"""Unit tests for statement accuracy fixes (merchant, count, date, third-party).""" from __future__ import annotations import re from app.pdf.calc import _detect_spend_category from app.saas.answer_style import apply_paragraph_style, transactions_to_paragraph, wants_paragraph_format from app.saas.chat_router import route_chat from app.saas.finance_intent import ( classify_finance_intent, extract_merchant_query, is_analyse_overview_ask, is_payment_count_question, is_payment_list_question, is_unrelated_party_spend_question, is_vague_single_payment_question, ) from app.saas.insight import ( _account_holder, build_date_lookup, build_merchant_spending, build_payment_count, build_payment_list, build_category_spending, build_month_spend, build_spending_summary, build_statement_summary, ) def test_zepto_is_merchant_not_full_groceries(): cat = _detect_spend_category("How much spent on Zepto?") assert cat is not None label, keys = cat assert label == "zepto" assert keys == ("zepto",) assert "blinkit" not in keys def test_groceries_still_expands(): cat = _detect_spend_category("How much spent on groceries?") assert cat is not None label, keys = cat assert label == "groceries" assert "zepto" in keys and "blinkit" in keys def test_payment_count_intent(): assert is_payment_count_question("How many payments were made?") assert classify_finance_intent("How many payments were made?") == "payment_count" def test_third_party_spend_refused(): q = "How much did Elon Musk spend last month?" assert is_unrelated_party_spend_question(q) assert classify_finance_intent(q) == "general" r = route_chat(q) assert r.target == "template_offtopic" assert not r.use_docs def test_first_person_spend_still_docs(): q = "How much did I spend last month?" assert not is_unrelated_party_spend_question(q) r = route_chat(q) assert r.use_docs def test_extract_merchant_nykaa(): assert extract_merchant_query("How much did I pay Nykaa?") name = extract_merchant_query("How much did I pay Nykaa?") assert name and "nykaa" in name.lower() def test_statement_identity_and_period_intent(): assert classify_finance_intent("Whose statement is this?") == "statement_summary" assert classify_finance_intent("What is the statement period?") == "statement_summary" def _sample_job() -> dict: return { "filename": "sample.pdf", "statement_totals": { "money_out": 1000.0, "money_in": 100.0, "payments_made": 84, "payments_received": 5, "source": "statement_header", }, "calculation": {"metrics": {}}, "rows": [ { "date": "16 Jul", "description": "Paid to Nykaa ERetail Private Limited", "debit": 993.0, "credit": None, "amount": -993.0, }, { "date": "16 Jul", "description": "Money sent to Aditya Yadav", "debit": 3000.0, "credit": None, "amount": -3000.0, }, { "date": "06 Jul", "description": "Paid to Zepto", "debit": 327.0, "credit": None, "amount": -327.0, }, { "date": "11 Jul", "description": "Paid to Blinkit", "debit": 177.0, "credit": None, "amount": -177.0, }, ], "document_text": ( "AAYUSHI VERMA\n" "Paytm Statement\n" "3 JUL'26 - 2 AUG'26 - Rs.30,298.30\n" "84 Payments made 5 Payments received\n" "16 Jul Paid to Nykaa ERetail Private Limited Note: HDFC Bank - - Rs.993\n" "10:54 PM\n" "16 Jul Money sent to Jatin Thakur Tag: State Bank - Rs.89\n" "7:13 PM\n" "16 Jul Money sent to Aditya Yadav Tag: State Bank - Rs.3,000\n" "4:04 PM\n" "16 Jul Money sent to Aditya Yadav Tag: HDFC Bank - - Rs.1\n" "4:03 PM\n" "16 Jul Money sent to Bhairab Pathak Tag: HDFC Bank - - Rs.114\n" "9:03 AM\n" "06 Jul Paid to Zepto Note: UPIIntent HDFC Bank - - Rs.327\n" "3:46 PM\n" "11 Jul Paid to Zepto Note: UPIIntent HDFC Bank - - Rs.326\n" "12:12 PM\n" "11 Jul Paid to Blinkit Note: HDFC Bank - - Rs.177\n" "1:00 PM\n" ), } def test_payment_count_uses_header(): insight = build_payment_count(_sample_job(), "How many payments were made?") assert "84" in insight.summary assert any(m.value == "84" for m in insight.metrics) def test_date_lookup_uses_full_text_day(): insight = build_date_lookup(_sample_job(), "How much did I spend on 16 July?") # 993+89+3000+1+114 = 4197 assert any("4,197" in (m.value or "") or "4197" in (m.value or "").replace(",", "") for m in insight.metrics) assert len(insight.transactions or []) == 5 def test_zepto_category_not_blinkit(): insight = build_category_spending(_sample_job(), "How much spent on Zepto?") assert insight is not None descs = " ".join(t.description for t in (insight.transactions or [])).lower() assert "zepto" in descs assert "blinkit" not in descs def test_account_holder_from_header(): holder = _account_holder(_sample_job()) assert holder and "aayushi" in holder.lower() def test_unique_ready_docs_dedupes_filename(): from app.saas.knowledge_state import unique_ready_docs docs = [ {"status": "ready", "filename": "Paytm.pdf", "created_at": "1"}, {"status": "ready", "filename": "Paytm.pdf", "created_at": "2"}, {"status": "ready", "filename": "Other.csv", "created_at": "1"}, ] out = unique_ready_docs(docs) assert len(out) == 2 assert {d["filename"] for d in out} == {"Paytm.pdf", "Other.csv"} def test_merchant_nykaa(): insight = build_merchant_spending(_sample_job(), "How much did I pay Nykaa?") assert insight is not None assert any("993" in (m.value or "").replace(",", "") for m in insight.metrics) def test_hdfc_branch_and_credits_from_statement(): from pathlib import Path from app.saas.kb_answer import build_statement_job from app.saas.insight import build_payment_count, build_statement_field_answer from app.saas.finance_intent import classify_finance_intent path = Path("/home/esfera/Documents/accountants_chatbot/Acct Statement_4640_14062026_10.58.32.pdf") if not path.exists(): return job = build_statement_job(path, path.name) assert job assert classify_finance_intent("what is the state account branch number?") == "statement_summary" field = build_statement_field_answer(job, "what is the state account branch number?") vals = " ".join(m.value for m in field.metrics) assert "1395" in vals # branch code assert "HDFC0001395" in vals or "SECTOR" in vals.upper() assert classify_finance_intent("how many credits were ,ade?") == "payment_count" credits = build_payment_count(job, "how many credits were ,ade?") assert credits.headline.lower().startswith("credit") assert any(m.label.lower().startswith("credit") and int(m.value) >= 1 for m in credits.metrics) def test_which_payment_routes_to_transaction_search(): assert is_payment_list_question("which payment did i make") assert classify_finance_intent("which payment did i make") == "transaction_search" assert is_vague_single_payment_question("which payment did i make") assert not is_vague_single_payment_question("which payments did i make") def test_which_payment_singular_clarifies(): insight = build_payment_list(_sample_job(), "which payment did i make") assert insight.headline == "Which payment?" assert "can't tell which one" in insight.summary.lower() assert not insight.table assert not insight.transactions def test_which_payments_plural_lists_without_categories(): insight = build_payment_list(_sample_job(), "which payments did i make") assert insight.headline == "Your payments" assert insight.transactions assert not insight.table assert "Spending summary" not in insight.headline def test_month_spend_intent_not_full_summary(): assert classify_finance_intent("how much did i spend in august?") == "date_lookup" assert classify_finance_intent("how much did I spend in July?") == "date_lookup" def test_august_spend_scoped_not_full_dump(): job = _sample_job() # Sample rows are all July — August should not dump the whole statement. insight = build_month_spend(job, "how much did i spend in august?", "aug") assert insight.headline == "Spend in August" assert "Spending summary" not in insight.headline assert "august" in insight.summary.lower() or "August" in insight.summary # No July payment list dumped when month has no matches assert not insight.transactions assert any("No payments" in (m.value or "") for m in insight.metrics) def test_july_spend_scoped(): insight = build_date_lookup(_sample_job(), "how much did i spend in july?") assert insight.headline == "Spend in July" assert insight.transactions assert all("jul" in (t.date or "").lower() for t in insight.transactions if t.date) def test_refined_routing_intents(): assert is_analyse_overview_ask("analyse the docment") assert not is_analyse_overview_ask("how much did i spend in total?") assert not is_analyse_overview_ask("list all payments") assert classify_finance_intent("how much did i spend in total?") == "spending_summary" assert classify_finance_intent("list all payments") == "transaction_search" assert classify_finance_intent("how many payments did i make in 1 august") == "date_lookup" def test_compact_overview_no_payment_list(): insight = build_statement_summary(_sample_job(), "analyse the document") assert insight.headline == "Statement overview" assert not insight.transactions assert any(m.label == "Total spent" for m in insight.metrics) # Narrative should observe, not dump "Overall: total received … total spent …" assert "overall:" not in insight.summary.lower() assert "net" in insight.summary.lower() or "outflow" in insight.summary.lower() or "inflow" in insight.summary.lower() def test_total_spend_compact_not_overview(): insight = build_spending_summary(_sample_job(), "how much did i spend in total?") assert insight.headline == "Spending summary" assert not insight.transactions assert insight.summary.lower().startswith("you spent") assert "largest" in insight.summary.lower() def test_list_payments_light_not_overview(): insight = build_payment_list(_sample_job(), "list all payments") assert insight.headline == "Your payments" assert insight.transactions assert "Statement overview" not in insight.headline def test_biggest_expenses_ranked_list(): from app.saas.finance_intent import is_biggest_expenses_question from app.saas.insight import build_biggest_expenses q = "Show my biggest expenses" assert is_biggest_expenses_question(q) assert classify_finance_intent(q) == "insights" insight = build_biggest_expenses(_sample_job(), q) assert insight.headline == "Biggest expenses" assert insight.transactions assert insight.table and insight.table.headers[:3] == ["#", "Date", "Merchant / description"] # Ranked largest-first amounts = [] for t in insight.transactions: raw = (t.amount or "").replace(",", "").replace("₹", "").replace("Rs.", "").strip() try: amounts.append(float(re.sub(r"[^\d.]", "", raw) or 0)) except ValueError: pass assert amounts == sorted(amounts, reverse=True) # Should not dump the full totals overview as the headline story assert "overall:" not in insight.summary.lower() def test_general_knowledge_skips_rag(): for q in ( "what is the capital of France", "who invented the telephone", "what is photosynthesis", "tell me a joke", ): r = route_chat(q) assert r.use_docs is False, (q, r) assert r.target == "ollama_concept", (q, r) def test_payment_list_never_claims_all_when_incomplete(): job = _sample_job() # Header says 84; sample only has a handful of rows. insight = build_payment_list(job, "list all payments") assert insight.transactions text = (insight.summary + " " + " ".join(insight.footnotes or [])).lower() assert "incomplete" in text or "indexed" in text assert not re.search(r"showing all 84", text) assert "showing all" not in text or "indexed" in text def test_payment_list_includes_text_august_with_month(): """Full payment list must use the same text+rows index as month lookups.""" job = _sample_job() # Inject an August payment only in document text (not in table rows). job = { **job, "document_text": job["document_text"] + "\n01 Aug Money sent to Test Payee Tag: HDFC Bank - - Rs.250\n10:00 AM\n", "rows": list(job["rows"]), # July-only table rows } month = build_month_spend(job, "how much did i spend in august?", "aug") assert month.transactions, "August text hit should surface in month spend" assert any("aug" in (t.date or "").lower() for t in month.transactions) listing = build_payment_list(job, "list all payments") assert any("aug" in (t.date or "").lower() for t in listing.transactions), ( "August text payments must appear in list-all, same as month query" ) def test_coverage_challenge_explains_gap(): from app.saas.conversation import is_coverage_challenge from app.saas.insight import build_coverage_explanation assert is_coverage_challenge("why only 40 payments?") assert is_coverage_challenge("where are the rest of the transactions") insight = build_coverage_explanation(_sample_job(), "why only 4 of 84?") assert "payments" in insight.summary.lower() assert "84" in insight.summary or any("84" in (m.value or "") for m in insight.metrics) assert insight.headline == "Payment coverage" assert "extraction" not in insight.summary.lower() assert "indexed" not in insight.summary.lower() def test_list_footnote_honest(): from app.saas.insight import _list_footnote note = _list_footnote(40, 40, header_n=84) assert note is not None assert "40" in note and "84" in note assert "extraction" not in note.lower() complete = _list_footnote(84, 84, header_n=84) assert complete is not None assert "all 84" in complete.lower() def test_august_payment_count_date_lookup(): insight = build_date_lookup(_sample_job(), "how many payments did i make in 1 august") assert insight.headline.startswith("Payments on 1 Aug") assert "no payments" in insight.summary.lower() assert not insight.transactions def test_vague_what_routes_to_clarify(): assert route_chat("what").target == "template_clarify" assert route_chat("gi").target == "template_greeting" def test_full_list_paragraph_intent(): assert classify_finance_intent("i need full list of payments in a paragraph") == "transaction_search" assert wants_paragraph_format("in a paragraph") def test_payment_list_paragraph_style(): insight = build_payment_list(_sample_job(), "list all payments in a paragraph") styled = apply_paragraph_style(insight, "list all payments in a paragraph") assert not styled.transactions assert "paid" in styled.summary.lower() assert len(styled.summary) > 80