only in the global file.
+CLI: `min-agent setup` (interactive wizard or flags). Flags: `--type openai-compatible|openai|ollama`, `--url`, `--api-key`, `--name`, `--default-model`, `--switch <name>`, `--remove <name>`. `--yes` skips overwrite/delete confirmation during setup; it does not change permission. Keys only in the global file. While the preset is active the wizard only shows the locked built-in provider (test connection / done).
 
 After changing keys or MCP connections, tell the user if a restart is required. Do not print full secrets back in chat.

package/skills/self-config/reference.md

@@ -72,7 +72,7 @@ Read this when you need exact keys. Files are strict JSON.
 - Old single-provider configs migrate on first run; write the `providers` array for new files.
 - Interactive and non-interactive provider setup both go through `min-agent setup`. API keys are written only to the global config.
 - Interactive `/provider` (and typed `/provider <name>`) writes global `activeProvider` and clears a project `activeProvider` overlay. It does not add, edit, or remove providers.
-- If `providers` is missing or empty, runtime uses a built-in OpenRouter provider (`name`: `openrouter`, `defaultModel`: `openrouter/free`) without writing it to disk. User entries always win. `MIN_AGENT_NO_BUILTIN_PROVIDER=1` disables that fallback.
+- If `providers` is missing or empty, runtime uses a built-in OpenRouter provider (`name`: `openrouter`, `defaultModel`: `openrouter/free`) without writing it to disk. The preset is locked: while it is the active provider, `min-agent setup` refuses to add/edit/switch/remove providers, the model list is fixed to `openrouter/free` (no network fetch), and selecting any other model (interactive picker, `/model`, `--model`, HTTP chat `model`) is rejected. `MIN_AGENT_NO_BUILTIN_PROVIDER=1` disables the fallback; with the preset disabled or any user provider stored, the lock does not apply and setup works normally.
 
 ## Project `.min-agent/config.json`