Fix the broken ESLint setup: add vue-eslint-parser and @typescript-eslint/parser
and wire the TS parser into eslint.config.mjs so .ts and <script lang=ts> parse
(was failing outright). Clear the resulting backlog to a clean 0/0 baseline —
fix genuine issues, relax two intentional-pattern rules (multi-word-component-names,
no-required-prop-with-default).
Add a committed .githooks/pre-commit (enabled via core.hooksPath, auto-set by the
prepare script) that runs Pint on staged PHP and ESLint --max-warnings 0 on staged
resources/scripts JS/TS/Vue, blocking on failure. Add composer/npm lint scripts and
document the gate in CLAUDE.md.
Replace every scattered v-html with a single audited BaseSanitizedHtml component
that DOMPurify-sanitizes its input (new utils/markdown.ts sanitizeHtml), so
server/registry-provided HTML is actually sanitized and vue/no-v-html stays enabled
everywhere but one reviewed sink.
The conversation-list sidebar header and the chat panel header lived
in separate columns with different heights — the "+ New conversation"
wrapper was ~60px (p-3 + py-2 button) while the "AI Assistant" header
was ~44px (p-3 + small icon/text). The resulting staircase looked
unintentional.
Pins both to h-12 (48px) so they form a single unified top bar across
the drawer. Shrinks the "+ New conversation" button to text-xs /
py-1 / px-2 so it fits the tighter height without clipping, and
switches its alignment to center to match the compacter footprint.
The AI chat drawer was rendering assistant responses as plain text,
so code blocks, lists, tables and inline formatting came through as
literal asterisks and backticks — noisy and hard to scan.
Adds a shared renderMarkdown() helper in resources/scripts/utils/
markdown.ts that parses GFM markdown via marked and sanitizes the
result with DOMPurify before handing it to Vue's v-html. AiChatMessage
uses the helper for assistant messages only; user messages stay as
plain text since markdown syntax in their own typed input would be
surprising.
Assistant bubbles get the Tailwind `prose prose-sm` classes from the
already-enabled @tailwindcss/typography plugin so headings, lists and
code blocks inherit sensible defaults without per-element styling.
Security: DOMPurify runs in its default browser profile, which strips
<script>, event handlers, javascript: URLs and every other XSS vector.
The AI provider isn't a trusted source — it can echo arbitrary user
input and tool-call results from the database — so sanitization is
non-negotiable even though the immediate source is our own backend.
Second phase of the AI feature. Users can now open a slide-in chat drawer from the SiteHeader and ask natural-language questions about their company's invoices, customers, payments, and expenses. The LLM invokes pre-defined read-only tool functions (scoped to the current company at execute time) to fetch data and synthesize answers.
**Database** — new ai_conversations and ai_messages tables. Messages are stored in OpenAI's chat format so AiAssistantService serializes a conversation into an API request with zero translation. Columns: role, content, tool_call_id, tool_calls JSON, model, tokens_in, tokens_out. Conversations are scoped (company_id, user_id) — one user's chats are invisible to everyone else, even inside the same company. Foreign-key cascade deletes.
**Tool infrastructure** — AiTool abstract base + AiToolRegistry singleton (registered in a new AiServiceProvider). The base class enforces the security rule: every tool's execute() receives companyId and userId as injected parameters; tools' JSON schemas NEVER include a company_id field. An LLM physically cannot pass a company_id and escape tenancy. Modules can register their own tools by resolving the registry from their own ServiceProvider::boot().
**Nine built-in tools**: search_invoices, get_invoice, search_customers, get_customer, list_recent_payments, list_overdue_invoices, get_company_stats (aggregates for named periods), search_items, list_expense_categories. All read-only; no mutations. Each returns JSON-encodable data the LLM can parse.
**AiAssistantService orchestration loop** — the heart of Phase 2. Flow: persist user message → build payload from system prompt + recent history (40-message window) + new user message → call driver.chatCompletion with tools → if tool_calls, execute each one via the registry (with injected scope), persist tool result, loop → if plain text, persist and return. Hard cap at 5 iterations to prevent runaway LLMs. System prompt pins the assistant to this company's data and forbids mutation.
**Controllers + policy + rate limit** — POST /api/v1/ai/chat runs the orchestration loop. GET/PATCH/DELETE /api/v1/ai/conversations for CRUD. AiConversationPolicy enforces user_id+company_id match on every action. A new 'ai' RateLimiter in RouteServiceProvider throttles to 30 req/min per (user, company). New 'use ai' Gate defined in AppServiceProvider returns true for any authenticated user — the per-company kill-switch still goes through AiConfigurationService::resolveForCompany.
**Frontend** — new features/company/ai/ folder with a Pinia store (ai-chat.store.ts) holding drawer state, current conversation, messages, and loading flags. AiChatDrawer.vue is a slide-in panel teleported to <body>, mounted globally in CompanyLayout.vue when bootstrap reports ai.enabled && ai.chat_enabled. Sub-components: AiChatMessage (user bubbles vs assistant bubbles), AiChatMessageInput (Enter submits, Shift+Enter newline), AiChatConversationList (sidebar with 'new chat' button, rename, delete). A SparklesIcon button in SiteHeader toggles the drawer.
**Driver test double** — tests use a ScriptedAiDriver registered via AiDriverFactory::register('scripted', ...) that returns pre-queued AiChatResponse objects. Feature tests cover: happy path (new conversation + message persistence), tool-call loop (multi-round-trip with search_invoices), runaway-loop cap, driver-throws path, ai_enabled=NO rejection, chat role disabled rejection, per-user conversation visibility, cross-user policy enforcement, cascade delete.
388 tests pass (was 372, +16 new). Pint clean. npm run build clean. Phase 3 (WYSIWYG text generation popup) is the remaining follow-up.
