From 38610ec5dd03f22a98765d83ba34e97ccbd91d21 Mon Sep 17 00:00:00 2001 From: yacosta738 <33158051+yacosta738@users.noreply.github.com> Date: Wed, 1 Jul 2026 08:49:51 +0000 Subject: [PATCH] docs: document pdf_inspect tool and sync bilingual parity --- .agents/journal/scribe-journal.md | 17 +++++++++++++++ .../clients/agent-runtime/tools/index.mdx | 2 +- .../docs/clients/agent-runtime/tools/media.md | 21 ++++++++++++++++++- .../es/clients/agent-runtime/tools/core.md | 14 +++++++++++++ .../es/clients/agent-runtime/tools/index.mdx | 2 +- .../es/clients/agent-runtime/tools/media.md | 21 ++++++++++++++++++- 6 files changed, 73 insertions(+), 4 deletions(-) diff --git a/.agents/journal/scribe-journal.md b/.agents/journal/scribe-journal.md index 3de2f6423..71d23ae75 100644 --- a/.agents/journal/scribe-journal.md +++ b/.agents/journal/scribe-journal.md @@ -91,6 +91,23 @@ All three identified issues have been fixed: - Glossary: "Firejail" = Linux user-space sandbox (confirmed implemented). "Landlock" = Linux kernel-level sandbox (confirmed). "Cerebro" = standalone MCP memory service client in `clients/cerebro/` (confirmed). - Remaining gap: `cost` command subcommands (`summary`, `history`, `reset`) exist in code but are not yet in the CLI reference doc. Low priority. +## 2026-07-01 - PDF Inspection & Parity Sync - Complete + +**Verification:** +- Verified `pdf_inspect` implementation in `clients/agent-runtime/src/tools/pdf_inspect.rs`. +- Verified `Task*` parity tools and aliases in `clients/agent-runtime/src/tools/mod.rs`. +**Changes:** +- Added `pdf_inspect` documentation to `media.md` (en/es), including security tier, features, constraints, and parameters. +- Added "Mapa de nombres de paridad" (Parity naming map) to `es/clients/agent-runtime/tools/core.md` to ensure bilingual parity with the English version. +- Updated `index.mdx` (en/es) to include `pdf_inspect` in the Read-Only (Safe) examples. +- Updated `lastReviewed` date for `media.md` (en/es) to 2026-07-01 to satisfy metadata validation rules. +**Validation:** +- `make docs-check` — ✅ PASSED (astro check + biome + metadata validation). +- `make docs-build` — ✅ PASSED (84 pages built, search index generated). +**Notes:** +- `pdf_inspect` is a native tool requiring the `pdf-inspect` feature flag. +- Parity mapping ensures developers can use Claude-style tool names (`glob`, `task_create`, etc.) which are aliased to Corvus-native implementations. + ## 2026-04-15 - CLI Reference Update - Complete **Verification:** Verified `clients/agent-runtime/src/main.rs` for `code` and `cost` command implementations. diff --git a/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/index.mdx b/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/index.mdx index cd4c5bb34..4e370f086 100644 --- a/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/index.mdx +++ b/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/index.mdx @@ -31,7 +31,7 @@ Every tool execution passes through the `SecurityPolicy` layer. Tools are classi ### Read-Only (Safe) These tools do not modify the system or external state. They are generally allowed even in low-autonomy modes. -- *Examples:* `file_read`, `Glob`, `Grep`, `WebFetch`, `memory_recall`, `web_search_tool`, `image_info`, `hardware_board_info`. +- *Examples:* `file_read`, `Glob`, `Grep`, `WebFetch`, `memory_recall`, `web_search_tool`, `image_info`, `pdf_inspect`, `hardware_board_info`. ## Claude-style parity mapping diff --git a/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/media.md b/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/media.md index fef8e2cc3..b1bf2152c 100644 --- a/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/media.md +++ b/clients/web/apps/docs/src/content/docs/clients/agent-runtime/tools/media.md @@ -3,7 +3,7 @@ title: Media Tools description: Reference for vision and image-related tools in Corvus. owner: team-runtime status: canonical -lastReviewed: 2026-03-26 +lastReviewed: 2026-07-01 appliesTo: main docType: reference --- @@ -44,3 +44,22 @@ Extracts metadata from an image file and optionally returns it as base64 for pro | :--- | :--- | :--- | | `path` | `string` | **Required.** Path to the image file. | | `include_base64` | `boolean` | Include the full image data in the output. Default: `false`. | + +--- + +## `pdf_inspect` + +Inspects, classifies, and extracts text from a PDF file. + +- **Security Tier:** Read-Only (Safe). +- **Features:** Detects PDF type (text, scanned, image, mixed), page count, and converts extractable text to Markdown. +- **Constraints:** + - Maximum file size: 50 MB. + - Timeout: 60 seconds. + +### Parameters + +| Parameter | Type | Description | +| :--- | :--- | :--- | +| `path` | `string` | **Required.** Relative path to the PDF file within the workspace. | +| `extract_text` | `boolean` | Whether to extract and convert text to Markdown (default: true). | diff --git a/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/core.md b/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/core.md index 3738f1a07..436e6170d 100644 --- a/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/core.md +++ b/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/core.md @@ -27,6 +27,20 @@ Consulta la página dedicada de [`code_search`](code-search.md) para el contrato --- +## Mapa de nombres de paridad + +| Herramienta canónica | Alias de compatibilidad | Superficie del runtime que lo respalda | Notas | +| --- | --- | --- | --- | +| `Glob` | `glob` | Helpers de discovery del workspace | Alias de paridad aditivo | +| `Grep` | `grep` | Internals de `code_search` | `code_search` se mantiene como contrato nativo | +| `TaskCreate` | `task_create` | Servicio de tareas persistentes | Solo disponible con backend SQLite | +| `TaskGet` | `task_get` | Servicio de tareas persistentes | Solo disponible con backend SQLite | +| `TaskList` | `task_list` | Servicio de tareas persistentes | Orden determinístico por created_at/id | +| `TaskUpdate` | `task_update` | Servicio de tareas persistentes | Actualiza solo campos mutables | +| `TaskStop` | `task_stop` | Servicio de tareas persistentes | Ruta de cancelación semántica | + +--- + ## `Glob` Herramienta de paridad estilo Claude para descubrimiento seguro de archivos por patrón dentro del workspace. diff --git a/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/index.mdx b/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/index.mdx index 401b0209c..f53549a06 100644 --- a/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/index.mdx +++ b/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/index.mdx @@ -31,7 +31,7 @@ Cada ejecución de herramienta pasa por la capa de `SecurityPolicy`. Las herrami ### Solo Lectura (Seguras) Estas herramientas no modifican el sistema ni el estado externo. Generalmente están permitidas incluso en modos de baja autonomía. -- *Ejemplos:* `file_read`, `Glob`, `Grep`, `WebFetch`, `memory_recall`, `web_search_tool`, `image_info`, `hardware_board_info`. +- *Ejemplos:* `file_read`, `Glob`, `Grep`, `WebFetch`, `memory_recall`, `web_search_tool`, `image_info`, `pdf_inspect`, `hardware_board_info`. ## Mapeo de paridad estilo Claude diff --git a/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/media.md b/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/media.md index e02544d04..4770e6ce6 100644 --- a/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/media.md +++ b/clients/web/apps/docs/src/content/docs/es/clients/agent-runtime/tools/media.md @@ -3,7 +3,7 @@ title: Herramientas Multimedia description: Referencia para herramientas de visión e imágenes en Corvus. owner: team-runtime status: canonical -lastReviewed: 2026-03-26 +lastReviewed: 2026-07-01 appliesTo: main docType: reference --- @@ -44,3 +44,22 @@ Extrae metadatos de un archivo de imagen y, opcionalmente, lo devuelve como base | :--- | :--- | :--- | | `path` | `string` | **Requerido.** Ruta al archivo de imagen. | | `include_base64` | `boolean` | Incluir los datos completos de la imagen en la salida. Por defecto: `false`. | + +--- + +## `pdf_inspect` + +Inspecciona, clasifica y extrae texto de un archivo PDF. + +- **Nivel de Seguridad:** Solo Lectura (Segura). +- **Características:** Detecta el tipo de PDF (texto, escaneado, imagen, mixto), recuento de páginas y convierte el texto extraíble a Markdown. +- **Restricciones:** + - Tamaño máximo de archivo: 50 MB. + - Tiempo de espera: 60 segundos. + +### Parámetros + +| Parámetro | Tipo | Descripción | +| :--- | :--- | :--- | +| `path` | `string` | **Requerido.** Ruta relativa al archivo PDF dentro del workspace. | +| `extract_text` | `boolean` | Indica si se debe extraer y convertir el texto a Markdown (por defecto: true). |