AIssistLab

Current Feature Inventory

This inventory is based on the current repo shape, not product aspiration.

App Shell

• Next.js app under src/app.
• Main areas: Chat, Skills, Skill Editor, Settings, Export.
• Sidebar exposes index rebuild and navigation.
• Styling is local CSS variables and Tailwind utility classes.

Settings

Current Settings capabilities:

• Structured editor for core env values and Claude CLI fields.
• Raw env editor for local troubleshooting.
• Runtime-applied provider settings for current server session.
• Windows-native folder picker when available, with in-app browser fallback.
• Path existence checks for workspace and skills paths.
• Workspace profile shortcuts stored in browser local storage.
• Claude profile selector with generic labels and sanitized display paths.
• Claude login action that opens visible login flow rather than running hidden auth in chat requests.
• Claude CLI smoke test action that uses the currently selected UI profile.
• V1 Release Readiness panel that summarizes workspace, provider, index, skill quality, Claude project, chat, and diagnostics state.

Current Settings APIs:

• GET /api/settings
• POST /api/settings
• GET /api/settings/runtime
• GET /api/settings/path-exists
• GET /api/settings/native-folder
• GET /api/settings/browse
• GET /api/settings/browse/search
• GET /api/settings/doctor
• GET /api/settings/claude-cli
• POST /api/settings/claude-cli
• GET /api/settings/claude-cli/profiles
• POST /api/settings/claude-cli/test
• GET /api/settings/claude-project
• GET /api/release/readiness

Claude Provider Modes

Current provider modes:

• anthropic_api: default local API-key mode.
• claude_code_cli: local-only Claude CLI mode.

Current CLI safeguards:

• Local CLI mode requires explicit enablement.
• CLI mode is rejected for non-local usage.
• Chat generation is rejected for hosted, production, and non-local requests across provider modes.
• The child process removes the Anthropic API key when local Claude CLI mode is active.
• Claude profile discovery keeps raw paths server-side.
• Public profile output uses generic labels like Default profile, Profile 1, and Manual profile.
• Login helper is optional; built-in Claude auth login is the fallback.

Setup Doctor

Current checks include:

• Workspace root presence and directory validity.
• Skills directory validity.
• RAG index state.
• Provider mode readiness.
• Anthropic API key presence for API mode.
• Claude CLI installation and auth state for CLI mode.
• Claude login helper availability.
• Runtime-applied provider env sync.
• Restart-required drift for non-provider env values.

Current output is structured as:

• Summary status.
• Error, warning, and ok counts.
• Top recommendation.
• Grouped checks with messages, suggested fixes, and related env keys.

RAG Index

Current behavior:

• TF-IDF data stays in memory.
• Sanitized index metadata persists under .next/cache.
• Index status can be ready, stale, missing, rebuilding, or failed.
• Skill count, chunk count, build time, stale reason, workspace display, and skills directory display are exposed through /api/index.
• Skill writes mark the index stale.

Current APIs:

• GET /api/index
• POST /api/index

Skill Management

Current behavior:

• Skills are read from the configured skills directory.
• Skills can be created, edited, deleted, and exported.
• Skills page shows a Library Readiness panel that summarizes index state, skill quality issues, and next actions using existing index and validation APIs.
• Skill editor has frontmatter fields, markdown body, live preview, mobile preview tab, unsaved-change warning, and save disabling while invalid.
• Guided Skill Builder can draft learning-style skills from purpose, audience, trigger examples, required inputs, boundaries, success criteria, and a selected template.
• Guided drafts are scored by a deterministic rubric and handed to the normal editor for review and save.
• Import preview supports local folders, zip archives, and GitHub URLs; duplicate overwrites require typing overwrite before the UI enables Apply.
• Server-side validation checks path-safe names, missing descriptions, empty bodies, duplicate tags, and oversized files.
• Quality scanner checks existing skills for deterministic issues.

Current APIs:

• GET /api/skills
• POST /api/skills
• GET /api/skills/:skillName
• PUT /api/skills/:skillName
• DELETE /api/skills/:skillName
• POST /api/skills/:skillName/restore
• GET /api/skills/templates
• GET /api/skills/validation
• POST /api/skills/guided/feedback
• POST /api/skills/guided/draft
• POST /api/skills/import/preview
• POST /api/skills/import/apply

Chat

Current behavior:

• Chat status shows provider mode, runtime source, index readiness, and last Claude CLI smoke test state.
• Chat warns when the index is not ready and can rebuild it.
• Chat blocked states link back to Settings and Diagnostics Export.
• Assistant responses stream as newline-delimited JSON chunks.
• Failed assistant responses can be retried without duplicating the user message.
• Citations show source skill, section, score, preview, and editor link.

Current APIs:

• POST /api/chat
• GET /api/chat/status

Export

Current behavior:

• Single skill export.
• Zip export for selected or all skills.
• Optional diagnostics bundle with manifest, V1 readiness report, sanitized index metadata, skill quality report, Claude project inventory, and settings summary.
• Export page shows a diagnostics readiness preview before download.

Current APIs:

• GET /api/export
• GET /api/export/zip

Existing Test Coverage

Current focused tests cover:

• Claude discovery and privacy behavior.
• Runtime config behavior.
• Setup Doctor rules.
• RAG index metadata and route behavior.
• Skill validation and quality checks.
• Local guards.
• Settings profile selection.
• Chat status.
• Export zip diagnostics.