@zvec/zvec-grep 0.1.7 → 0.1.8

package/README.md

@@ -3,9 +3,6 @@
 </p>
 
 <div align="center">
-  <p>
-    <a href="./docs/08-roadmap.md"><img src="https://img.shields.io/badge/status-work%20in%20progress-F59E0B?style=for-the-badge" alt="Work in progress" /></a>
-  </p>
   <picture>
     <source media="(prefers-color-scheme: dark)" srcset="./.github/assets/zg-logo-dark.svg">
     <img src="./.github/assets/zg-logo.svg" width="150" alt="zg logo" />
@@ -42,9 +39,6 @@ terminal, or let your agent use it for you.
   <img src="./.github/assets/zvec-grep-tour.gif" width="1000" alt="Install the agent integration, index a workspace, and let the agent search it with zvec-grep" />
 </div>
 
-Install the integration once, index the workspace, then search from the terminal
-or ask your agent naturally.
-
 <a id="features"></a>
 
 ## 💫 Why zg?
@@ -64,47 +58,36 @@ or ask your agent naturally.
 
 ## 🚀 Try it yourself
 
-Requires Node.js 22 or newer and a configured
-[OpenCode](https://opencode.ai/) installation. The agent example explicitly
-selects a free model, so it does not depend on your default provider.
-
 ### 1. Set up a sample bookshelf
 
 ```bash
+# Requires Node.js 22 or newer.
 npm install -g @zvec/zvec-grep
-zg install --target opencode --yes
 
 mkdir zg-mystery && cd zg-mystery
-curl -L https://www.gutenberg.org/files/11/11-0.txt -o alice-in-wonderland.txt
-curl -L https://www.gutenberg.org/files/345/345-0.txt -o dracula.txt
-curl -L https://www.gutenberg.org/files/84/84-0.txt -o frankenstein.txt
-curl -L https://www.gutenberg.org/files/834/834-0.txt -o sherlock-holmes.txt
+curl --retry 3 --retry-all-errors --progress-bar -fL \
+  -o alice-in-wonderland.txt https://raw.githubusercontent.com/GITenberg/Alice-s-Adventures-in-Wonderland_11/master/11.txt \
+  -o sherlock-holmes.txt https://raw.githubusercontent.com/GITenberg/The-Memoirs-of-Sherlock-Holmes_834/master/834.txt
 
 zg index --embedding local/potion-retrieval-32m
 ```
 
-This creates a small bookshelf of four public-domain novels and indexes it
-locally with Potion Retrieval.
-
-Installation starts the shared daemon immediately. The default MCP transport is
-stdio: the agent runs `zg server --stdio`, which safely starts or reuses it
-again after a reboot. Advanced users can keep direct HTTP configuration with
-`zg install --mcp-transport http` and `zg server on`.
-
-Use `zg install --mcp-toolset full` when the installed MCP integration should
-also expose managed rg, index, and status tools. The default is `agent`.
-
 ### 2. Choose how to search
 
 #### For agents: ask with OpenCode
 
+With [OpenCode](https://opencode.ai/) configured:
+
 ```bash
+zg install --target opencode --yes
 opencode run --model opencode/deepseek-v4-flash-free \
   "An unseen creature left a few marks. What did the detective infer? Cite local evidence."
 ```
 
-OpenCode chooses zg on its own—the prompt does not name a tool. Query rewrites
-may vary by model, but a typical run looks like this:
+OpenCode chooses zg on its own—the prompt does not name a tool.
+
+<details>
+<summary><strong>See the full agent run and answer</strong></summary>
 
 ```text
 ⚙ zvec_grep_zvec_grep_search {"root":".../zg-mystery","query":"An unseen creature left a few marks. What did the detective infer?","fts":["marks","unseen creature","inferred","detective"],"fuse":true}
@@ -128,6 +111,8 @@ Holmes also inferred that a man entered through the window with it and later
 left with the key (sherlock-holmes.txt:5464-5470, 5527-5528).
 ```
 
+</details>
+
 #### For humans: search directly
 
 Search the same bookshelf directly, without an agent:
@@ -136,60 +121,32 @@ Search the same bookshelf directly, without an agent:
 zg query --human "An unseen creature left a few marks. What did the detective infer?" --limit 3
 ```
 
-zg returns the relevant passages from `sherlock-holmes.txt`, ranked ahead of the
-other novels.
+zg returns the relevant passages from `sherlock-holmes.txt`, ranked ahead of
+`alice-in-wonderland.txt`.
 
 <a id="benchmarks"></a>
 
 ## 📊 Benchmarks
 
-zg is a general-purpose retrieval tool. Current benchmarks cover code and text
-retrieval.
-
-### Repository code retrieval
-
-We use [SWE-QA-Bench](https://github.com/peng-weihan/SWE-QA-Bench), which
-requires cross-file and multi-hop reasoning over real-world codebases.
-
-- **Coverage:** 20 retrieval-intensive tasks covering the paper's four
-  top-level categories—**What, Where, How, and Why**—across 8 intentions and
-  11 repositories.
-- **Agent:** Claude Code with **Claude Opus 5**.
-- **Embedding:** **Qwen3.7 Text Embedding** for the zg profile.
-- **Protocol:** 3 runs per task and profile.
-
-| Profile | Judge /100 ↑ | Input tokens ↓ | Tool calls ↓ | Time (s) ↓ | Cost ↓ |
-| --- | ---: | ---: | ---: | ---: | ---: |
-| Native tools | 80.42 | 558,651 | 23.42 | 127.5 | $0.905 |
-| Native tools + zg | 81.92 | 294,262 | 9.70 | 79.7 | $0.558 |
-| Change | **+1.50 pp** | **−47.3%** | **−58.6%** | **−37.5%** | **−38.3%** |
-
-Across 20 tasks × 3 runs, zg increased the Judge score by 1.50 points while
-reducing input tokens by 47.3%, tool calls by 58.6%, execution time by 37.5%,
-and cost by 38.3%.
-
-### Large-scale text retrieval
-
-We use [BrowseComp-Plus](https://github.com/texttron/BrowseComp-Plus), which
-requires multi-document evidence retrieval over a fixed web corpus. The
-evaluation includes 30 queries covering 10 mapped topics and three
-evidence-breadth levels, with 3 runs per profile.
-
-| Profile | Accuracy ↑ | Evidence recall ↑ | Search calls ↓ | Input tokens ↓ | Time (s) ↓ | Cost ↓ |
-| --- | ---: | ---: | ---: | ---: | ---: | ---: |
-| Agent + BM25 | TBD | TBD | TBD | TBD | TBD | TBD |
-| Agent + zg | TBD | TBD | TBD | TBD | TBD | TBD |
-| Change | TBD | TBD | TBD | TBD | TBD | TBD |
-
-Across 30 queries × 3 runs, zg changed accuracy by TBD pp, evidence recall by
-TBD pp, search calls by TBD%, input tokens by TBD%, time by TBD%, and cost by
-TBD%.
-
-To reproduce the evaluation, follow the
-[benchmark guide](./benchmarks/README.md) for environment setup, pinned
-revisions, task manifests, baseline configuration, exact commands, and result
-aggregation. All profiles use the same agent, model, prompt, tool budget, and
-three-run protocol.
+zg helps agents find relevant evidence faster by **narrowing the effective
+search space** — using **fewer tokens, tool calls, and less time** while
+preserving answer quality.
+
+Each benchmark follows a **controlled, reproducible paired A/B protocol**: the
+same agent runs the same pinned tasks with identical model, prompt, environment,
+and limits; **only `zg` access and usage guidance differ**.
+
+<img src="./.github/assets/benchmark-comparison-v2.svg" alt="Paired baseline and zvec-grep comparison for SWE-QA-Bench and BrowseComp-Plus" width="900" />
+
+| Benchmark | Answer&nbsp;quality&nbsp;↑ | Avg.&nbsp;input&nbsp;tokens&nbsp;↓ | Avg.&nbsp;tool&nbsp;calls&nbsp;↓ | Avg.&nbsp;agent&nbsp;time&nbsp;↓ |
+| --- | ---: | ---: | ---: | ---: |
+| [**SWE-QA-Bench**](./benchmarks/swe-qa-bench/README.md)<br>20 tasks · Codebase QA | Judge Score<br>80.42&nbsp;→&nbsp;81.92 | 558,651&nbsp;→&nbsp;294,262 | 23.42&nbsp;→&nbsp;9.70 | 127.5s&nbsp;→&nbsp;79.7s |
+| [**BrowseComp-Plus**](./benchmarks/browse-comp-plus/README.md)<br>80 cases · Deep-research QA | Accuracy<br>90.00%&nbsp;→&nbsp;90.00% | 2.04M&nbsp;→&nbsp;1.19M | 22.70&nbsp;→&nbsp;14.24 | 284.8s&nbsp;→&nbsp;199.3s |
+
+Values show **Baseline → zg**.
+
+See the [benchmark documentation](./benchmarks/README.md) for full results and
+reproduction details.
 
 ## 📚 Documentation
 

package/README_CN.md

@@ -3,9 +3,6 @@
 </p>
 
 <div align="center">
-  <p>
-    <a href="./docs/08-roadmap.md"><img src="https://img.shields.io/badge/status-work%20in%20progress-F59E0B?style=for-the-badge" alt="项目开发中" /></a>
-  </p>
   <picture>
     <source media="(prefers-color-scheme: dark)" srcset="./.github/assets/zg-logo-dark.svg">
     <img src="./.github/assets/zg-logo.svg" width="150" alt="zg logo" />
@@ -42,9 +39,6 @@
   <img src="./.github/assets/zvec-grep-tour.gif" width="1000" alt="安装 Agent 集成、为工作区建索引并让 Agent 使用 zvec-grep 检索本地内容" />
 </div>
 
-只需安装一次集成并为工作区建索引，之后既可以在终端中搜索，也可以直接向
-Agent 提问。
-
 <a id="features"></a>
 
 ## 💫 为什么选择 zg？
@@ -64,67 +58,61 @@ Agent 提问。
 
 ## 🚀 动手体验
 
-需要 Node.js 22 或更新版本，以及已经完成配置的
-[OpenCode](https://opencode.ai/)。Agent 示例会显式选择免费模型，不依赖你配置的
-默认模型提供方。
-
 ### 1. 准备示例书架
 
 ```bash
+# 需要 Node.js 22 或更新版本。
 npm install -g @zvec/zvec-grep
-zg install --target opencode --yes
 
 mkdir zg-mystery && cd zg-mystery
-curl -L https://www.gutenberg.org/files/11/11-0.txt -o alice-in-wonderland.txt
-curl -L https://www.gutenberg.org/files/345/345-0.txt -o dracula.txt
-curl -L https://www.gutenberg.org/files/84/84-0.txt -o frankenstein.txt
-curl -L https://www.gutenberg.org/files/834/834-0.txt -o sherlock-holmes.txt
+curl --retry 3 --retry-all-errors --progress-bar -fL \
+  -o alice-in-wonderland.txt https://raw.githubusercontent.com/GITenberg/Alice-s-Adventures-in-Wonderland_11/master/11.txt \
+  -o sherlock-holmes.txt https://raw.githubusercontent.com/GITenberg/The-Memoirs-of-Sherlock-Holmes_834/master/834.txt
 
 zg index --embedding local/potion-retrieval-32m
 ```
 
-这会创建一个包含四本公版小说的小书架，并使用 Potion Retrieval 在本地建立索引。
-
-安装时会立即启动共享 Daemon。默认 MCP 传输方式为 stdio：Agent 会运行
-`zg server --stdio`，电脑重启后该命令会再次安全启动或复用 Daemon。高级用户
-可以通过 `zg install --mcp-transport http` 保留 HTTP 直连，并自行使用
-`zg server on` 管理后续启动。
-
-如需让安装后的 MCP 集成额外暴露 managed rg、index 和 status 工具，可使用
-`zg install --mcp-toolset full`；默认值为 `agent`。
-
 ### 2. 选择检索方式
 
 #### Agent：通过 OpenCode 提问
 
+配置好 [OpenCode](https://opencode.ai/) 后：
+
 ```bash
+zg install --target opencode --yes
 opencode run --model opencode/deepseek-v4-flash-free \
   "An unseen creature left a few marks. What did the detective infer? Cite local evidence."
 ```
 
-Prompt 中没有指定任何工具，OpenCode 会自主选择 zg。不同模型改写的
-Query 可能略有不同，一次典型的执行过程如下：
+Prompt 中没有指定任何工具，OpenCode 会自主选择 zg。
+
+<details>
+<summary><strong>展开查看完整 Agent 调用与回答</strong></summary>
 
 ```text
 ⚙ zvec_grep_zvec_grep_search {"root":".../zg-mystery","query":"An unseen creature left a few marks. What did the detective infer?","fts":["marks","unseen creature","inferred","detective"],"fuse":true}
 → Read sherlock-holmes.txt [offset=5440, limit=120]
 
 Answer:
-侦探是《驼背人》中的福尔摩斯。根据动物留下的脚印，他推断：
-
-- 它不是狗、猫、猴子或其他常见动物：脚印有五个肉垫和长爪，整个脚印
-  接近一把甜点勺大小（sherlock-holmes.txt:5479-5486）。
-- 它身体细长、四肢很短：身长接近两英尺，步幅却只有约三英寸
-  （sherlock-holmes.txt:5497-5502）。
-- 它是食肉动物，因为它爬上窗帘，试图接近金丝雀笼
-  （sherlock-holmes.txt:5509-5510）。
-- 它很可能属于鼬或白鼬一类，但体型异常大
-  （sherlock-holmes.txt:5514-5516）。
-
-福尔摩斯还推断，一名男子和它一起从窗户进入房间，离开时带走了钥匙
-（sherlock-holmes.txt:5464-5470, 5527-5528）。
+The detective is Sherlock Holmes in *The Adventure of the Crooked Man*.
+From the animal's foot-marks, he inferred:
+
+- It was not a dog, cat, monkey, or other familiar creature: its print had
+  five foot-pads, long nails, and was nearly as large as a dessert-spoon
+  (sherlock-holmes.txt:5479-5486).
+- It had a long body and very short legs: nearly two feet long, but with a
+  stride of only about three inches (sherlock-holmes.txt:5497-5502).
+- It was carnivorous because it climbed the curtain toward a canary's cage
+  (sherlock-holmes.txt:5509-5510).
+- It was probably from the weasel or stoat tribe, though unusually large
+  (sherlock-holmes.txt:5514-5516).
+
+Holmes also inferred that a man entered through the window with it and later
+left with the key (sherlock-holmes.txt:5464-5470, 5527-5528).
 ```
 
+</details>
+
 #### 用户：直接检索
 
 不通过 Agent，直接搜索同一个书架：
@@ -133,52 +121,27 @@ Answer:
 zg query --human "An unseen creature left a few marks. What did the detective infer?" --limit 3
 ```
 
-zg 会从四本小说中对相关段落进行排序，并将 `sherlock-holmes.txt` 中的证据排在前面。
+zg 会将 `sherlock-holmes.txt` 中的相关段落排在
+`alice-in-wonderland.txt` 前面。
 
 <a id="benchmarks"></a>
 
 ## 📊 性能测试
 
-zg 是通用检索工具，目前的性能测试覆盖代码与文本检索。
-
-### 代码仓库检索
-
-我们使用 [SWE-QA-Bench](https://github.com/peng-weihan/SWE-QA-Bench)，它要求
-Agent 在真实代码库中完成跨文件、多跳推理。
-
-- **覆盖范围：**20 个检索密集型任务，覆盖论文中的 **What、Where、How、Why**
-  四个一级类别、8 种意图和 11 个代码仓库。
-- **Agent：**使用 **Claude Opus 5** 的 Claude Code。
-- **Embedding：**zg 方案使用 **Qwen3.7 Text Embedding**。
-- **测试方式：**每个任务、每种方案运行 3 次。
-
-| 方案 | 评审得分 /100 ↑ | 输入 Token ↓ | 工具调用 ↓ | 耗时（秒）↓ | 成本 ↓ |
-| --- | ---: | ---: | ---: | ---: | ---: |
-| 原生工具 | 80.42 | 558,651 | 23.42 | 127.5 | $0.905 |
-| 原生工具 + zg | 81.92 | 294,262 | 9.70 | 79.7 | $0.558 |
-| 变化 | **+1.50 个百分点** | **−47.3%** | **−58.6%** | **−37.5%** | **−38.3%** |
-
-在 20 个任务 × 3 次运行中，zg 将评审得分提升了 1.50 分，同时将输入 Token
-减少 47.3%、工具调用减少 58.6%、执行时间减少 37.5%、成本减少 38.3%。
+zg 通过**缩小有效搜索空间**，帮助 Agent 更快找到相关证据 —— 在保持回答质量的同时，使用**更少的 Token、工具调用和时间**。
 
-### 大规模文本检索
+每项测试均采用**受控、可复现的配对 A/B 评测**：同一个 Agent 在完全相同的模型、Prompt、环境和资源限制下执行同一组预先固定的任务；**实验组仅额外提供 `zg` 工具及其使用说明**。
 
-我们使用 [BrowseComp-Plus](https://github.com/texttron/BrowseComp-Plus)，它要求
-Agent 从固定网页语料库中检索多文档证据。测试包含 30 个问题，覆盖 10 个映射主题
-和三个证据广度级别，每种方案运行 3 次。
+<img src="./.github/assets/benchmark-comparison-cn-v2.svg" alt="SWE-QA-Bench 和 BrowseComp-Plus 中 Baseline 与 zvec-grep 的配对对比" width="900" />
 
-| 方案 | 准确率 ↑ | 证据召回率 ↑ | 搜索调用 ↓ | 输入 Token ↓ | 耗时（秒）↓ | 成本 ↓ |
-| --- | ---: | ---: | ---: | ---: | ---: | ---: |
-| Agent + BM25 | TBD | TBD | TBD | TBD | TBD | TBD |
-| Agent + zg | TBD | TBD | TBD | TBD | TBD | TBD |
-| 变化 | TBD | TBD | TBD | TBD | TBD | TBD |
+| 测试 | 回答质量&nbsp;↑ | 平均输入&nbsp;Token&nbsp;↓ | 平均工具调用&nbsp;↓ | 平均&nbsp;Agent&nbsp;耗时&nbsp;↓ |
+| --- | ---: | ---: | ---: | ---: |
+| [**SWE-QA-Bench**](./benchmarks/swe-qa-bench/README_CN.md)<br>20 个任务 · 代码库 QA | 评审得分<br>80.42&nbsp;→&nbsp;81.92 | 558,651&nbsp;→&nbsp;294,262 | 23.42&nbsp;→&nbsp;9.70 | 127.5s&nbsp;→&nbsp;79.7s |
+| [**BrowseComp-Plus**](./benchmarks/browse-comp-plus/README_CN.md)<br>80 个样例 · 深度研究 QA | 准确率<br>90.00%&nbsp;→&nbsp;90.00% | 2.04M&nbsp;→&nbsp;1.19M | 22.70&nbsp;→&nbsp;14.24 | 284.8s&nbsp;→&nbsp;199.3s |
 
-在 30 个问题 × 3 次运行中，zg 对准确率、证据召回率、搜索调用、输入 Token、
-执行时间和成本的影响尚待测试结果补充。
+表中数值均按 **Baseline → zg** 展示。
 
-如需复现测试，请按照[性能测试指南](./benchmarks/README.md)准备环境、固定版本、
-任务清单、基线配置、运行命令和结果汇总。所有方案使用相同的 Agent、模型、Prompt、
-工具预算和三次运行协议。
+完整结果和复现细节参见[性能测试文档](./benchmarks/README_CN.md)。
 
 ## 📚 文档
 

package/dist/daemon/backend.d.ts

@@ -57,6 +57,7 @@ export declare class DaemonBackend implements ZvecGrepDaemonBackend {
     rg(input: ZvecGrepRgInput): Promise<ZvecGrepRgResult>;
     close(): Promise<void>;
     private runIndex;
+    private runIndexOperation;
     private submitIndex;
     private ensureWatcher;
     private settleKnownChanges;

package/dist/daemon/backend.js

@@ -203,40 +203,26 @@ export class DaemonBackend {
         const targetRevision = createsWork
             ? runtime.markDirty()
             : runtime.snapshot().dirtyRevision;
-        runtime.setWriterPending(true);
-        let submitted;
-        try {
-            submitted = this.scheduler.submit({
-                canonicalRoot: runtime.canonicalRoot,
-                reason: "manual",
-                followupIfRunning: input.rebuild === true || followsNarrowJob,
-                run: (report, signal) => runtime.withWrite(async () => {
-                    const proof = await this.runIndex(runtime, input, report, options.authorization, signal);
-                    this.lastScanDiagnostics.set(runtime.canonicalRoot, proof.scanDiagnostics);
-                    if (proof.reconciled) {
-                        runtime.markReconciled(targetRevision, proof.reconciliationEpoch);
-                    }
-                    else {
-                        runtime.markIndexed(targetRevision);
-                    }
-                }),
-            });
-        }
-        catch (error) {
-            runtime.setWriterPending(false);
-            throw error;
-        }
-        if (!submitted.reused) {
-            void this.scheduler.waitForRootIdle(runtime.canonicalRoot).finally(() => {
-                runtime.setWriterPending(false);
-            });
-        }
+        const submitted = this.scheduler.submit({
+            canonicalRoot: runtime.canonicalRoot,
+            reason: "manual",
+            followupIfRunning: input.rebuild === true || followsNarrowJob,
+            run: async (report, signal) => {
+                const proof = await this.runIndex(runtime, input, report, options.authorization, signal);
+                this.lastScanDiagnostics.set(runtime.canonicalRoot, proof.scanDiagnostics);
+                if (proof.reconciled) {
+                    runtime.markReconciled(targetRevision, proof.reconciliationEpoch);
+                }
+                else {
+                    runtime.markIndexed(targetRevision);
+                }
+            },
+        });
         const job = input.wait
             ? await this.scheduler.wait(submitted.job.id, options.onProgress)
             : submitted.job;
         if (input.wait) {
             await this.settleKnownChanges(runtime);
-            runtime.setWriterPending(false);
         }
         this.options.logger?.event("job.submitted", {
             root_id: rootIdentity(runtime.canonicalRoot),
@@ -378,10 +364,6 @@ export class DaemonBackend {
                 input.autoUpdate &&
                 (runtime.requiresFullReconciliation() ||
                     !this.scheduler.hasActiveRoot(runtime.canonicalRoot))) {
-                if (!runtime.requiresFullReconciliation() ||
-                    runtime.canProbeFullReconciliation()) {
-                    await this.probeCurrentFreshness(runtime);
-                }
                 const currentJob = this.scheduler.getByRoot(runtime.canonicalRoot);
                 const terminalKnownPathJob = !runtime.requiresFullReconciliation() &&
                     (currentJob?.state === "failed" || currentJob?.state === "cancelled");
@@ -535,14 +517,25 @@ export class DaemonBackend {
         return this.closePromise;
     }
     async runIndex(runtime, input, report, authorization, signal) {
+        const releaseRuntimeActivity = runtime.beginActivity();
+        try {
+            return await this.runIndexOperation(runtime, input, report, authorization, signal);
+        }
+        finally {
+            releaseRuntimeActivity();
+        }
+    }
+    async runIndexOperation(runtime, input, report, authorization, signal) {
         const startedAt = Date.now();
-        // Watcher path updates already identify their indexing scope, so avoid a
-        // workspace-wide status scan here. Silent watcher misses are repaired by
-        // the periodic full-reconciliation probes scheduled by WatchManager.
-        const includeInitialStatus = input.rebuild !== true && input.changedPaths === undefined;
+        // Watcher path updates already identify their indexing scope. Uncertain
+        // watcher reconciliations are verified by the final status scan, so neither
+        // path needs another workspace-wide status scan before indexing starts.
+        const includeInitialStatus = input.skipInitialStatus !== true &&
+            input.rebuild !== true &&
+            input.changedPaths === undefined;
         const includeFinalStatus = input.changedPaths === undefined;
         const before = await this.inspectRoot(runtime.canonicalRoot, includeInitialStatus);
-        if (input.changedPaths === undefined) {
+        if (includeInitialStatus) {
             this.statusCache.set(runtime.canonicalRoot, before);
         }
         const modelLoadRequest = this.indexModelLoadRequest(before, input);
@@ -573,7 +566,7 @@ export class DaemonBackend {
                 device: input.runtimeOverridesAreEphemeral ? undefined : input.device,
                 daemonInstanceToken: this.runtimeManager.instanceToken,
             });
-            const result = await withRemoteEmbeddingOperationPermit(authorization, () => service.index({
+            const result = await runtime.withWrite(() => withRemoteEmbeddingOperationPermit(authorization, () => service.index({
                 root: runtime.canonicalRoot,
                 rebuild: input.rebuild,
                 resetPaths: input.resetPaths,
@@ -592,7 +585,7 @@ export class DaemonBackend {
                 signal,
                 onProgress: report,
                 onWriterContext: (context) => runtime.setWriterContext(context, lease.key),
-            }));
+            })));
             scanDiagnostics = result.scanDiagnostics;
         }
         finally {
@@ -608,8 +601,10 @@ export class DaemonBackend {
         }
         const proofReconciliationEpoch = runtime.reconciliationEpoch();
         const after = await this.inspectRoot(runtime.canonicalRoot, includeFinalStatus);
-        if (includeFinalStatus)
+        if (includeFinalStatus) {
             this.statusCache.set(runtime.canonicalRoot, after);
+            await this.watchers.get(runtime.canonicalRoot)?.refreshPaths?.();
+        }
         if (!after.workspaceIndex?.embedding) {
             throw new DaemonError("INDEX_MISSING", "Index completed without an embedding schema.");
         }
@@ -637,38 +632,29 @@ export class DaemonBackend {
             ? runtime.markDirty()
             : runtime.snapshot().dirtyRevision;
         const followsNarrowWatch = this.scheduler.getByRoot(runtime.canonicalRoot)?.reason === "watch";
-        runtime.setWriterPending(true);
-        let submitted;
-        try {
-            submitted = this.scheduler.submit({
-                canonicalRoot: runtime.canonicalRoot,
-                reason,
-                followupIfRunning: followsNarrowWatch,
-                run: (report, signal) => runtime.withWrite(async () => {
-                    const proof = await this.runIndex(runtime, input, report, authorization, signal);
-                    if (proof.reconciled) {
-                        runtime.markReconciled(targetRevision, proof.reconciliationEpoch);
-                    }
-                    else {
-                        runtime.markIndexed(targetRevision);
-                    }
-                }),
-            });
-        }
-        catch (error) {
-            runtime.setWriterPending(false);
-            throw error;
-        }
-        if (!submitted.reused) {
-            void this.scheduler.waitForRootIdle(runtime.canonicalRoot).finally(() => {
-                runtime.setWriterPending(false);
-            });
-        }
+        const submitted = this.scheduler.submit({
+            canonicalRoot: runtime.canonicalRoot,
+            reason,
+            followupIfRunning: followsNarrowWatch,
+            run: async (report, signal) => {
+                if (reason === "background_reconcile" &&
+                    runtime.canProbeFullReconciliation() &&
+                    (await this.probeCurrentFreshness(runtime)) === "fresh") {
+                    return;
+                }
+                const proof = await this.runIndex(runtime, input, report, authorization, signal);
+                if (proof.reconciled) {
+                    runtime.markReconciled(targetRevision, proof.reconciliationEpoch);
+                }
+                else {
+                    runtime.markIndexed(targetRevision);
+                }
+            },
+        });
         if (!wait)
             return submitted.job;
         const job = await this.scheduler.wait(submitted.job.id);
         await this.scheduler.waitForRootIdle(runtime.canonicalRoot);
-        runtime.setWriterPending(false);
         return job;
     }
     ensureWatcher(runtime) {
@@ -678,7 +664,6 @@ export class DaemonBackend {
         const coordinator = new IndexCoordinator({
             runtime,
             scheduler: this.scheduler,
-            getIndexedFileCount: () => this.statusCache.get(runtime.canonicalRoot)?.status?.filesStored,
             run: async (changes, report, signal) => {
                 const changedPaths = [
                     ...changes.touchedFiles,
@@ -695,6 +680,7 @@ export class DaemonBackend {
                 return await this.runIndex(runtime, {
                     root: runtime.canonicalRoot,
                     changedPaths: changes.forceFullReconcile ? undefined : changedPaths,
+                    skipInitialStatus: changes.forceFullReconcile,
                 }, report, automaticAuthorization.authorization, signal);
             },
         });
@@ -734,6 +720,7 @@ export class DaemonBackend {
                 coordinator.enqueue(changes, reason);
             },
             onPendingChange: (pending) => runtime.setWatcherPending(pending),
+            getRootPaths: () => this.statusCache.get(runtime.canonicalRoot)?.workspaceIndex?.rootPaths,
         });
         watcher.start();
         this.indexCoordinators.set(runtime.canonicalRoot, coordinator);
@@ -821,15 +808,21 @@ export class DaemonBackend {
         }
     }
     async probeCurrentFreshness(runtime) {
-        const freshness = await runtime.probeFreshness(async () => {
-            const info = await inspectRoot(runtime.canonicalRoot, this.options.serviceOptions);
-            this.statusCache.set(runtime.canonicalRoot, info);
-            return indexStatusIsFresh(info);
-        });
-        this.options.logger?.event(`runtime.recovery_probe_${freshness}`, {
-            root_id: rootIdentity(runtime.canonicalRoot),
-        });
-        return freshness;
+        const releaseRuntimeActivity = runtime.beginActivity();
+        try {
+            const freshness = await runtime.probeFreshness(async () => {
+                const info = await this.inspectRoot(runtime.canonicalRoot, true);
+                this.statusCache.set(runtime.canonicalRoot, info);
+                return indexStatusIsFresh(info);
+            });
+            this.options.logger?.event(`runtime.recovery_probe_${freshness}`, {
+                root_id: rootIdentity(runtime.canonicalRoot),
+            });
+            return freshness;
+        }
+        finally {
+            releaseRuntimeActivity();
+        }
     }
     async closeWatcher(canonicalRoot) {
         const watcher = this.watchers.get(canonicalRoot);