Upload folder using huggingface_hub

builders/build_excel.py

@@ -507,7 +507,169 @@ def _dcf_sheet(wb, data):
     _widths(ws, [30, 20, 20, 20, 20, 20])
 
 
-def _narrative_sheet(wb, data):
+def _dcf_sheet(wb, data):
+    ws = wb.create_sheet("DCF Valuation Model")
+    _title(ws, "Interactive Discounted Cash Flow (DCF)", "Change the yellow highlighted cells to update assumptions")
+    
+    currency = data.get("currency", "USD")
+    units = data.get("company", {}).get("units", "Millions")
+    
+    # 1. Historical data
+    rev_hist = data.get("income_statement", {}).get("revenue", [])
+    fcf_hist = data.get("cash_flow", {}).get("free_cash_flow", [])
+    
+    last_rev = _last(rev_hist) or 0
+    last_fcf = _last(fcf_hist) or 0
+    
+    # Default assumptions
+    default_wacc = 0.09
+    default_tg = 0.025
+    
+    # Historical FCF Margin & Growth
+    rev_cagr = data.get("growth_metrics", {}).get("revenue_cagr_5y")
+    if rev_cagr is None:
+        rev_cagr = 0.05
+    rev_cagr = max(0.02, min(0.15, rev_cagr)) # Cap between 2% and 15%
+    
+    fcf_margins = [_safe_div(f, r) for f, r in zip(fcf_hist, rev_hist) if f is not None and r is not None and r != 0]
+    avg_fcf_margin = sum(fcf_margins[-3:]) / len(fcf_margins[-3:]) if len(fcf_margins) > 0 else 0.10
+    
+    shares_out = _last(data.get("income_statement", {}).get("diluted_shares", [])) or 1
+    if shares_out <= 0: shares_out = 1 # Prevent division by zero
+    net_debt = _last(data.get("net_debt", [])) or 0
+    
+    # --- Assumptions Block ---
+    r = 4
+    _sec(ws, r, "1. Model Assumptions", 3)
+    ws.cell(row=r, column=4, value="Yellow cells are editable").font = Font(italic=True, color="595959")
+    r += 1
+    
+    def add_assumption(row_idx, label, default_val, is_pct=True):
+        ws.cell(row=row_idx, column=1, value=label).font = LBL_FONT
+        c = ws.cell(row=row_idx, column=2, value=default_val)
+        c.fill = YLW_FILL
+        c.font = Font(name="Calibri", size=11, bold=True, color="1F4E78")
+        c.border = THIN
+        c.alignment = RIGHT
+        c.number_format = FMT_PCT if is_pct else FMT_DEC
+        return c.coordinate
+        
+    wacc_cell = add_assumption(r, "Weighted Average Cost of Capital (WACC)", default_wacc); r += 1
+    tg_cell = add_assumption(r, "Terminal Growth Rate", default_tg); r += 1
+    rev_gr_cell = add_assumption(r, "Projected Revenue Growth (Years 1-5)", rev_cagr); r += 1
+    fcf_m_cell = add_assumption(r, "Projected FCF Margin", avg_fcf_margin); r += 1
+    
+    r += 1
+    
+    # --- Projections Block ---
+    _sec(ws, r, "2. Free Cash Flow Projections", 7)
+    r += 1
+    
+    _hdr(ws, r, ["Metric", "Year 0 (Current)", "Year 1", "Year 2", "Year 3", "Year 4", "Year 5"])
+    r += 1
+    
+    # Revenue row
+    ws.cell(row=r, column=1, value=f"Revenue ({currency} {units})").font = LBL_FONT
+    ws.cell(row=r, column=2, value=last_rev).number_format = FMT_MONEY
+    ws.cell(row=r, column=2).font = DAT_FONT
+    
+    rev_cells = []
+    for i in range(1, 6):
+        c = ws.cell(row=r, column=2+i, value=f"={get_column_letter(2+i-1)}{r}*(1+{rev_gr_cell})")
+        c.number_format = FMT_MONEY; c.font = DAT_FONT
+        rev_cells.append(c.coordinate)
+    r += 1
+    
+    # FCF row
+    ws.cell(row=r, column=1, value=f"Free Cash Flow ({currency} {units})").font = LBL_FONT
+    ws.cell(row=r, column=2, value=last_fcf).number_format = FMT_MONEY
+    ws.cell(row=r, column=2).font = DAT_FONT
+    
+    fcf_cells = []
+    for i in range(1, 6):
+        c = ws.cell(row=r, column=2+i, value=f"={rev_cells[i-1]}*{fcf_m_cell}")
+        c.number_format = FMT_MONEY; c.font = DAT_FONT
+        fcf_cells.append(c.coordinate)
+    r += 1
+    
+    # Discount Factor
+    ws.cell(row=r, column=1, value="Discount Factor").font = LBL_FONT
+    ws.cell(row=r, column=2, value=1.0).number_format = FMT_DEC
+    ws.cell(row=r, column=2).font = Font(italic=True, color="595959")
+    
+    df_cells = []
+    for i in range(1, 6):
+        c = ws.cell(row=r, column=2+i, value=f"=1/((1+{wacc_cell})^{i})")
+        c.number_format = FMT_DEC; c.font = Font(italic=True, color="595959")
+        df_cells.append(c.coordinate)
+    r += 1
+    
+    # PV of FCF
+    ws.cell(row=r, column=1, value="Present Value of FCF").font = Font(name="Calibri", bold=True)
+    ws.cell(row=r, column=2, value="--").alignment = RIGHT
+    
+    pv_fcf_cells = []
+    for i in range(1, 6):
+        c = ws.cell(row=r, column=2+i, value=f"={fcf_cells[i-1]}*{df_cells[i-1]}")
+        c.number_format = FMT_MONEY; c.font = Font(name="Calibri", bold=True)
+        pv_fcf_cells.append(c.coordinate)
+    
+    sum_pv_fcf_cell = f"SUM({pv_fcf_cells[0]}:{pv_fcf_cells[-1]})"
+    r += 2
+    
+    # --- Valuation Bridge ---
+    _sec(ws, r, "3. Valuation Bridge", 3)
+    r += 1
+    
+    # TV
+    ws.cell(row=r, column=1, value="Terminal Value (TV)").font = LBL_FONT
+    tv_formula = f"={fcf_cells[-1]}*(1+{tg_cell})/({wacc_cell}-{tg_cell})"
+    c = ws.cell(row=r, column=2, value=tv_formula); c.number_format = FMT_MONEY; c.font = DAT_FONT
+    tv_cell = c.coordinate
+    r += 1
+    
+    # PV of TV
+    ws.cell(row=r, column=1, value="PV of Terminal Value").font = LBL_FONT
+    pv_tv_formula = f"={tv_cell}*{df_cells[-1]}"
+    c = ws.cell(row=r, column=2, value=pv_tv_formula); c.number_format = FMT_MONEY; c.font = DAT_FONT
+    pv_tv_cell = c.coordinate
+    r += 1
+    
+    # EV
+    ws.cell(row=r, column=1, value="Implied Enterprise Value").font = Font(name="Calibri", bold=True, size=12)
+    ev_formula = f"={sum_pv_fcf_cell}+{pv_tv_cell}"
+    c = ws.cell(row=r, column=2, value=ev_formula); c.number_format = FMT_MONEY; c.font = Font(name="Calibri", bold=True, size=12)
+    ev_cell = c.coordinate
+    r += 1
+    
+    # Net Debt
+    ws.cell(row=r, column=1, value="Less: Net Debt").font = LBL_FONT
+    c = ws.cell(row=r, column=2, value=net_debt); c.number_format = FMT_MONEY; c.font = DAT_FONT
+    nd_cell = c.coordinate
+    r += 1
+    
+    # Equity Value
+    ws.cell(row=r, column=1, value="Implied Equity Value").font = Font(name="Calibri", bold=True, size=12)
+    eq_formula = f"={ev_cell}-{nd_cell}"
+    c = ws.cell(row=r, column=2, value=eq_formula); c.number_format = FMT_MONEY; c.font = Font(name="Calibri", bold=True, size=12)
+    eq_cell = c.coordinate
+    r += 1
+    
+    # Shares
+    ws.cell(row=r, column=1, value="Diluted Shares Outstanding").font = LBL_FONT
+    c = ws.cell(row=r, column=2, value=shares_out); c.number_format = FMT_DEC; c.font = DAT_FONT
+    shares_cell = c.coordinate
+    r += 1
+    
+    # Implied Share Price
+    ws.cell(row=r, column=1, value="Implied Share Price").font = Font(name="Calibri", bold=True, size=14, color="1F4E78")
+    share_price_formula = f"={eq_cell}/{shares_cell}"
+    c = ws.cell(row=r, column=2, value=share_price_formula); c.number_format = FMT_DEC; c.font = Font(name="Calibri", bold=True, size=14, color="1F4E78")
+    c.fill = GRN_FILL
+    r += 2
+    
+    # Format and fit
+    _widths(ws, [35, 18, 18, 18, 18, 18, 18])
     ws = wb.create_sheet("Industry & AI Insights")
     _title(ws, "AI Narrative Analysis", "Generated by AI Models")
     r = 4

pipeline/analyst.py

@@ -222,14 +222,23 @@ Guidelines:
 - Output ONLY the final synthesized text, with NO markdown formatting, NO introductory text, and NO commentary. Just the plain text paragraphs.
 """
 
-GENERATE_PPT_JSON = """You are a professional investment banker creating a pitch deck from financial data.
-Your task is to generate the structure and content for a PowerPoint presentation summarizing the company's financial health.
+GENERATE_PPT_JSON = """You are a top-tier Management Consultant (e.g., McKinsey, BCG) tasked with creating a highly visual, insightful, and strategic slide deck for a C-suite presentation.
+You have been provided with a detailed textual narrative analysis of the company's financial health, trends, SWOT, and peer benchmarking.
+Your task is to synthesize this narrative into a structured PowerPoint presentation.
+
 Output a strict JSON array of slide objects. Each slide object MUST have:
-- "title": A short, punchy title for the slide
-- "bullets": An array of 3-5 strings, each containing a compelling insight, metric, or trend based on the data.
+- "title": A short, punchy, action-oriented title for the slide (e.g., "Margins Contract Amidst Pricing Pressure" rather than "Margin Analysis").
+- "bullets": An array of 3-5 strings. These MUST NOT be generic summaries. Each bullet must be a deep, consultant-level insight, data-backed takeaway, or strategic implication extracted directly from the provided narrative.
+
+DO NOT include an Executive Summary slide, as that is handled separately.
+Focus on:
+1. Strategic Overview & Market Position
+2. Revenue & Profitability Dynamics
+3. Balance Sheet & Cash Flow Health
+4. Competitive Benchmarking Insights
+5. Key Risks & Strategic Imperatives
 
-Do not include any markdown fences, introduction, or conclusion. Just the raw JSON array.
-Focus on: Executive Summary, Revenue & Margin Trends, Cash Flow Health, Valuation, and Key Risks.
+Do not include any markdown fences (like ```json), introduction, or conclusion. Output ONLY the raw JSON array.
 """
 
 # -- Section prompts -----------------------------------------------------------
@@ -676,10 +685,10 @@ def generate_narrative(
         return _synthesize_drafts(section, drafts)
         
     def _generate_ppt_json() -> str:
-        # Pick the smartest model available for structured JSON generation
+        # Pick the smartest model available for structured JSON generation (prioritize Anthropic Claude for Consultant output)
         synthesizer = None
-        if use_openai:  synthesizer = ("ChatGPT", _openai_generate)
-        elif use_claude: synthesizer = ("Claude", _claude_generate)
+        if use_claude:   synthesizer = ("Claude", _claude_generate)
+        elif use_openai: synthesizer = ("ChatGPT", _openai_generate)
         elif use_gemini: synthesizer = ("Gemini", _gemini_generate)
         elif use_nvidia: synthesizer = ("NVIDIA", _nvidia_generate)
         elif use_groq:   synthesizer = ("Groq", _groq_generate)
@@ -688,12 +697,15 @@ def generate_narrative(
             return ""
             
         name, fn = synthesizer
-        log(f"\n[PPT] Generating PowerPoint Structure using {name} ...")
+        log(f"\n[PPT] Generating PowerPoint Structure using {name} (Consultant-level Insights) ...")
+        
+        # Build narrative context to feed the PPT generator
+        narrative_context = f"=== Company Overview ===\n{company_overview}\n\n=== SWOT ===\n{swot_raw}\n\n=== Trends ===\n{trend_revenue}\n{trend_bs}\n{trend_cf}\n\n=== Peer Benchmarking ===\n{peer_commentary}\n\n=== Risks & Insights ===\n{risks_raw}\n{insights_raw}"
         
         original_prompt = SECTION_PROMPTS.get("ppt_json", "")
         SECTION_PROMPTS["ppt_json"] = GENERATE_PPT_JSON
         try:
-            res = fn(context, "ppt_json")
+            res = fn(narrative_context, "ppt_json")
         finally:
             if "ppt_json" in SECTION_PROMPTS:
                 del SECTION_PROMPTS["ppt_json"]