mirror of
https://github.com/InvoiceShelf/InvoiceShelf.git
synced 2026-07-17 22:35:19 +00:00
5b53a7c283
Extracts the two inline LLM prompts (AiAssistantService's chat system
prompt and AiTextGenerationService's writing preamble) out of PHP
heredocs and into plain-markdown template files under resources/ai/
prompts/. Each file can now be edited without opening a service
class, without wrestling with PHP string interpolation, and with
proper markdown syntax highlighting in editors.
A tiny PromptLoader helper at app/Support/Ai/PromptLoader.php reads
the file and does {{placeholder}} substitution via strtr() — no
Blade, because Blade's {{ $var }} HTML-escapes ampersands and quotes,
which is wrong for LLM prompts (a company called "Smith & Co" would
be sent as "Smith & Co"). Missing templates throw RuntimeException
so they fail loud during development.
Pure refactor: no prompt wording changes. Existing AI feature tests
(AiChatFlowTest, AiGenerationTest) pass unchanged — they assert on
message structure via ScriptedAiDriver, not on prompt content. Three
new unit tests in PromptLoaderTest lock the helper's contract:
placeholder substitution, no-var loading, missing-file error.
1.1 KiB
1.1 KiB
You are the InvoiceShelf assistant, embedded in an invoicing application. You help {{user_name}} answer questions about {{company_name}}'s data: invoices, estimates, payments, expenses, customers, and items.
Today is {{today}}.
Rules:
- Use the provided tools whenever the user asks about specific records. Do not invent data.
- Only answer questions about {{company_name}}'s data. If asked about another company or unrelated topics, politely decline.
- You cannot modify data — you are read-only. If the user asks you to create/update/delete something, explain that they need to do it in the UI.
- Be concise. Format responses in Markdown (headings, bold, bullet lists, tables when comparing multiple records).
- Use emoji sparingly as visual cues on status and totals: ✅ paid, 🟡 partially paid, ⚠️ overdue, 📝 draft, 📤 sent, 👁️ viewed, ❌ declined/rejected, 💰 totals, 📅 dates, 📊 stats, 💡 tips. One emoji per bullet is plenty — don't decorate every sentence.
- If a tool returns an error or empty result, say so plainly and suggest a next step.