SylphxAI · sylphx-desk-services · Oct 3, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,6 +1,6 @@
 # Changelog
 
-## Unreleased
+## 0.7.0
 
 - **First use finds the docs sites too.** The automatic first-use fetch (`docs`/`api` before any `lockdocs fetch`) skipped the separate docs-site repositories, so React 19, Express, tokio and Tailwind answers came from package files only. It now reads them as well: the site archive within the same 64 MB cap, else one REST tree request plus raw files (tailwindcss.com is 190 MB and prisma/docs over 800 MB as archives, but a few hundred files). The REST call follows a redirect to the same API host once, which a renamed repository such as prisma/docs answers. A docs-site failure keeps the package docs and says what was skipped in the manifest note; `lockdocs fetch` retries it. On the benchmark's first-use run (empty isolated cache, no credentials, 97 of 105 questions installable locally) 76 become 88 correct and held-out 13/16 becomes 14/16; the first-use floor in `bench/run.py` is 88/105. `upstream::FORMAT` is 6.
 - **Free upgrade preview.** Without a Pro licence, `lockdocs upgrade <pkg> <ver>` (and `docs --upgrade-to`, and the `docs` tool's `upgrade_to`) now answers with a short local preview instead of only "Pro required": how many of the APIs your project calls were removed, renamed, re-signed or deprecated, across how many call sites, plus one sample, then the Pro price and link. The full list stays Pro. The MCP answer still carries `structuredContent.pro_required`, the CLI still exits 3, and the preview reads the same installed, cached or opted-in files as the report (no new downloads).

diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -3,7 +3,7 @@ members = ["crates/lockdocs-core", "crates/lockdocs"]
 resolver = "2"
 
 [workspace.package]
-version = "0.6.0"
+version = "0.7.0"
 edition = "2021"
 license = "MIT"
 authors = ["SylphxAI"]

diff --git a/README.md b/README.md
@@ -186,6 +186,7 @@ Registry package downloads remain opt-in; default upstream enrichment uses insta
 | Context7 (anonymous) | 77/105 | 19/45 | 53/55 | 5/5 | 15/18 | 908 | 2,583 ms |
 <!-- bench:end -->
 
+- **Default first use is 88/105 in 0.7.0** (local first-use run; the table's 70/105 row is the earlier 0.4.0 run and is replaced by the next full benchmark run), above Context7's 77/105.
 - **Where versions matter most, lockdocs wins by 2x** (38/45 vs 19/45 on older majors). Context7 often answers with the newest API (pydantic 1 questions get pydantic 2 answers).
 - **Tied on the newest majors (53/55 each) and on tokio (5/5 each); ahead on held-out questions (16/18 vs 15/18)**, which were written before the ranking changes they measure and not used for tuning.
 - **Faster, fewer tokens, no quota.** Median 58 ms per question after `lockdocs fetch` (185 ms on first use), a fresh CLI process each time. Context7's hosted API took a median 2,583 ms in the earlier runs its reused answers come from.

diff --git a/crates/lockdocs/Cargo.toml b/crates/lockdocs/Cargo.toml
@@ -20,7 +20,7 @@ path = "src/main.rs"
 
 [dependencies]
 anyhow = "1"
-lockdocs-core = { path = "../lockdocs-core", version = "0.6.0" }
+lockdocs-core = { path = "../lockdocs-core", version = "0.7.0" }
 serde_json = "1"
 sylphx-mcp-kit = { version = "0.7.0", features = ["licence"] }
 

diff --git a/docs/benchmarks.md b/docs/benchmarks.md
@@ -32,7 +32,7 @@ The per-question table, index build times and `lockdocs fetch` times are in the
 ## Reading the results
 
 - **lockdocs is ahead overall (96/105 vs 77/105) and on older majors by 2x (38/45 vs 19/45), ties Context7 on the newest majors (53/55 each) and on tokio (5/5 each), and is ahead on the held-out questions (16/18 vs 15/18).** It also uses fewer tokens (880 vs 908 median) and answers in 58 ms instead of 2,583 ms (Context7 latency from the earlier runs its reused answers were measured in).
-- **Default first use: 70/105.** From an empty cache with no GitHub token, the first query downloads release-tag docs anonymously; median 185 ms including those downloads (p95 2258 ms). Running `lockdocs fetch` once raises it to 96/105.
+- **Default first use: 88/105 in 0.7.0 (70/105 in the run above, which predates it).** From an empty cache with no GitHub token, the first query downloads release-tag docs anonymously, and since 0.7.0 also the docs-site repositories (React, Express, tokio, Tailwind, Prisma); the 0.7.0 figure is a local first-use run (97 of 105 questions installable on that machine, held-out 14/16), and the table above is replaced by the next full benchmark run. The 0.4.0 run's median was 185 ms including downloads (p95 2258 ms). Running `lockdocs fetch` once raises it to 96/105.
 - **Where each still misses.** lockdocs: seven older-major questions (among them Next.js 14 `cookies()`, React Router 6 `json()`, FastAPI 0.88 `on_event`, Prisma 5 `Buffer`, ESLint 8 `env`), and two held-out newer-major questions (ESLint 9 plugins, Tailwind 4 `@source`). Context7 mostly answers older-major questions with the newest API, and two of its Tailwind answers came from unrelated libraries its search ranked first.
 - **Configurations matter (from run 36204404336).** Keyword-only on package files: 59/105. Adding the local embedding model: 60/105. Adding `lockdocs fetch` (upstream docs at each version's git tag, plus docs-site repositories for your major): 96/105.
 - Latency for lockdocs is a fresh CLI process per question, including loading the embedding model; the MCP server keeps it loaded.

diff --git a/docs/index.md b/docs/index.md
@@ -24,7 +24,7 @@ proof:
   - value: "96/105"
     label: version-sensitive questions right after lockdocs fetch (Context7 77/105)
     link: /benchmarks
-  - value: "70/105"
+  - value: "88/105"
     label: with default settings from an empty cache
     link: /benchmarks
   - value: "58 ms"

diff --git a/package.json b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "lockdocs-workspace",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "private": true,
   "type": "module",
   "description": "lockdocs monorepo: Rust engine (crates/), npm packages (packages/), docs (docs/), benchmark (bench/).",

diff --git a/packages/lockdocs/package.json b/packages/lockdocs/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/lockdocs",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "mcpName": "io.github.SylphxAI/lockdocs",
   "description": "Exact-version library docs from your lockfile — local, offline, no rate limits.",
   "bin": {
@@ -15,11 +15,11 @@
     "node": ">=18"
   },
   "optionalDependencies": {
-    "@sylphx/lockdocs-darwin-arm64": "0.6.0",
-    "@sylphx/lockdocs-darwin-x64": "0.6.0",
-    "@sylphx/lockdocs-linux-x64-gnu": "0.6.0",
-    "@sylphx/lockdocs-linux-arm64-gnu": "0.6.0",
-    "@sylphx/lockdocs-win32-x64-msvc": "0.6.0"
+    "@sylphx/lockdocs-darwin-arm64": "0.7.0",
+    "@sylphx/lockdocs-darwin-x64": "0.7.0",
+    "@sylphx/lockdocs-linux-x64-gnu": "0.7.0",
+    "@sylphx/lockdocs-linux-arm64-gnu": "0.7.0",
+    "@sylphx/lockdocs-win32-x64-msvc": "0.7.0"
   },
   "keywords": [
     "mcp",

diff --git a/packages/npm/darwin-arm64/package.json b/packages/npm/darwin-arm64/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/lockdocs-darwin-arm64",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "description": "lockdocs native binary for darwin-arm64",
   "os": [
     "darwin"

diff --git a/packages/npm/darwin-x64/package.json b/packages/npm/darwin-x64/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/lockdocs-darwin-x64",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "description": "lockdocs native binary for darwin-x64",
   "os": [
     "darwin"

diff --git a/packages/npm/linux-arm64-gnu/package.json b/packages/npm/linux-arm64-gnu/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/lockdocs-linux-arm64-gnu",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "description": "lockdocs native binary for linux-arm64-gnu",
   "os": [
     "linux"

diff --git a/packages/npm/linux-x64-gnu/package.json b/packages/npm/linux-x64-gnu/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/lockdocs-linux-x64-gnu",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "description": "lockdocs native binary for linux-x64-gnu",
   "os": [
     "linux"

diff --git a/packages/npm/win32-x64-msvc/package.json b/packages/npm/win32-x64-msvc/package.json
@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/lockdocs-win32-x64-msvc",
-  "version": "0.6.0",
+  "version": "0.7.0",
   "description": "lockdocs native binary for win32-x64-msvc",
   "os": [
     "win32"

diff --git a/server.json b/server.json
@@ -7,13 +7,13 @@
     "url": "https://github.com/SylphxAI/lockdocs",
     "source": "github"
   },
-  "version": "0.6.0",
+  "version": "0.7.0",
   "websiteUrl": "https://sylphxai.github.io/lockdocs/",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "@sylphx/lockdocs",
-      "version": "0.6.0",
+      "version": "0.7.0",
       "transport": {
         "type": "stdio"
       },