[codex] Fix LSP live sync and ctags export summaries

DEVELOPER_GUIDE.md

@@ -236,6 +236,12 @@ When an error code is available, the first line is `Error [<code>]: <message>`.
 
 CLI JSON output must be machine-clean: redirected stdout is written as UTF-8 without a BOM, and JSON-mode commands must not emit ANSI escape sequences even when `--color=always` or `CLICOLOR_FORCE=1` would color human output. Keep JSON-safe styling suppression close to shared formatting helpers such as `ConsoleUi.ColorizeKind` so future query output paths inherit the invariant.
 
+`cdidx export ctags --json` follows the same contract: stdout contains only a
+single JSON summary or structured error, while the tag file itself remains the
+artifact. The summary includes resolved output/database paths, tag/emitted/
+skipped counts, filters, and advertised metadata field names so editor
+integrations can validate filtered exports without parsing human output.
+
 Interactive terminal controls are allowed only when stdout is not redirected or captured, terminal capability hints are present, and the environment has not opted out. Treat `TERM=dumb`, truthy `CI`, missing Unix terminal hints, `NO_COLOR`, and `CLICOLOR=0` as reasons to suppress ANSI/progress controls unless an explicitly human-facing override is documented for that control.
 
 ### C# / .NET integration
@@ -248,6 +254,13 @@ Query commands that accept path filters (`search`, `definition`, `references`, `
 
 Editor integrations can request standard location shapes directly. `definition`, `references`, `search`, `find`, and `validate` accept `--format <text|json|lsp|qf|sarif>`; `lsp` emits LSP `Location` arrays, `qf` emits Vim quickfix lines, and `sarif` emits SARIF 2.1.0. `goto <symbol>` returns the single unambiguous definition as one LSP `Location`, while `goto --all <symbol>` returns all matching locations.
 
+The `cdidx lsp` server advertises full text document synchronization and keeps
+open document text in a bounded in-memory cache only. Position-based providers
+must read that live cache before disk so unsaved editor buffers can identify the
+requested token, but provider results remain conservative and index-backed:
+return empty arrays or null when the database cannot answer safely instead of
+inventing language-server analysis.
+
 ### Extractor performance contract
 
 Symbol and reference extractors run during `cdidx index`, so language-specific
@@ -2463,6 +2476,11 @@ CLI JSON output は機械処理向けにきれいでなければなりません
 `ConsoleUi.ColorizeKind` など共有 formatter の近くに置き、将来の query output path も同じ
 invariant を継承できるようにしてください。
 
+`cdidx export ctags --json` も同じ contract に従います。stdout は単一の JSON summary または
+structured error だけを含み、tags file 自体は artifact として残します。summary には解決済みの
+output / database path、tag / emitted / skipped counts、filters、metadata field names を含め、
+editor integration が human output を parse せず filtered export を検証できるようにします。
+
 interactive terminal control は stdout が redirected / captured されておらず、terminal
 capability hint があり、environment が opt out していない場合にだけ許可します。`TERM=dumb`、
 truthy `CI`、Unix terminal hint の欠落、`NO_COLOR`、`CLICOLOR=0` は、明示的な human-facing
@@ -2478,6 +2496,12 @@ path filter を受け付ける query コマンド（`search`, `definition`, `ref
 
 editor integration は標準的な location 形状を直接要求できる。`definition`、`references`、`search`、`find`、`validate` は `--format <text|json|lsp|qf|sarif>` を受け付け、`lsp` は LSP `Location` 配列、`qf` は Vim quickfix 行、`sarif` は SARIF 2.1.0 を出力する。`goto <symbol>` は曖昧でない単一定義を 1 つの LSP `Location` として返し、`goto --all <symbol>` は一致する全 location を返す。
 
+`cdidx lsp` server は full text document synchronization を advertise し、open document text は
+上限付きの in-memory cache にだけ保持する。position-based provider は未保存 editor buffer から
+request token を特定できるよう disk より先に live cache を読む必要があるが、provider result は
+保守的かつ index-backed のままにする。database が安全に答えられない場合は、language-server
+analysis を作り上げず、空配列または null を返す。
+
 ### 抽出器の性能契約
 
 symbol / reference extractor は `cdidx index` 中に実行されるため、言語別 helper は

USER_GUIDE.md

@@ -350,8 +350,19 @@ instead of querying `cdidx` directly:
 ```bash
 cdidx export ctags --output tags
 cdidx export ctags --db .cdidx/codeindex.db --output .tags
+cdidx export ctags --lang csharp --path src/ --exclude-tests --json
 ```
 
+`cdidx export ctags` accepts the same language and path filtering style used by
+query commands: `--lang <lang>`, repeatable `--path <glob>`, repeatable
+`--exclude-path <glob>`, and `--exclude-tests`. The default human mode keeps
+writing the tags file and prints the output path. `--json` prints a machine
+summary with `output_path`, `db_path`, total candidate `tag_count`,
+`emitted_count`, `skipped_count`, `filters`, and `metadata_fields`; filtered
+exports satisfy `tag_count == emitted_count + skipped_count`. Tag lines keep
+the standard `kind` and `line` fields and may also include indexed metadata such as
+`language`, `container_kind`, `container`, and `visibility`.
+
 Use `cdidx export <archive>` to package the current `codeindex.db` with a
 manifest, and `cdidx import <archive>` to restore it on another checkout or CI
 job:
@@ -2105,7 +2116,14 @@ server over stdio. It reuses the existing CodeIndex database and exposes
 `textDocument/definition`, `textDocument/declaration`,
 `textDocument/typeDefinition`, `textDocument/implementation`, and
 `textDocument/references` for editors that can launch an arbitrary LSP command
-but do not speak MCP.
+but do not speak MCP. It also advertises full `textDocument` sync and
+conservative `hover`, `completion`, `documentHighlight`, `semanticTokens/full`,
+`codeLens`, and `inlayHint` providers backed by indexed symbols and references
+where available.
+Open buffers sent through `textDocument/didOpen`, `textDocument/didChange`, and
+`textDocument/didClose` are kept in a bounded in-memory cache. Position-based
+requests read the live buffer first, so unsaved edits can drive token lookup
+without writing back to the CodeIndex database.
 Incoming `textDocument.uri` values must be strings, must be absolute `file:`
 URIs, and are rejected before URI parsing when they exceed 4096 characters,
 matching the MCP resource URI limit and keeping error responses bounded. LSP
@@ -2805,8 +2823,19 @@ Editor が `cdidx` を直接 query するのではなく従来の ctags file を
 ```bash
 cdidx export ctags --output tags
 cdidx export ctags --db .cdidx/codeindex.db --output .tags
+cdidx export ctags --lang csharp --path src/ --exclude-tests --json
 ```
 
+`cdidx export ctags` は query command と同じ language / path filter の形を受け付けます。
+`--lang <lang>`、繰り返し指定できる `--path <glob>` / `--exclude-path <glob>`、
+`--exclude-tests` を使えます。既定の human mode は tags file を書き出し、output path を
+表示します。`--json` は `output_path`、`db_path`、総候補数の `tag_count`、
+`emitted_count`、`skipped_count`、`filters`、`metadata_fields` を含む機械処理向け
+summary を出力します。filter 付き export では
+`tag_count == emitted_count + skipped_count` になります。tag line は標準の
+`kind` / `line` fields を維持し、indexed metadata として
+`language`、`container_kind`、`container`、`visibility` も含めることがあります。
+
 `cdidx export <archive>` は現在の `codeindex.db` と manifest を archive 化します。
 別 checkout や CI job では `cdidx import <archive>` で復元できます。
 
@@ -4560,7 +4589,14 @@ cdidxには**MCP（Model Context Protocol）サーバー**が組み込まれて
 `initialize`、`workspace/symbol`、`textDocument/documentSymbol`、
 `textDocument/definition`、`textDocument/declaration`、
 `textDocument/typeDefinition`、`textDocument/implementation`、
-`textDocument/references` を公開します。
+`textDocument/references` を公開します。さらに full `textDocument` sync と、
+indexed symbols / references で答えられる範囲に限定した `hover`、`completion`、
+`documentHighlight`、`semanticTokens/full`、`codeLens`、`inlayHint` provider を
+advertise します。
+`textDocument/didOpen`、`textDocument/didChange`、`textDocument/didClose` で送られた
+open buffer は上限付きの in-memory cache に保持されます。position-based request は
+live buffer を先に読むため、未保存の編集内容でも CodeIndex database に書き戻さず token lookup に
+利用できます。
 受信した `textDocument.uri` は string かつ absolute `file:` URI である必要があり、
 4096 文字を超える場合は URI parse の前に拒否されます。これは MCP resource URI の上限と
 揃えており、エラー応答が過大にならないようにします。

changelog.d/unreleased/3536.added.md

@@ -0,0 +1,18 @@
+---
+category: added
+issues:
+  - 3536
+affected:
+  - src/CodeIndex/Lsp/LspServer.cs
+  - tests/CodeIndex.Tests/LspServerTests.cs
+  - USER_GUIDE.md
+  - DEVELOPER_GUIDE.md
+---
+
+## English
+
+- **LSP now supports live document sync and richer editor providers (#3536)** — `cdidx lsp` now advertises full text document sync plus conservative hover, completion, document highlight, semantic tokens, code lens, and inlay hint providers backed by indexed symbols and references where available.
+
+## 日本語
+
+- **LSP が live document sync と richer editor providers に対応しました (#3536)** — `cdidx lsp` は full text document sync と、indexed symbols / references で答えられる範囲に限定した hover、completion、document highlight、semantic tokens、code lens、inlay hint providers を advertise するようになりました。

changelog.d/unreleased/3551.fixed.md

@@ -0,0 +1,20 @@
+---
+category: fixed
+issues:
+  - 3551
+affected:
+  - src/CodeIndex/Cli/ExportImportCommandRunner.cs
+  - src/CodeIndex/Cli/JsonOutputContracts.cs
+  - src/CodeIndex/Cli/ConsoleUi.cs
+  - tests/CodeIndex.Tests/ExportImportCommandRunnerTests.cs
+  - USER_GUIDE.md
+  - DEVELOPER_GUIDE.md
+---
+
+## English
+
+- **ctags export now reports filtered JSON summaries and richer metadata (#3551)** — `cdidx export ctags` now supports `--json`, `--lang`, repeated `--path` / `--exclude-path`, and `--exclude-tests`, while emitted tags include indexed metadata fields such as language, container, and visibility when available.
+
+## 日本語
+
+- **ctags export が filtered JSON summary と richer metadata を出力できるようになりました (#3551)** — `cdidx export ctags` は `--json`、`--lang`、繰り返し指定できる `--path` / `--exclude-path`、`--exclude-tests` を受け付け、出力 tag には利用可能な場合に language、container、visibility などの indexed metadata fields も含めるようになりました。

src/CodeIndex/Cli/ConsoleUi.cs

@@ -116,7 +116,7 @@ private static readonly (string Command, string Usage)[] CommandUsageLines =
         ("hotspots", "cdidx hotspots [--db <path>] [--json] [--verbose] [--limit <n>|--top <n>] [--kind <kind>] [--visibility <v[,v]>] [--exclude-visibility <v[,v]>] [--lang <lang>] [--path <glob>] [--exclude-path <glob>] [--exclude-tests] [--count] [--group-by <symbol|file|statement>] [--group-by-name]"),
         ("suggestions", "cdidx suggestions <list|show|export> [id] [--db <path>] [--json] [--status <all|submitted|unsubmitted>] [--language <lang>] [--category <category>] [--since <datetime>] [--agent <name>] [--limit <n>] [--offset <n>] [--format <json|markdown|issue-drafts>] [--open-issues <path|github|github:owner/name>] [--repo <owner/name>]"),
         ("export", "cdidx export <archive> [--db <path>] [--json]"),
-        ("export", "cdidx export ctags [--output <path>] [--db <path>]"),
+        ("export", "cdidx export ctags [--output <path>] [--db <path>] [--json] [--lang <lang>] [--path <glob>] [--exclude-path <glob>] [--exclude-tests]"),
         ("import", "cdidx import <archive> [--db <path>] [--prune-paths] [--dry-run|--check] [--json]"),
         ("languages", "cdidx languages [--db <path>] [--json] [--indexed-only] [--capability <graph|references|symbols|missing-graph|missing-references|missing-symbols|search-only>]"),
         ("batch", "cdidx batch [--db <path>]  # reads JSON string arrays from stdin, one query command per line; max 1,048,576 chars/line and 256 arguments"),