Hacktoberfest 2026: los issues que los mantenedores marcaron para octubre, abiertos y aptos para principiantes. Explorar issues de Hacktoberfest

🎃 Export the audit log as CSV or JSON from /logs

Abierto
#998 0 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
4/5
Tiempo estimado
3-5 días
Aptitud para principiantes
22/100
Tipo de issue
Nueva funcionalidad
Claridad
Bien especificado
Estado de actividad
Activo
Stack tecnológico
typescript

Línea de trabajo

Start with getRecentInvocations and orgScope() in packages/backend/src/audit/audit.service.ts, and the list endpoint in audit.controller.ts, to see how the where filters are built today. Then look at the admin-only guard pattern in site-settings.controller.ts. Done means the export endpoint streams batches under the row cap, returns 403 for non-admins and for a missing organizationId, and the audit specs and Playwright test pass.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

enhancement hacktoberfest help wanted

Part of the October Challenge #997.

Why

Admins regularly need the tool-call history outside the product: for a security review, to attach to a support ticket, to load into a spreadsheet or a SIEM, or to keep it beyond the retention window. Today the Audit Log page (/logs) shows 50 rows per page and offers no download. The only way out is paging through GET /api/audit/invocations by hand. GitHub's audit log, Auth0 and Grafana all offer CSV and JSON export of the filtered view.

What already exists: the page filters by status, tool/connector search, connector and MCP server (logs/page.tsx:116-134), and the API supports the same filters plus limit/offset (audit.controller.ts:13-37). There is no date filter and no export. The header button passed to AppShell as exportButton (logs/page.tsx:183-207) is only a Refresh button with a misleading variable name. docs/api-reference.md:226-253 documents toolName, from and to filters that the API does not have, and a stats response (last24h, last7d, total) that does not match getStats() (invocations24h, errors24h, invocations7d, totalInvocations).

Self-hosted vs Cloud

  • Self-hosted: enabled in Community and Business. No ee/ code. There is no retention job on self-hosted, so an export can cover the whole history of the organization.
  • Cloud (DEPLOYMENT_MODE=cloud): enabled. Rows older than INVOCATION_RETENTION_DAYS (default 90) are deleted and payloads older than INVOCATION_PAYLOAD_RETENTION_DAYS (default 14) are cut to an excerpt (ee/cloud/kg-cron.service.ts:372, :407), so an export only contains what is still stored. Say so in a tooltip next to the button. Do not change anything under ee/. The export must be streamed in batches: an organization can have hundreds of thousands of rows, and loading them in one query would put the shared backend's heap at risk.
  • Who can use it: org ADMIN only (@Roles('ADMIN') with RolesGuard, the pattern on site-settings.controller.ts:111-112). Other roles keep reading /logs as today but get no bulk download. Rows are always scoped to the caller's organization. If the JWT carries no organizationId, return 403: orgScope() returns {} (no filter) for a missing org (audit.service.ts:297-300), and an export must never fall back to all rows.

What to build

  1. Shared filter builder: move the where construction out of getRecentInvocations (audit.service.ts:302-345) into one helper used by the list and the export, and add from / to on createdAt (ISO date or datetime; from inclusive, to exclusive). The list endpoint gets from/to too.
  2. Endpoint: GET /api/audit/invocations/export?format=csv|json with the list filters (status, search, connectorId, mcpServerId, toolId, from, to) plus payloads=true|false (default false). Headers: Content-Type: text/csv; charset=utf-8 or application/json, and Content-Disposition: attachment; filename="anythingmcp-audit-<from>-<to>.<ext>".
  3. Streaming: read in batches of 1,000 with a keyset cursor on (createdAt desc, id desc) and write each batch to the response (@Res()), never collecting all rows in memory. Cap one export at AUDIT_EXPORT_MAX_ROWS (default 100,000). When the cap is reached, end the file cleanly and add a final marker (a last CSV comment line or a truncated: true field) so the UI can say "narrow the date range". For JSON, write an array incrementally or NDJSON; pick one and document it.
  4. Columns (same names in CSV and JSON): createdAt (ISO 8601 UTC), status, tool, connector, connectorType, mcpServer, userEmail, authMethod and apiKeyName (both parsed from clientInfo), durationMs, repeatCount, error; with payloads=true also input and output (JSON strings in CSV). Keep repeatCount: one stored row can stand for many identical failures (audit.service.ts:12-18).
  5. CSV safety: RFC 4180 quoting (quotes, commas, newlines), UTF-8 with BOM so Excel shows umlauts correctly, and formula-injection protection: prefix a cell that starts with =, +, -, @, tab or carriage return with '. Tool arguments and upstream errors are user- and vendor-controlled text.
  6. UI: on /logs, rename exportButton to what it is and add an "Export" menu next to Refresh (CSV, JSON, and an "Include input and output" option), shown to admins only (useAuth().user.role). Add a From / To date range to the toolbar (logs/page.tsx:210-267) that drives both the list and the export. request() in lib/api.ts:27-57 always parses JSON, so add a separate client helper that uses fetch + blob() and an object URL, like handleExportAll on the Connectors page (connectors/page.tsx:139-160).

Where to look

  • packages/backend/src/audit/audit.controller.ts:13-37: list endpoint and its filters.
  • packages/backend/src/audit/audit.service.ts:297-345: orgScope() and getRecentInvocations() (the includes give tool, connector, user and server names).
  • packages/backend/prisma/schema.prisma:678-718: ToolInvocation (clientInfo is a JSON string; repeatCount).
  • packages/backend/src/auth/roles.guard.ts:11 and packages/backend/src/settings/site-settings.controller.ts:111-112: admin-only guard pattern.
  • packages/backend/src/ee/cloud/kg-cron.service.ts:372, :407: Cloud payload trim and retention (read only).
  • packages/frontend/src/app/logs/page.tsx:83-207: state, filters, fetch, header actions.
  • packages/frontend/src/lib/api.ts:753-765: audit.invocations() client.
  • packages/frontend/src/components/ui/action-menu.tsx: dropdown for the menu.
  • docs/api-reference.md:219-253: Audit section to correct and extend.

Acceptance criteria

  • Unit specs next to the service (cd packages/backend && npm test): filters and from/to build the expected where; another organization's rows never appear; CSV escaping (quotes, newlines, a cell starting with =); the row cap ends the file with the truncation marker; payloads=false omits input/output.
  • EDITOR and VIEWER get 403 from the export endpoint; a JWT without organizationId gets 403.
  • Exporting 50,000 seeded rows keeps backend memory flat (describe how you checked in the PR).
  • The CSV opens in LibreOffice or Excel with correct columns and umlauts.
  • Playwright test in packages/frontend/tests/e2e/ (stub /api/audit/invocations* like mobile.spec.ts): an admin sees the Export menu and the request carries the active filters and dates; a viewer does not see it. Run with cd packages/frontend && npm run test:e2e.
  • docs/api-reference.md: export endpoint documented, toolName example removed, from/to documented, stats response corrected.

Out of scope

  • Scheduled or emailed exports, pushing to S3 or a SIEM.
  • Exporting security events or product events.
  • Changing retention on Cloud or adding retention to self-hosted.
  • Any change under ee/.

Size

M (1–2 days)

How to claim

Comment "I'd like to work on this" and we'll assign you. Rules in #997.

Lenguaje dominante
TypeScript
Estrellas
856
Forks
82
Merge medio
8 h 27 min
PR fusionados (30 d)
104

Preparar el entorno

Este proyecto no incluye contenedor de desarrollo, Dockerfile ni guía de contribución, así que la configuración corre por tu cuenta: empieza por su README y consulta nuestra guía para la primera contribución para los pasos generales.

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de HelpCode-ai/anythingmcp

Todos los issues de HelpCode-ai/anythingmcp

Issues similares

Más issues de TypeScript

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.