@hifullmoon/aicommit 2.5.0 → 2.6.0

package/CHANGELOG.md

@@ -4,6 +4,22 @@ This file lists notable user-facing changes. Internal refactors, test-only chang
 
 ## [Unreleased]
 
+## [2.6.0] - 2026-09-14
+
+### Added
+
+- Added `generate --scope=staged|all` and `apply` for a reviewable single-commit plan that rechecks the repository snapshot before committing.
+- Added read-only `split status --output=json` so interrupted split transactions expose completed commits and the next recovery action.
+
+### Changed
+
+- Machine output schema 1.1 now reports commit IDs, plan paths, explicit change scope, and structured recovery errors. Consumers pinned to schema 1.0 should update their parsers.
+- Staged snapshot checks use Git object IDs instead of regenerating full patches, and normal generation skips `git diff --stat` unless its diff is truncated.
+
+### Fixed
+
+- Split planning now displays the final model request's thinking for large changes and preserves it in interactive review; `auto` mode displays thinking when a provider emits it. JSON output remains free of model reasoning.
+
 ## [2.5.0] - 2026-09-10
 
 ### Added
@@ -217,7 +233,8 @@ This file lists notable user-facing changes. Internal refactors, test-only chang
 - Added file-level split planning and execution with Git-state concurrency checks.
 - Added provider presets and user/project configuration boundaries.
 
-[Unreleased]: https://github.com/hi-fullmoon/AICommit/compare/v2.5.0...HEAD
+[Unreleased]: https://github.com/hi-fullmoon/AICommit/compare/v2.6.0...HEAD
+[2.6.0]: https://github.com/hi-fullmoon/AICommit/releases/tag/v2.6.0
 [2.5.0]: https://github.com/hi-fullmoon/AICommit/releases/tag/v2.5.0
 [2.4.1]: https://github.com/hi-fullmoon/AICommit/releases/tag/v2.4.1
 [2.4.0]: https://github.com/hi-fullmoon/AICommit/releases/tag/v2.4.0

package/README.md

@@ -311,11 +311,15 @@ aicommit --dry-run       # generate and review without creating a commit
 aicommit split --dry-run # review a split plan without creating commits
 aicommit --yes           # non-interactively commit already staged changes
 aicommit --yes --dry-run # non-interactively preview all changes; restores staging
+aicommit --yes --dry-run --scope=staged --output=json # preview only the index
+aicommit generate --scope=staged --file=/tmp/commit-plan.json --yes --output=json
+aicommit apply --file=/tmp/commit-plan.json --yes --output=json
 aicommit split --scope=all --yes # non-interactively plan and commit all working-tree changes
 aicommit split --scope=all --yes --allow-single-fallback # explicitly permit a conservative fallback commit
 aicommit split plan --scope=staged --file=/tmp/split-plan.json --yes
 aicommit split apply --file=/tmp/split-plan.json --yes
 aicommit split resume --yes # resume an interrupted split transaction
+aicommit split status --output=json # inspect recovery state without changing Git
 aicommit split abort --yes # discard a stale checkpoint; keep commits and changes
 aicommit --reasoning=low # stream low-effort reasoning; Ctrl+O expands/collapses it
 aicommit --no-reasoning # explicitly disable reasoning when supported
@@ -331,8 +335,8 @@ aicommit -h              # help
 | `-l`, `--lang`            | Commit message language (`zh` or `en`)                                         |
 | `-p`, `--provider`        | Use the named provider from `providers`                                        |
 | `-m`, `--model`           | Use a named model profile from the selected provider                           |
-| `--scope`                 | `staged` or `all` scope for `aicommit split` and `aicommit split plan`         |
-| `--file`                  | JSON plan path for `aicommit split plan` and `aicommit split apply`            |
+| `--scope`                 | `staged` or `all` scope for dry-run, generate, and split planning              |
+| `--file`                  | JSON plan path for generate/apply and split plan/apply                         |
 | `--dry-run`               | Generate and review a message or split plan without creating commits           |
 | `-y`, `--yes`             | Accept without prompts; normal mode requires explicitly staged changes         |
 | `--allow-single-fallback` | Explicitly permit non-interactive split to create one complete fallback commit |
@@ -388,11 +392,13 @@ Verify the registration with `whence -w _aicommit`; it should print `_aicommit:
 
 ### Machine-readable output
 
-Use `--output=json` for scripts and CI. Commit and split flows also require `--yes`, preventing a machine consumer from hanging on an interactive prompt. stdout contains exactly one JSON object; progress, debug details, and diagnostics go to stderr. `doctor --output=json` and `update --output=json` do not require `--yes`.
+Use `--output=json` for scripts and CI. Commit, generate, apply, and split execution flows also require `--yes`, preventing a machine consumer from hanging on an interactive prompt. stdout contains exactly one JSON object; progress, debug details, and diagnostics go to stderr. `split status`, `doctor`, and `update` in JSON mode do not require `--yes`.
+
+For an agent-reviewed single commit, run `generate --scope=staged|all --file=<path> --yes --output=json`, inspect the returned message and plan, then run `apply --file=<path> --yes --output=json`. The plan is a validated one-group artifact; apply checks its base HEAD, change list, and content fingerprint before committing. Keep the file outside the worktree or in `.git/aicommit/`. `generate --scope=all` includes staged, unstaged, and untracked changes, restores its temporary staging, and refuses detected sensitive content in non-interactive mode. For a quick preview without an artifact, use `--dry-run --scope=staged|all --yes --output=json`; an omitted scope retains the older automatic staging behavior when the index is empty.
 
 ```json
 {
-  "schemaVersion": "1.0",
+  "schemaVersion": "1.1",
   "ok": true,
   "message": "fix: handle provider retry limits",
   "plan": null,
@@ -407,11 +413,16 @@ Use `--output=json` for scripts and CI. Commit and split flows also require `--y
   "warnings": [],
   "exitReason": "dry_run",
   "committed": false,
+  "commitState": "none",
+  "commitSha": null,
+  "planFile": null,
+  "scope": "staged",
+  "changeCount": 1,
   "error": null
 }
 ```
 
-The published [JSON schema](schemas/aicommit-output.schema.json) covers success, split-plan, doctor/check, and error results. Machine output never includes the diff or model reasoning. Split output exposes only each group message and its assigned paths.
+The published [JSON schema](schemas/aicommit-output.schema.json) covers success, plans, status, doctor/check, and error results. Machine output never includes the diff or model reasoning. Plans expose only each group message and its assigned paths. `committed` reports whether this invocation created a commit; `commitState` describes `none`, `partial`, `complete`, or `unknown` transaction state. A split failure can return `committed: true` with `error.code: "split_partial_failure"`, completed commit IDs in `data.split`, and `error.nextAction`. After a crash that emits no JSON, use `split status --output=json` before retrying.
 
 Stable process exits are shared by text and JSON modes:
 
@@ -444,7 +455,7 @@ The staged index (or complete split-mode working tree, including untracked file
 
 Reasoning defaults to `on` with `medium` effort. It is mapped natively for OpenAI reasoning models, DeepSeek, OpenRouter, and MiniMax; models that do not expose reasoning continue normally and show an unavailable notice instead of failing. Official OpenAI endpoints validate the selected effort against the model generation before sending the request, so unsupported combinations such as `o3 --no-reasoning` or `gpt-5.1 --reasoning=max` fail locally with a clear list of supported levels. DeepSeek's current `deepseek-v4-flash` and `deepseek-v4-pro` models receive `thinking: { "type": "enabled" }` plus `reasoning_effort`; `medium`/`xhigh` are normalized to DeepSeek's `high` level.
 
-When reasoning mode is `on` (including via `--reasoning=<level>`), aicommit requests a streaming response and displays reasoning as it arrives. The live view follows the newest two terminal lines by default; press `Ctrl+O` to expand or collapse the accumulated text during generation or review. Long expanded output is kept inside the terminal viewport—use `PageUp`/`PageDown` to read every page. Holding `Ctrl+O` counts as one toggle, so key repeat cannot leave duplicate panels behind. Output is sanitized and capped by `reasoning.maxDisplayChars`; providers that do not expose reasoning show a short unavailable notice.
+When reasoning mode is `on` (including via `--reasoning=<level>`), aicommit requests a streaming response and displays reasoning as it arrives. In `auto` mode it also displays thinking if the provider emits it, without forcing reasoning on. Split mode streams the final planning request's thinking, including when large changes use batched analysis; the review prompt keeps that reasoning available. Intermediate analysis batches show progress rather than mixing several reasoning streams. The live panel requires an interactive terminal, and JSON output never includes reasoning. The live view follows the newest two terminal lines by default; press `Ctrl+O` to expand or collapse the accumulated text during generation or review. Long expanded output is kept inside the terminal viewport—use `PageUp`/`PageDown` to read every page. Holding `Ctrl+O` counts as one toggle, so key repeat cannot leave duplicate panels behind. Output is sanitized and capped by `reasoning.maxDisplayChars`; providers that do not expose reasoning show a short unavailable notice.
 
 ```json
 {

package/README.zh-CN.md

@@ -313,11 +313,15 @@ aicommit --dry-run       # 生成并审阅，但不创建提交
 aicommit split --dry-run # 审阅拆分计划，但不创建提交
 aicommit --yes           # 非交互提交已明确暂存的变更
 aicommit --yes --dry-run # 非交互预览所有变更；退出时恢复暂存状态
+aicommit --yes --dry-run --scope=staged --output=json # 只预览暂存区
+aicommit generate --scope=staged --file=/tmp/commit-plan.json --yes --output=json
+aicommit apply --file=/tmp/commit-plan.json --yes --output=json
 aicommit split --scope=all --yes # 非交互规划并提交所有工作区变更
 aicommit split --scope=all --yes --allow-single-fallback # 明确允许规划预算耗尽后的保守提交
 aicommit split plan --scope=staged --file=/tmp/split-plan.json --yes
 aicommit split apply --file=/tmp/split-plan.json --yes
 aicommit split resume --yes # 恢复中断的拆分事务
+aicommit split status --output=json # 只读查询恢复状态
 aicommit split abort --yes # 丢弃过期 checkpoint；保留提交和变更
 aicommit --reasoning=low # 流式显示低强度推理；Ctrl+O 展开或收起
 aicommit --no-reasoning # Provider / 模型支持时显式关闭推理
@@ -333,8 +337,8 @@ aicommit -h              # 帮助
 | `-l`, `--lang`            | 提交信息语言：`zh` 或 `en`                                          |
 | `-p`, `--provider`        | 使用 `providers` 中的命名 Provider                                  |
 | `-m`, `--model`           | 使用所选 Provider 下的命名模型配置                                  |
-| `--scope`                 | `aicommit split` 和 `aicommit split plan` 的范围：`staged` 或 `all` |
-| `--file`                  | `aicommit split plan` 和 `aicommit split apply` 的 JSON 计划路径    |
+| `--scope`                 | dry-run、generate 和 split 规划的范围：`staged` 或 `all`            |
+| `--file`                  | generate/apply 和 split plan/apply 的 JSON 计划路径                 |
 | `--dry-run`               | 生成并审阅消息或拆分计划，但不创建提交                              |
 | `-y`, `--yes`             | 不提示直接接受；普通模式要求变更已明确暂存                          |
 | `--allow-single-fallback` | 明确允许非交互拆分创建一个覆盖完整变更的保守提交                    |
@@ -390,11 +394,13 @@ exec zsh
 
 ### 机器可读输出
 
-脚本和 CI 请使用 `--output=json`。提交和 split 流程还必须使用 `--yes`，避免机器消费者卡在交互提示上。stdout 只包含一个 JSON 对象；进度、调试信息和诊断输出会写入 stderr。`doctor --output=json` 和 `update --output=json` 不要求 `--yes`。
+脚本和 CI 请使用 `--output=json`。提交、generate、apply 和 split 执行流程还必须使用 `--yes`，避免机器消费者卡在交互提示上。stdout 只包含一个 JSON 对象；进度、调试信息和诊断输出会写入 stderr。`split status`、`doctor` 和 `update` 的 JSON 模式不要求 `--yes`。
+
+让 AI 审阅单次提交时，先执行 `generate --scope=staged|all --file=<路径> --yes --output=json`，检查返回的消息和计划，再执行 `apply --file=<路径> --yes --output=json`。计划复用经过校验的单组工件；apply 会在提交前核对基础 HEAD、变更清单和内容指纹。计划文件应放在工作区外或 `.git/aicommit/`。`generate --scope=all` 包含已暂存、未暂存和未跟踪变更，结束时恢复临时暂存；非交互模式遇到检测出的敏感内容会停止。只想快速预览时，可用 `--dry-run --scope=staged|all --yes --output=json`；不指定范围时，保留旧版在暂存区为空时自动暂存的行为。
 
 ```json
 {
-  "schemaVersion": "1.0",
+  "schemaVersion": "1.1",
   "ok": true,
   "message": "fix: handle provider retry limits",
   "plan": null,
@@ -409,11 +415,16 @@ exec zsh
   "warnings": [],
   "exitReason": "dry_run",
   "committed": false,
+  "commitState": "none",
+  "commitSha": null,
+  "planFile": null,
+  "scope": "staged",
+  "changeCount": 1,
   "error": null
 }
 ```
 
-已发布的 [JSON schema](schemas/aicommit-output.schema.json) 覆盖成功、拆分计划、doctor / check 和错误结果。机器输出绝不包含 diff 或模型推理。split 输出只暴露每组消息及其分配路径。
+已发布的 [JSON schema](schemas/aicommit-output.schema.json) 覆盖成功、计划、状态、doctor / check 和错误结果。机器输出绝不包含 diff 或模型推理；计划只暴露每组消息及其分配路径。`committed` 表示本次调用是否创建了提交；`commitState` 表示事务处于 `none`、`partial`、`complete` 或 `unknown`。拆分失败时可能返回 `committed: true`、`error.code: "split_partial_failure"`、`data.split` 中已完成的提交 ID，以及 `error.nextAction`。若进程崩溃而没有输出 JSON，应先执行 `split status --output=json` 再决定下一步。
 
 文本与 JSON 模式共享稳定的进程退出码：
 
@@ -446,7 +457,7 @@ exec zsh
 
 推理默认为 `on`，强度为 `medium`。OpenAI 推理模型、DeepSeek、OpenRouter 和 MiniMax 使用各自原生映射；不暴露推理能力的模型会正常继续，并显示“不可用”提示，而不是失败。官方 OpenAI 端点会在发送请求前根据模型代际校验所选强度，因此 `o3 --no-reasoning` 或 `gpt-5.1 --reasoning=max` 等不支持的组合会在本地失败，并清楚列出支持的级别。DeepSeek 当前的 `deepseek-v4-flash` 和 `deepseek-v4-pro` 会收到 `thinking: { "type": "enabled" }` 与 `reasoning_effort`；`medium` / `xhigh` 会归一化为 DeepSeek 的 `high` 级别。
 
-推理模式为 `on` 时（包括通过 `--reasoning=<level>` 启用），AICommit 会请求流式响应并实时显示推理。默认实时视图跟随终端最新两行；生成或审阅期间按 `Ctrl+O` 可展开或收起累计文本。较长的展开输出会限制在终端视口内，可用 `PageUp` / `PageDown` 阅读所有页面。长按 `Ctrl+O` 只计为一次切换，避免按键重复留下多个面板。输出会经过清理，并受 `reasoning.maxDisplayChars` 限制；不暴露推理的 Provider 会显示简短的不可用提示。
+推理模式为 `on` 时（包括通过 `--reasoning=<level>` 启用），AICommit 会请求流式响应并实时显示推理；`auto` 模式下若 Provider 主动返回 think，也会显示，但不会强制开启推理。split 模式会显示最终规划请求的 think，大变更走分批分析时也一样，并在审阅界面保留这段内容；前面的分析批次只显示进度，避免多个推理流混在一起。实时面板需要交互式终端，JSON 输出始终不包含推理。默认实时视图跟随终端最新两行；生成或审阅期间按 `Ctrl+O` 可展开或收起累计文本。较长的展开输出会限制在终端视口内，可用 `PageUp` / `PageDown` 阅读所有页面。长按 `Ctrl+O` 只计为一次切换，避免按键重复留下多个面板。输出会经过清理，并受 `reasoning.maxDisplayChars` 限制；不暴露推理的 Provider 会显示简短的不可用提示。
 
 ```json
 {

package/docs/distribution.md

@@ -33,7 +33,7 @@ The release workflow uses npm Trusted Publishing without a long-lived `NPM_TOKEN
 ```bash
 workdir=$(mktemp -d)
 cd "$workdir"
-npm install --package-lock-only @hifullmoon/aicommit@2.5.0
+npm install --package-lock-only @hifullmoon/aicommit@2.6.0
 npm audit signatures
 ```
 

package/docs/troubleshooting.md

@@ -15,20 +15,20 @@ JSON mode keeps one machine object on stdout and diagnostics on stderr. The `err
 
 JSON 模式保证 stdout 只有一个机器对象，诊断进入 stderr。`error.category` 与进程退出码可稳定用于自动化。
 
-| Symptom / 症状                                                                  | Category / exit                 | Likely cause / 常见原因                                                                                                         | Check and recovery / 检查与恢复                                                                                                                         |
-| ------------------------------------------------------------------------------- | ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| Config fails before Git/provider access / 在 Git/provider 前配置失败            | `config` / `2`                  | Malformed JSON, invalid URL, unsupported option, missing credential / JSON 损坏、URL 或选项无效、凭据缺失                       | `aicommit config path`; `aicommit config validate`; fix only the user-owned file shown                                                                  |
-| Project settings are ignored / 项目设置被忽略                                   | warning                         | Repository config tried to set endpoint, credential, reasoning, or raise a ceiling / 仓库配置尝试设置连接、凭据、推理或提高预算 | Move trusted connection settings to `~/.aicommit/config.json`; keep team rules in `.aicommit.policy.json`                                               |
-| Not a Git repository, conflict, empty index / 非仓库、冲突或 index 为空         | `git_state` / `3`               | Wrong directory or unsafe Git state / 目录错误或 Git 状态不安全                                                                 | `git status`; `aicommit /absolute/repo/path`; resolve conflicts before retrying                                                                         |
-| DNS, connection, or timeout / DNS、连接或超时                                   | `network` / `4`                 | Endpoint unavailable, proxy/TLS issue, budget too short / endpoint 不可达、代理/TLS、超时过短                                   | `aicommit doctor -p NAME`; verify HTTPS URL; raise user-owned `timeoutMs` only if expected                                                              |
-| HTTP authentication/rate/parameter failure / 鉴权、限流或参数失败               | `provider` / `5`                | Wrong key/model/body; non-retryable 4xx / key、model、body 错误                                                                 | Check `apiKeyEnv`, model and compatibility table; authentication is never retried automatically                                                         |
-| Empty or malformed model reply / 空或畸形回复                                   | `response_format` / `6`         | Unsupported response dialect, token limit, or policy validation failure / 响应方言、token 限制或 policy 校验失败                | Raise `maxTokens`, choose the matching adapter, inspect validator issue code; one correction is already attempted                                       |
-| `split run --scope=all --yes` stops before API call / split 非交互在 API 前停止 | `sensitive_data` / `7`          | Complete untracked scan found sensitive-looking data / 完整未跟踪扫描发现疑似敏感数据                                           | Review/stage intended files explicitly; do not bypass without checking the actual content                                                               |
-| Commit aborts after generation / 生成后提交中止                                 | `concurrent_modification` / `8` | Index/worktree changed during the protected window / 受保护窗口中 index/worktree 被修改                                         | Review `git status`, restore the intended snapshot, and generate again                                                                                  |
-| Split stopped after one or more commits / split 部分提交后停止                  | reported Git failure            | Hook, crash, SIGINT, or concurrent pending edit / hook、崩溃、中断或待处理文件变化                                              | Run `aicommit split resume`; if another Git workflow already replaced the transaction, use `aicommit split abort`（只删除恢复元数据，不改提交或工作区） |
-| `aicommit update` refuses the installation / update 拒绝当前安装                | `config` / `2`                  | Source checkout, npm link, or another active Node/npm environment / 源码、npm link 或当前是另一套 Node/npm 环境                 | Activate the Node environment that installed AICommit, or run `npm install --global @hifullmoon/aicommit@latest` manually                               |
-| `aicommit update` cannot reach npm / update 无法访问 npm                        | `network` / `4`                 | Registry authentication, proxy, DNS, or network failure / registry 鉴权、代理、DNS 或网络失败                                   | Check `npm config get registry` and npm authentication/proxy settings, then retry                                                                       |
-| npm provenance is absent or invalid / npm provenance 缺失或失败                 | npm audit failure               | Old npm CLI, non-trusted release, or wrong version / npm 过旧、非可信发布或版本错误                                             | Upgrade npm; run `npm audit signatures`; install only a version linked to the official workflow                                                         |
+| Symptom / 症状                                                                  | Category / exit                 | Likely cause / 常见原因                                                                                                         | Check and recovery / 检查与恢复                                                                                                            |
+| ------------------------------------------------------------------------------- | ------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------ |
+| Config fails before Git/provider access / 在 Git/provider 前配置失败            | `config` / `2`                  | Malformed JSON, invalid URL, unsupported option, missing credential / JSON 损坏、URL 或选项无效、凭据缺失                       | `aicommit config path`; `aicommit config validate`; fix only the user-owned file shown                                                     |
+| Project settings are ignored / 项目设置被忽略                                   | warning                         | Repository config tried to set endpoint, credential, reasoning, or raise a ceiling / 仓库配置尝试设置连接、凭据、推理或提高预算 | Move trusted connection settings to `~/.aicommit/config.json`; keep team rules in `.aicommit.policy.json`                                  |
+| Not a Git repository, conflict, empty index / 非仓库、冲突或 index 为空         | `git_state` / `3`               | Wrong directory or unsafe Git state / 目录错误或 Git 状态不安全                                                                 | `git status`; `aicommit /absolute/repo/path`; resolve conflicts before retrying                                                            |
+| DNS, connection, or timeout / DNS、连接或超时                                   | `network` / `4`                 | Endpoint unavailable, proxy/TLS issue, budget too short / endpoint 不可达、代理/TLS、超时过短                                   | `aicommit doctor -p NAME`; verify HTTPS URL; raise user-owned `timeoutMs` only if expected                                                 |
+| HTTP authentication/rate/parameter failure / 鉴权、限流或参数失败               | `provider` / `5`                | Wrong key/model/body; non-retryable 4xx / key、model、body 错误                                                                 | Check `apiKeyEnv`, model and compatibility table; authentication is never retried automatically                                            |
+| Empty or malformed model reply / 空或畸形回复                                   | `response_format` / `6`         | Unsupported response dialect, token limit, or policy validation failure / 响应方言、token 限制或 policy 校验失败                | Raise `maxTokens`, choose the matching adapter, inspect validator issue code; one correction is already attempted                          |
+| `split run --scope=all --yes` stops before API call / split 非交互在 API 前停止 | `sensitive_data` / `7`          | Complete untracked scan found sensitive-looking data / 完整未跟踪扫描发现疑似敏感数据                                           | Review/stage intended files explicitly; do not bypass without checking the actual content                                                  |
+| Commit aborts after generation / 生成后提交中止                                 | `concurrent_modification` / `8` | Index/worktree changed during the protected window / 受保护窗口中 index/worktree 被修改                                         | Review `git status`, restore the intended snapshot, and generate again                                                                     |
+| Split stopped after one or more commits / split 部分提交后停止                  | `git_state` / `3`               | Hook, crash, SIGINT, or concurrent pending edit / hook、崩溃、中断或待处理文件变化                                              | Inspect `aicommit split status --output=json`, then run `aicommit split resume --yes`; only use `split abort` to discard recovery metadata |
+| `aicommit update` refuses the installation / update 拒绝当前安装                | `config` / `2`                  | Source checkout, npm link, or another active Node/npm environment / 源码、npm link 或当前是另一套 Node/npm 环境                 | Activate the Node environment that installed AICommit, or run `npm install --global @hifullmoon/aicommit@latest` manually                  |
+| `aicommit update` cannot reach npm / update 无法访问 npm                        | `network` / `4`                 | Registry authentication, proxy, DNS, or network failure / registry 鉴权、代理、DNS 或网络失败                                   | Check `npm config get registry` and npm authentication/proxy settings, then retry                                                          |
+| npm provenance is absent or invalid / npm provenance 缺失或失败                 | npm audit failure               | Old npm CLI, non-trusted release, or wrong version / npm 过旧、非可信发布或版本错误                                             | Upgrade npm; run `npm audit signatures`; install only a version linked to the official workflow                                            |
 
 If a failure remains, capture `aicommit doctor --output=json`, Node/Git versions, the error category, and redacted config sources. Never attach a diff, commit message, config file, API key, reasoning trace, or credential-helper output to a public issue.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hifullmoon/aicommit",
-  "version": "2.5.0",
+  "version": "2.6.0",
   "description": "Safe, local-first AI commit message generator for Git workflows",
   "type": "module",
   "bin": {

package/schemas/aicommit-output.schema.json

@@ -16,11 +16,16 @@
     "warnings",
     "exitReason",
     "committed",
+    "commitState",
+    "commitSha",
+    "planFile",
+    "scope",
+    "changeCount",
     "error"
   ],
   "properties": {
     "schemaVersion": {
-      "const": "1.0"
+      "const": "1.1"
     },
     "ok": {
       "type": "boolean"
@@ -107,10 +112,26 @@
     "committed": {
       "type": "boolean"
     },
+    "commitState": {
+      "enum": ["none", "partial", "complete", "unknown"]
+    },
+    "commitSha": {
+      "type": ["string", "null"]
+    },
+    "planFile": {
+      "type": ["string", "null"]
+    },
+    "scope": {
+      "type": ["string", "null"]
+    },
+    "changeCount": {
+      "type": ["integer", "null"],
+      "minimum": 0
+    },
     "error": {
       "type": ["object", "null"],
       "additionalProperties": false,
-      "required": ["category", "message"],
+      "required": ["category", "code", "message", "retryable", "nextAction"],
       "properties": {
         "category": {
           "enum": [
@@ -126,6 +147,16 @@
         },
         "message": {
           "type": "string"
+        },
+        "code": {
+          "type": "string",
+          "minLength": 1
+        },
+        "retryable": {
+          "type": "boolean"
+        },
+        "nextAction": {
+          "type": ["string", "null"]
         }
       }
     },
@@ -150,7 +181,7 @@
               "description": "Files represented by partial excerpts; not fully analyzed files."
             },
             "strategy": {
-              "enum": ["auto", "deep"]
+              "enum": ["auto", "deep", "fallback"]
             },
             "metadataOnlyFiles": {
               "type": "number",
@@ -181,6 +212,33 @@
               "minimum": 0
             }
           }
+        },
+        "split": {
+          "type": "object",
+          "additionalProperties": false,
+          "required": ["state"],
+          "properties": {
+            "state": { "enum": ["idle", "pending"] },
+            "transactionId": { "type": "string" },
+            "checkpointPath": { "type": "string" },
+            "scope": { "enum": ["staged", "all"] },
+            "totalGroups": { "type": "integer", "minimum": 0 },
+            "completedGroups": { "type": "integer", "minimum": 0 },
+            "pendingGroups": { "type": "integer", "minimum": 0 },
+            "completedCommits": {
+              "type": "array",
+              "items": { "type": "string", "pattern": "^[0-9a-f]{40}([0-9a-f]{24})?$" }
+            },
+            "inFlight": {
+              "type": ["object", "null"],
+              "additionalProperties": false,
+              "properties": {
+                "group": { "type": "integer", "minimum": 1 },
+                "state": { "enum": ["pending", "committed", "unknown"] }
+              }
+            },
+            "nextAction": { "type": "string" }
+          }
         }
       }
     }

package/src/change-analysis.js

@@ -313,7 +313,7 @@ function parseAnalysisJson(raw) {
   throw new SyntaxError('Response contains no complete JSON array.');
 }
 
-async function jsonCall(config, instruction, items, validate = null) {
+async function jsonCall(config, instruction, items, validate = null, stream = null) {
   const parseAndValidate = (raw) => {
     const parsed = parseAnalysisJson(raw);
     validate?.(parsed);
@@ -329,7 +329,7 @@ async function jsonCall(config, instruction, items, validate = null) {
     Math.min(2048, config.maxTokens || 1024),
     'Return the complete requested JSON array, preserving every required input ID exactly once. ' +
       `Required IDs: ${JSON.stringify(items.map((item) => item.id))}`,
-    null,
+    stream,
     (response) => {
       try {
         parseAndValidate(response);
@@ -339,6 +339,7 @@ async function jsonCall(config, instruction, items, validate = null) {
       }
     },
   );
+  stream?.onReasoningComplete?.(result.reasoning);
   try {
     return parseAndValidate(result.text);
   } catch (cause) {
@@ -712,7 +713,7 @@ export async function summarizeChanges(config, facts) {
   return `Structured change summaries (not raw diff):\n${JSON.stringify(items)}`;
 }
 
-export async function planAnalyzedChanges(config, facts, coverage = null) {
+export async function planAnalyzedChanges(config, facts, coverage = null, stream = null) {
   let candidates = facts;
   const deep = isDeepAnalysis(config) && coverage?.strategy !== 'auto';
   const policy = normalizeCommitPolicy(config.commitPolicy, config.language);
@@ -749,6 +750,9 @@ export async function planAnalyzedChanges(config, facts, coverage = null) {
               'Analysis plan is missing a commit subject.',
             );
         },
+        // Earlier batches may run many requests. Surface thinking from the
+        // final plan, which is the reasoning relevant to the review screen.
+        batches.length === 1 ? stream : null,
       );
       for (const group of groups) {
         next.push({

package/src/cli.js

@@ -16,13 +16,17 @@ function showHelp() {
     ${chalk.dim('$')} aicommit [path] [options]
     ${chalk.dim('$')} aicommit setup
     ${chalk.dim('$')} aicommit update
-    ${chalk.dim('$')} aicommit split [run|plan|apply|resume|abort] [options]
+    ${chalk.dim('$')} aicommit generate --scope=staged|all --file=<path> [options]
+    ${chalk.dim('$')} aicommit apply --file=<path> [options]
+    ${chalk.dim('$')} aicommit split [run|plan|apply|resume|abort|status] [options]
 
   ${chalk.bold('Everyday commands:')}
     setup                 Interactive configuration wizard
     update                Update the global npm installation to latest
     doctor                Diagnose runtime, config, credentials, and connectivity
     split                 Plan and create file-level logical commits
+    generate              Export one validated commit plan without committing
+    apply                 Apply a validated commit plan
 
   ${chalk.bold('Advanced commands:')}
     config show           Print the effective configuration with secrets redacted
@@ -35,6 +39,7 @@ function showHelp() {
     split apply           Validate and apply an exported JSON plan
     split resume          Resume the repository's unfinished transaction
     split abort           Discard recovery metadata; keep commits and changes
+    split status          Inspect recovery metadata without changing Git state
 
   ${chalk.bold('Arguments:')}
     path                  Target directory (default: current directory)
@@ -45,8 +50,8 @@ function showHelp() {
     -l, --lang=<zh|en>    Commit message language (default: zh)
     -p, --provider=<name> Use the named provider from config "providers"
     -m, --model=<name>    Use a named model from the selected provider
-    --scope=<scope>       Scope for "split|split plan": staged, all
-    --file=<path>         Split-plan artifact or commit-message file
+    --scope=<scope>       Scope for dry-run, generate, or split planning: staged, all
+    --file=<path>         Commit-plan artifact or commit-message file
     --range=<revision>    Git revision/range for "policy check" (default: HEAD)
     --reasoning=<level>   Set reasoning effort (enabled by default: medium)
     --no-reasoning        Explicitly disable reasoning when supported
@@ -91,6 +96,9 @@ function parsedDefaults(overrides = {}) {
     debug: false,
     split: null,
     splitCommand: null,
+    generate: false,
+    generateScope: null,
+    previewScope: null,
     splitPlanFile: null,
     dryRun: false,
     yes: false,
@@ -165,10 +173,17 @@ export function parseArgs(args = process.argv.slice(2)) {
     args = args.slice(2);
   }
 
+  const generate = !update && args[0] === 'generate';
+  if (generate) args = args.slice(1);
+
   let splitCommand = null;
+  if (!update && args[0] === 'apply') {
+    splitCommand = 'apply';
+    args = args.slice(1);
+  }
   if (!update && args[0] === 'split') {
     const requestedAction = args[1];
-    const actions = ['run', 'plan', 'apply', 'resume', 'abort'];
+    const actions = ['run', 'plan', 'apply', 'resume', 'abort', 'status'];
     if (!requestedAction || requestedAction.startsWith('-')) {
       // Keep the common path short: `aicommit split` is the interactive
       // split flow, while explicit actions remain available for automation
@@ -181,7 +196,7 @@ export function parseArgs(args = process.argv.slice(2)) {
     } else {
       throw fail(
         ERROR_CATEGORIES.CONFIG,
-        'split requires one action: run, plan, apply, resume, or abort.',
+        'split requires one action: run, plan, apply, resume, abort, or status.',
       );
     }
   }
@@ -197,6 +212,8 @@ export function parseArgs(args = process.argv.slice(2)) {
   let output = 'text';
   let debug = false;
   let split = null;
+  let generateScope = null;
+  let previewScope = null;
   let splitPlanFile = null;
   let policyMessageFile = null;
   let policyRange = null;
@@ -238,20 +255,26 @@ export function parseArgs(args = process.argv.slice(2)) {
 
     if (arg === '--scope') {
       splitScopeOption = true;
-      split = takeValue(args, i, arg, 'staged|all');
+      const scope = takeValue(args, i, arg, 'staged|all');
       i++;
-      if (!['staged', 'all'].includes(split)) {
-        throw fail(ERROR_CATEGORIES.CONFIG, `Invalid split scope: "${split}". Use staged or all.`);
+      if (!['staged', 'all'].includes(scope)) {
+        throw fail(ERROR_CATEGORIES.CONFIG, `Invalid split scope: "${scope}". Use staged or all.`);
       }
+      if (generate) generateScope = scope;
+      else if (splitCommand) split = scope;
+      else previewScope = scope;
       continue;
     }
 
     if (arg.startsWith('--scope=')) {
       splitScopeOption = true;
-      split = arg.slice('--scope='.length);
-      if (!['staged', 'all'].includes(split)) {
-        throw fail(ERROR_CATEGORIES.CONFIG, `Invalid split scope: "${split}". Use staged or all.`);
+      const scope = arg.slice('--scope='.length);
+      if (!['staged', 'all'].includes(scope)) {
+        throw fail(ERROR_CATEGORIES.CONFIG, `Invalid split scope: "${scope}". Use staged or all.`);
       }
+      if (generate) generateScope = scope;
+      else if (splitCommand) split = scope;
+      else previewScope = scope;
       continue;
     }
 
@@ -427,14 +450,32 @@ export function parseArgs(args = process.argv.slice(2)) {
       }
     }
   }
-  if (splitScopeOption && !['run', 'plan'].includes(splitCommand)) {
+  if (generate) {
+    if (!generateScope || !splitPlanFile) {
+      throw fail(
+        ERROR_CATEGORIES.CONFIG,
+        'generate requires --scope=staged|all and --file=<path>.',
+      );
+    }
+    if (splitCommand || allowSingleFallback) {
+      throw fail(ERROR_CATEGORIES.CONFIG, 'generate does not accept split commit options.');
+    }
+    dryRun = true;
+  }
+  if (splitScopeOption && !generate && splitCommand && !['run', 'plan'].includes(splitCommand)) {
     throw fail(
       ERROR_CATEGORIES.CONFIG,
-      '--scope is only valid with "aicommit split run" or "aicommit split plan".',
+      '--scope is only valid with dry-run, generate, split run, or split plan.',
     );
   }
-  if (splitPlanFile && !['plan', 'apply'].includes(splitCommand)) {
-    throw fail(ERROR_CATEGORIES.CONFIG, '--file is only valid with split plan or split apply.');
+  if (previewScope && !dryRun) {
+    throw fail(ERROR_CATEGORIES.CONFIG, '--scope without a subcommand requires --dry-run.');
+  }
+  if (splitPlanFile && !generate && !['plan', 'apply'].includes(splitCommand)) {
+    throw fail(
+      ERROR_CATEGORIES.CONFIG,
+      '--file is only valid with generate, apply, or split plan/apply.',
+    );
   }
   if (allowSingleFallback && (splitCommand !== 'run' || !yes || dryRun)) {
     throw fail(
@@ -508,6 +549,9 @@ export function parseArgs(args = process.argv.slice(2)) {
     debug,
     split,
     splitCommand,
+    generate,
+    generateScope,
+    previewScope,
     splitPlanFile,
     dryRun,
     yes,

package/src/completion.js

@@ -5,6 +5,8 @@ const TOP_LEVEL = [
   'config',
   'policy',
   'completion',
+  'generate',
+  'apply',
   'split',
   '--help',
   '--version',
@@ -33,7 +35,7 @@ _aicommit() {
     config) COMPREPLY=( $(compgen -W "show validate path --provider --model --output --debug" -- "$cur") ); return ;;
     policy) COMPREPLY=( $(compgen -W "template check --file --range --output --debug" -- "$cur") ); return ;;
     completion) COMPREPLY=( $(compgen -W "bash zsh fish" -- "$cur") ); return ;;
-    split) COMPREPLY=( $(compgen -W "run plan apply resume abort --scope --file --yes --allow-single-fallback --output --debug" -- "$cur") ); return ;;
+    split) COMPREPLY=( $(compgen -W "run plan apply resume abort status --scope --file --yes --allow-single-fallback --output --debug" -- "$cur") ); return ;;
   esac
   case "$prev" in
     --lang|-l) COMPREPLY=( $(compgen -W "zh en" -- "$cur") ); return ;;
@@ -58,7 +60,9 @@ _aicommit() {
     'config:inspect or validate configuration'
     'policy:print or enforce a repository team policy'
     'completion:generate shell completion'
-    'split:run, plan, apply, resume, or abort split commits'
+    'generate:export one commit plan'
+    'apply:apply a validated commit plan'
+    'split:run, plan, apply, resume, abort, or inspect split commits'
   )
   options=(
     '--help[show help]'
@@ -66,8 +70,8 @@ _aicommit() {
     '--lang=[commit language]:language:(zh en)'
     '--provider=[provider name]:provider:'
     '--model=[model name]:model:'
-    '--scope=[split scope]:scope:(staged all)'
-    '--file=[split plan or commit-message file]:path:_files'
+    '--scope=[change scope]:scope:(staged all)'
+    '--file=[commit plan or commit-message file]:path:_files'
     '--range=[Git range for policy check]:revision:'
     '--reasoning=[reasoning effort]:effort:(low medium high xhigh max)'
     '--no-reasoning[disable reasoning]'
@@ -87,7 +91,7 @@ _aicommit() {
         config) _values 'config action' show validate path ;;
         policy) _values 'policy action' template check ;;
         completion) _values 'shell' bash zsh fish ;;
-        split) _values 'split action' run plan apply resume abort ;;
+        split) _values 'split action' run plan apply resume abort status ;;
       esac
       ;;
   esac
@@ -103,18 +107,20 @@ complete -c aicommit -n '__fish_use_subcommand' -a doctor -d 'Diagnose configura
 complete -c aicommit -n '__fish_use_subcommand' -a config -d 'Inspect or validate configuration'
 complete -c aicommit -n '__fish_use_subcommand' -a policy -d 'Print or enforce a repository team policy'
 complete -c aicommit -n '__fish_use_subcommand' -a completion -d 'Generate shell completion'
+complete -c aicommit -n '__fish_use_subcommand' -a generate -d 'Export one commit plan'
+complete -c aicommit -n '__fish_use_subcommand' -a apply -d 'Apply a validated commit plan'
 complete -c aicommit -n '__fish_use_subcommand' -a split -d 'Run, plan, apply, resume, or abort split commits'
 complete -c aicommit -n '__fish_seen_subcommand_from config' -a 'show validate path'
 complete -c aicommit -n '__fish_seen_subcommand_from policy' -a 'template check'
 complete -c aicommit -n '__fish_seen_subcommand_from completion' -a 'bash zsh fish'
-complete -c aicommit -n '__fish_seen_subcommand_from split' -a 'run plan apply resume abort'
+complete -c aicommit -n '__fish_seen_subcommand_from split' -a 'run plan apply resume abort status'
 complete -c aicommit -s h -l help -d 'Show help'
 complete -c aicommit -s v -l version -d 'Show version'
 complete -c aicommit -s l -l lang -x -a 'zh en' -d 'Commit language'
 complete -c aicommit -s p -l provider -x -d 'Provider name'
 complete -c aicommit -s m -l model -x -d 'Model name'
-complete -c aicommit -l scope -x -a 'staged all' -d 'Split plan scope'
-complete -c aicommit -l file -r -d 'Split plan file'
+complete -c aicommit -l scope -x -a 'staged all' -d 'Change scope'
+complete -c aicommit -l file -r -d 'Commit plan file'
 complete -c aicommit -l range -r -d 'Git range for policy check'
 complete -c aicommit -l reasoning -x -a 'low medium high xhigh max' -d 'Reasoning effort'
 complete -c aicommit -l no-reasoning -d 'Disable reasoning'

package/src/errors.js

@@ -38,8 +38,13 @@ export class AicommitError extends Error {
     super(message);
     this.name = 'AicommitError';
     this.category = category;
+    this.code = options.code || category;
     this.exitCode = EXIT_CODES[category] ?? EXIT_CODES.internal;
     this.reported = Boolean(options.reported);
+    this.retryable = Boolean(options.retryable);
+    this.nextAction = options.nextAction || null;
+    this.committed = Boolean(options.committed);
+    this.commitState = options.commitState || 'none';
     if (options.data && typeof options.data === 'object' && !Array.isArray(options.data)) {
       this.data = options.data;
     }
@@ -62,11 +67,10 @@ export function classifyError(err) {
   const lower = message.toLowerCase();
   const code = errorCode(err);
   if (
-    err instanceof TypeError ||
     NETWORK_CODES.has(code) ||
     /timed out|fetch failed|socket|network|dns|econn|enotfound/.test(lower)
   ) {
-    return fail(ERROR_CATEGORIES.NETWORK, message, { cause: err });
+    return fail(ERROR_CATEGORIES.NETWORK, message, { cause: err, retryable: true });
   }
   if (/^http \d{3}:|streaming api error|rate limit|provider request/.test(lower)) {
     return fail(ERROR_CATEGORIES.PROVIDER, message, { cause: err });

package/src/generation-ui.js

@@ -18,7 +18,7 @@ export async function runModelTask({
   const spinner = ora({ text: chalk.dim(spinnerText), color: 'cyan' }).start();
   let liveReasoning;
   const stream =
-    reasoning?.mode === 'on' && !machineOutput
+    reasoning && reasoning.mode !== 'off' && !machineOutput
       ? {
           onReasoningDelta(chunk) {
             if (!liveReasoning) {

package/src/git.js

@@ -70,13 +70,24 @@ export function getStagedDiff(cwd, contextLines) {
   return readGit(['diff', unifiedArg(contextLines), '--staged'], cwd).trim();
 }
 
-// Hash the complete staged patch (including binary changes and full object
-// ids) so a long-running AI/review step cannot silently commit a different
-// index from the one used to generate the message.
+// Git's raw staged diff contains the full old/new blob IDs, modes, status,
+// and NUL-delimited paths for every changed entry. Hashing that metadata is
+// equivalent to hashing the staged patch for snapshot identity, including
+// binary changes, without regenerating and spooling the entire patch during
+// every pre/post-model safety check.
 export function getIndexFingerprint(cwd) {
   return updateGitHash(
     createHash('sha256'),
-    ['diff', '--staged', '--binary', '--full-index', '--no-ext-diff'],
+    [
+      'diff',
+      '--raw',
+      '--staged',
+      '--no-renames',
+      '--no-ext-diff',
+      '--no-textconv',
+      '--abbrev=64',
+      '-z',
+    ],
     cwd,
   ).digest('hex');
 }
@@ -263,7 +274,8 @@ export function condenseDiff(diff, maxChars, stat, maxSectionChars = Infinity) {
   }
   const marker = `... (diff truncated — ${diff.length} chars total)`;
   const body = kept ? `${kept}\n${marker}` : marker;
-  return { diff: stat ? `${stat}\n\n${body}` : body, truncated: true };
+  const resolvedStat = typeof stat === 'function' ? stat() : stat;
+  return { diff: resolvedStat ? `${resolvedStat}\n\n${body}` : body, truncated: true };
 }
 
 export function getChangedFiles(cwd) {

package/src/main.js

@@ -17,6 +17,8 @@ import {
   runGit,
   isGitRepo,
   getIndexFingerprint,
+  hasHead,
+  readGit,
   createIndexTransaction,
   protectSensitiveDiff,
   unifiedArg,
@@ -54,8 +56,13 @@ import {
   applySplitPlan,
   resumeSplit,
   splitFlow,
+  splitStatus,
   getStagedChangedFiles,
+  getAllChangedFiles,
+  getSplitStateFingerprint,
+  safeExportPlanPath,
 } from './split.js';
+import { createSplitPlanArtifact, writeSplitPlanArtifact } from './split-plan.js';
 import { runModelTask } from './generation-ui.js';
 import { runSetup } from './setup.js';
 import { detectProviderType } from './providers.js';
@@ -108,6 +115,9 @@ async function runMain() {
     debug,
     split,
     splitCommand,
+    generate,
+    generateScope,
+    previewScope,
     splitPlanFile,
     dryRun,
     yes,
@@ -130,7 +140,15 @@ async function runMain() {
     return { exitReason: 'completion' };
   }
   const machineOutput = output === 'json';
-  if (machineOutput && !yes && !doctor && !configAction && !policyAction && !update) {
+  if (
+    machineOutput &&
+    !yes &&
+    !doctor &&
+    !configAction &&
+    !policyAction &&
+    !update &&
+    splitCommand !== 'status'
+  ) {
     throw fail(ERROR_CATEGORIES.CONFIG, '--output=json requires --yes for commit and split flows.');
   }
 
@@ -178,13 +196,14 @@ async function runMain() {
   console.log('  ' + chalk.dim('─'.repeat(45)));
   console.log('  ' + chalk.dim(`Working directory: ${sanitizeTerminalText(process.cwd())}`));
 
-  if (['apply', 'resume', 'abort'].includes(splitCommand)) {
+  if (['apply', 'resume', 'abort', 'status'].includes(splitCommand)) {
     const projectRoot = getProjectRoot();
     if (!isGitRepo(projectRoot)) {
       throw fail(ERROR_CATEGORIES.GIT_STATE, `Not a git repository: ${process.cwd()}`);
     }
     if (splitCommand === 'resume') return resumeSplit(projectRoot, { yes, machineOutput });
     if (splitCommand === 'abort') return abortSplit(projectRoot, { yes, machineOutput });
+    if (splitCommand === 'status') return splitStatus(projectRoot);
     return applySplitPlan(projectRoot, splitPlanFile, { yes, machineOutput });
   }
 
@@ -340,6 +359,35 @@ async function runMain() {
   // All git commands run at the repo root (projectRoot), even when aicommit
   // is invoked from a subdirectory.
   let indexTransaction = null;
+  const effectiveScope = generate ? generateScope : previewScope;
+  let stagedAllForPreview = false;
+  const generateSnapshot = generate
+    ? (() => {
+        const changes =
+          generateScope === 'staged'
+            ? getStagedChangedFiles(projectRoot)
+            : getAllChangedFiles(projectRoot);
+        if (!changes.length) {
+          throw fail(
+            ERROR_CATEGORIES.GIT_STATE,
+            `No ${generateScope} changes to generate a plan for.`,
+          );
+        }
+        return {
+          changes,
+          baseHead: hasHead(projectRoot)
+            ? readGit(['rev-parse', 'HEAD'], projectRoot).trim()
+            : null,
+          fingerprint: getSplitStateFingerprint(
+            projectRoot,
+            hasHead(projectRoot),
+            changes,
+            generateScope,
+          ),
+          file: safeExportPlanPath(projectRoot, splitPlanFile),
+        };
+      })()
+    : null;
   const finishCancelled = ({
     notice = 'Commit cancelled.',
     message = null,
@@ -374,6 +422,20 @@ async function runMain() {
     indexTransaction ||= createIndexTransaction(projectRoot);
     return indexTransaction;
   };
+  if (effectiveScope === 'all') {
+    try {
+      beginIndexTransaction();
+      runGit(['add', '-A'], projectRoot);
+      indexTransaction.markOwned();
+      stagedAllForPreview = true;
+    } catch (err) {
+      indexTransaction?.restore({ force: true });
+      indexTransaction = null;
+      throw fail(ERROR_CATEGORIES.GIT_STATE, `Failed to stage plan snapshot: ${err.message}`, {
+        cause: err,
+      });
+    }
+  }
   if (!getChangedFiles(projectRoot).length) {
     // Nothing staged. But git diff --staged is also empty for unstaged work
     // and untracked files — surface what git status actually shows instead of
@@ -390,6 +452,10 @@ async function runMain() {
       throw fail(ERROR_CATEGORIES.GIT_STATE, 'No changes to commit.', { reported: true });
     }
 
+    if (effectiveScope === 'staged' || generate) {
+      throw fail(ERROR_CATEGORIES.GIT_STATE, `No ${effectiveScope} changes to preview.`);
+    }
+
     console.log('\n  ' + chalk.yellow(`✗ No staged changes — ${tips.join(', ')}.`));
     printChangeList(withUntrackedFiles(unstaged, untracked));
     console.log('');
@@ -531,6 +597,13 @@ async function runMain() {
   let diffForModel = diff;
   let protectAnalysis = true;
   if (protectedInput.findings.length) {
+    if (generate && generateScope === 'all' && yes) {
+      throw fail(
+        ERROR_CATEGORIES.SENSITIVE_DATA,
+        'Non-interactive generate --scope=all refuses sensitive content; review and stage intended files explicitly.',
+        { code: 'sensitive_all_scope' },
+      );
+    }
     warnings.push('Sensitive data was detected and protected before the provider request.');
     console.log('\n  ' + chalk.yellow.bold('⚠ Potential sensitive data detected:'));
     for (const finding of protectedInput.findings) {
@@ -571,7 +644,7 @@ async function runMain() {
     : condenseDiff(
         strippedDiff,
         config.maxDiffChars,
-        getDiffStat(projectRoot),
+        () => getDiffStat(projectRoot),
         config.maxFileDiffChars,
       );
   let analysis;
@@ -723,7 +796,7 @@ async function runMain() {
       ? 'use'
       : await confirmAction(
           message,
-          reasoningEnabled
+          reasoningEnabled || reasoningText
             ? {
                 text: reasoningText,
                 maxChars: config.reasoning.maxDisplayChars,
@@ -773,14 +846,51 @@ async function runMain() {
   console.log('');
 
   if (dryRun) {
+    stagedAllForPreview ||= Boolean(indexTransaction);
     const restored = indexTransaction ? indexTransaction.restore() : true;
     indexTransaction = null;
     if (!restored) {
+      if (generate) {
+        throw fail(
+          ERROR_CATEGORIES.CONCURRENT_MODIFICATION,
+          'The Git index changed while generating a commit plan; no plan was written.',
+        );
+      }
       warnings.push('The Git index changed during the run and was left untouched.');
       console.log(
         '  ' + chalk.yellow('⚠ The Git index changed during the run and was left untouched.'),
       );
     }
+    let plan = null;
+    let planFile = null;
+    if (generate) {
+      const fingerprint = getSplitStateFingerprint(
+        projectRoot,
+        hasHead(projectRoot),
+        undefined,
+        generateScope,
+      );
+      const baseHead = hasHead(projectRoot)
+        ? readGit(['rev-parse', 'HEAD'], projectRoot).trim()
+        : null;
+      if (fingerprint !== generateSnapshot.fingerprint || baseHead !== generateSnapshot.baseHead) {
+        throw fail(
+          ERROR_CATEGORIES.CONCURRENT_MODIFICATION,
+          'Repository changes changed while generating a commit plan; no plan was written.',
+        );
+      }
+      plan = [{ message, files: generateSnapshot.changes.map((change) => change.path) }];
+      const artifact = createSplitPlanArtifact({
+        scope: generateScope,
+        baseHead,
+        fingerprint,
+        language: config.language,
+        commitPolicy: config.commitPolicy,
+        changes: generateSnapshot.changes,
+        groups: plan,
+      });
+      planFile = await writeSplitPlanArtifact(generateSnapshot.file, artifact);
+    }
     console.log(
       '  ' +
         chalk.green.bold(
@@ -789,6 +899,10 @@ async function runMain() {
     );
     return {
       message,
+      plan,
+      planFile,
+      scope: effectiveScope || (stagedAllForPreview || indexTransaction ? 'all' : 'staged'),
+      changeCount: generate ? generateSnapshot.changes.length : changedFiles.length,
       provider: selectedProvider,
       model: config.modelId,
       latencyMs: elapsed,
@@ -835,6 +949,9 @@ async function runMain() {
     console.log('\n  ' + chalk.green.bold('✓ Done!\n'));
     return {
       message,
+      commitSha: readGit(['rev-parse', 'HEAD'], projectRoot).trim(),
+      scope: 'staged',
+      changeCount: changedFiles.length,
       provider: selectedProvider,
       model: config.modelId,
       latencyMs: elapsed,

package/src/output.js

@@ -1,6 +1,10 @@
 import { classifyError, EXIT_CODES } from './errors.js';
 
-export const OUTPUT_SCHEMA_VERSION = '1.0';
+export const OUTPUT_SCHEMA_VERSION = '1.1';
+
+function safeString(value) {
+  return typeof value === 'string' && value ? value : null;
+}
 
 function normalizedUsage(usage) {
   if (!usage) return null;
@@ -43,6 +47,11 @@ export function successOutput(result = {}) {
     warnings: Array.isArray(result.warnings) ? result.warnings.map(String) : [],
     exitReason: result.exitReason || 'success',
     committed: Boolean(result.committed),
+    commitState: result.commitState || (result.committed ? 'complete' : 'none'),
+    commitSha: safeString(result.commitSha),
+    planFile: safeString(result.planFile),
+    scope: safeString(result.scope),
+    changeCount: Number.isInteger(result.changeCount) ? result.changeCount : null,
     error: null,
   };
   if (result.data && typeof result.data === 'object' && !Array.isArray(result.data)) {
@@ -64,10 +73,18 @@ export function errorOutput(err) {
     usage: null,
     warnings: [],
     exitReason: classified.category,
-    committed: false,
+    committed: classified.committed,
+    commitState: classified.commitState,
+    commitSha: null,
+    planFile: null,
+    scope: null,
+    changeCount: null,
     error: {
       category: classified.category,
+      code: classified.code,
       message: classified.message,
+      retryable: classified.retryable,
+      nextAction: classified.nextAction,
     },
   };
   if (classified.data && typeof classified.data === 'object' && !Array.isArray(classified.data)) {

package/src/split.js

@@ -1163,6 +1163,83 @@ function currentHeadParent(projectRoot) {
   return fields[1] || null;
 }
 
+function splitCheckpointSummary(projectRoot, loaded) {
+  const { path, checkpoint } = loaded;
+  const knownCommits = checkpoint.completed.map((record) => record.commit);
+  let inFlight = null;
+  if (checkpoint.inFlight) {
+    const record = checkpoint.inFlight;
+    const head = currentHead(projectRoot);
+    const state =
+      head === record.parent
+        ? 'pending'
+        : currentHeadParent(projectRoot) === record.parent &&
+            currentHeadTree(projectRoot) === record.tree
+          ? 'committed'
+          : 'unknown';
+    if (state === 'committed') knownCommits.push(head);
+    inFlight = { group: record.index + 1, state };
+  }
+  const totalGroups = checkpoint.plan.groups.length;
+  return {
+    state: 'pending',
+    transactionId: checkpoint.transactionId,
+    checkpointPath: path,
+    scope: checkpoint.plan.scope,
+    totalGroups,
+    completedGroups: knownCommits.length,
+    pendingGroups: totalGroups - knownCommits.length,
+    completedCommits: knownCommits,
+    inFlight,
+    nextAction: 'aicommit split resume --yes',
+  };
+}
+
+export function splitStatus(projectRoot) {
+  let summary;
+  try {
+    summary = splitCheckpointSummary(projectRoot, readSplitCheckpoint(projectRoot));
+  } catch (err) {
+    if (!/No split checkpoint found:/.test(err.message)) {
+      throw fail(ERROR_CATEGORIES.GIT_STATE, `Cannot inspect split checkpoint: ${err.message}`, {
+        code: 'split_checkpoint_invalid',
+        cause: err,
+      });
+    }
+    summary = { state: 'idle' };
+  }
+  const known = summary.completedCommits || [];
+  const commitState =
+    summary.inFlight?.state === 'unknown'
+      ? 'unknown'
+      : !known.length
+        ? 'none'
+        : summary.pendingGroups
+          ? 'partial'
+          : 'complete';
+  return {
+    exitReason: 'split_status',
+    committed: false,
+    commitState,
+    commitSha: known.at(-1) || null,
+    scope: summary.scope || null,
+    data: { split: summary },
+  };
+}
+
+function splitCommitFailure(projectRoot, message) {
+  const status = splitStatus(projectRoot);
+  const knownCommits = status.data.split.completedCommits || [];
+  throw fail(ERROR_CATEGORIES.GIT_STATE, message, {
+    code: knownCommits.length ? 'split_partial_failure' : 'split_commit_failed',
+    reported: true,
+    committed: knownCommits.length > 0,
+    commitState: status.commitState,
+    nextAction: 'aicommit split resume --yes',
+    data: status.data,
+  });
+}
+
 function readHeadEntries(projectRoot, paths) {
   const entries = new Map();
   if (!hasHead(projectRoot) || !paths.length) return entries;
@@ -1717,6 +1794,18 @@ export async function splitFlow(
       task: async (stream) => {
         if (large) {
           let plan;
+          let planReasoning = null;
+          const planningStream = stream
+            ? {
+                onReasoningDelta(chunk) {
+                  planReasoning = (planReasoning || '') + chunk;
+                  stream.onReasoningDelta(chunk);
+                },
+                onReasoningComplete(text) {
+                  if (text) planReasoning = text;
+                },
+              }
+            : null;
           try {
             analysis = await analyzeChanges(
               planningConfig,
@@ -1729,7 +1818,12 @@ export async function splitFlow(
                 ),
               persistentAnalysisCache,
             );
-            plan = await planAnalyzedChanges(planningConfig, analysis.facts, analysis.coverage);
+            plan = await planAnalyzedChanges(
+              planningConfig,
+              analysis.facts,
+              analysis.coverage,
+              planningStream,
+            );
           } catch (err) {
             const exhausted =
               err.data?.analysis?.exhausted ||
@@ -1759,6 +1853,7 @@ export async function splitFlow(
               fallbackReason: exhausted || 'planning_capacity',
             };
             plan = normalizePlan([], allFiles, config.language, config.commitPolicy);
+            planReasoning = null;
             const warning = `Large-change planning used one conservative all-files commit after ${exhausted || 'planning capacity'} exhaustion; review the fallback message and grouping.`;
             warnings.push(warning);
             console.error(`  Fallback: ${warning}`);
@@ -1781,7 +1876,7 @@ export async function splitFlow(
             raw: JSON.stringify(plan),
             elapsed: planningConfig.analysisBudget.snapshot().elapsedMs,
             usage: planningConfig.analysisBudget.snapshot().usage,
-            reasoning: null,
+            reasoning: planReasoning,
           };
         }
         return generateSplitPlan(
@@ -1876,7 +1971,7 @@ export async function splitFlow(
               { name: 'Cancel', value: 'cancel', description: 'Abort without committing' },
             ],
           },
-          reasoningEnabled
+          reasoningEnabled || reasoningText
             ? {
                 text: reasoningText,
                 maxChars: config.reasoning.maxDisplayChars,
@@ -2013,6 +2108,8 @@ export async function splitFlow(
     return {
       plan: groups,
       planFile: writtenPlanPath,
+      scope,
+      changeCount: allFiles.length,
       provider,
       model: config.modelId,
       latencyMs: elapsed,
@@ -2054,12 +2151,13 @@ export async function splitFlow(
   if (ok) {
     console.log('\n  ' + chalk.green.bold(`✓ Done! Created ${groups.length} commits.\n`));
   } else {
-    throw fail(ERROR_CATEGORIES.GIT_STATE, 'One or more split commits failed.', {
-      reported: true,
-    });
+    splitCommitFailure(projectRoot, 'One or more split commits failed.');
   }
   return {
     plan: groups,
+    commitSha: currentHead(projectRoot),
+    scope,
+    changeCount: allFiles.length,
     provider,
     model: config.modelId,
     latencyMs: elapsed,
@@ -2168,9 +2266,13 @@ export async function resumeSplit(projectRoot, { yes = false, machineOutput = fa
     );
     return {
       plan: checkpoint.plan.groups,
+      commitSha: currentHead(projectRoot),
+      scope: checkpoint.plan.scope,
+      changeCount: checkpoint.plan.changes.length,
       warnings: [],
       exitReason: 'success',
-      committed: true,
+      committed: false,
+      commitState: 'complete',
       edited: false,
       rewrites: 0,
     };
@@ -2225,9 +2327,7 @@ export async function resumeSplit(projectRoot, { yes = false, machineOutput = fa
     { transaction: { path: transaction.path, checkpoint } },
   );
   if (!ok) {
-    throw fail(ERROR_CATEGORIES.GIT_STATE, 'One or more resumed split commits failed.', {
-      reported: true,
-    });
+    splitCommitFailure(projectRoot, 'One or more resumed split commits failed.');
   }
   console.log(
     '\n  ' +
@@ -2237,6 +2337,9 @@ export async function resumeSplit(projectRoot, { yes = false, machineOutput = fa
   );
   return {
     plan: checkpoint.plan.groups,
+    commitSha: currentHead(projectRoot),
+    scope: checkpoint.plan.scope,
+    changeCount: checkpoint.plan.changes.length,
     provider: null,
     model: null,
     latencyMs: null,
@@ -2443,13 +2546,15 @@ export async function applySplitPlan(
     { planArtifact: artifact },
   );
   if (!ok) {
-    throw fail(ERROR_CATEGORIES.GIT_STATE, 'One or more split commits failed.', {
-      reported: true,
-    });
+    splitCommitFailure(projectRoot, 'One or more split commits failed.');
   }
   console.log('\n  ' + chalk.green.bold(`✓ Done! Created ${artifact.groups.length} commits.\n`));
   return {
     plan: artifact.groups,
+    planFile: loaded.path,
+    commitSha: readGit(['rev-parse', 'HEAD'], projectRoot).trim(),
+    scope: artifact.scope,
+    changeCount: artifact.changes.length,
     provider: null,
     model: null,
     latencyMs: null,