Implementation Guide: Generate CDRL Deliverables, OMB Budget Justifications & Program Documents

# Builds structured JSON templates from DID requirements for use in AI
# generation prompts

# did_template_builder.py
# Builds structured JSON templates from DID requirements for use in AI generation prompts

DID_TEMPLATES = {
    "DI-MGMT-81466": {
        "title": "Program Management Plan (PMP)",
        "frequency": "Once + Updates",
        "sections": [
            "1. Introduction and Program Overview",
            "2. Program Organization and Management Structure",
            "3. Work Breakdown Structure (WBS)",
            "4. Integrated Master Schedule Summary",
            "5. Risk Management Approach",
            "6. Technical Performance Measures",
            "7. Subcontractor/Supplier Management",
            "8. Configuration Management",
            "9. Data Management",
            "10. Earned Value Management System Description"
        ],
        "required_data_inputs": [
            "contract_number", "program_name", "period_of_performance",
            "org_chart", "wbs_structure", "ims_summary", "risk_register_summary",
            "tpm_list", "subcontractor_list"
        ],
        "typical_page_count": "30-60",
        "classification": "CUI//PROCURE"
    },
    "DI-MGMT-81650": {
        "title": "Risk Management Plan (RMP)",
        "frequency": "Once + Quarterly Updates",
        "sections": [
            "1. Purpose and Scope",
            "2. Risk Management Process Overview",
            "3. Risk Identification Methodology",
            "4. Risk Analysis (Likelihood x Consequence)",
            "5. Risk Mitigation Strategies",
            "6. Risk Register (Current Period)",
            "7. Risk Retirement Criteria",
            "8. Roles and Responsibilities"
        ],
        "required_data_inputs": [
            "risk_register", "risk_matrix_thresholds",
            "current_top_risks", "mitigations_status"
        ],
        "typical_page_count": "15-30",
        "classification": "CUI//PROCURE"
    },
    "DI-IPSC-81431": {
        "title": "System Design Document (SDD)",
        "frequency": "Milestone-based (PDR, CDR)",
        "sections": [
            "1. Introduction",
            "2. System Overview",
            "3. Design Considerations",
            "4. Architecture Design",
            "5. Component Design",
            "6. Interface Design",
            "7. Data Design",
            "8. Human Interface Design",
            "9. Security Design",
            "10. Appendices"
        ],
        "required_data_inputs": [
            "system_description", "architecture_diagrams",
            "interface_control_documents", "security_requirements",
            "design_decisions_log"
        ],
        "typical_page_count": "50-150",
        "classification": "CUI//SP-EXPT"
    }
}

def get_did_template(did_number: str) -> dict:
    template = DID_TEMPLATES.get(did_number)
    if not template:
        raise ValueError(f"DID {did_number} not found in template library. Add manually.")
    return template

def list_available_dids() -> list:
    return [(did, info["title"]) for did, info in DID_TEMPLATES.items()]

# cdrl_generator.py

from openai import AzureOpenAI
import os, json, datetime
from did_template_builder import get_did_template

client = AzureOpenAI(
    azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
    api_key=os.environ["AZURE_OPENAI_KEY"],
    api_version="2024-08-01-preview"
)

CDRL_SYSTEM_PROMPT = """You are a defense program management documentation specialist
with expertise in DoD acquisition documentation standards. You generate CDRL deliverables
that comply with Data Item Description (DID) requirements and meet government customer
acceptance criteria.

RULES:
- Follow the DID section structure exactly
- Use formal, precise technical language appropriate for government deliverables
- Do not fabricate technical data, metrics, or status not provided in the inputs
- Flag missing data with [DATA REQUIRED: description] rather than inventing values
- Use passive voice for factual descriptions, active voice for process descriptions
- Format section headers per DID numbering scheme
- Mark output as DRAFT requiring program manager and technical lead review
- Include revision history table at document start (Version 1.0 — Initial Draft)"""

def generate_cdrl(
    did_number: str,
    contract_number: str,
    program_name: str,
    period: str,
    program_data: dict,
    reporting_period: str = None
) -> str:
    """Generate a draft CDRL deliverable."""

    template = get_did_template(did_number)
    reporting_period = reporting_period or datetime.date.today().strftime("%B %Y")

    sections_formatted = "\n".join([f"  {s}" for s in template["sections"]])
    data_formatted = json.dumps(program_data, indent=2)

    user_prompt = f"""Generate a draft {template['title']} ({did_number}) for:

CONTRACT: {contract_number}
PROGRAM: {program_name}
PERIOD: {period}
REPORTING PERIOD: {reporting_period}
CLASSIFICATION: {template['classification']}

REQUIRED SECTIONS (per DID):
{sections_formatted}

PROGRAM DATA INPUTS:
{data_formatted}

INSTRUCTIONS:
- Generate each section fully based on the data provided
- For sections where data is missing or insufficient, generate a framework and mark [DATA REQUIRED]
- Typical length: {template['typical_page_count']} pages
- Include a Document Information section at the top with contract number, CDRL item, DID reference, date, version
- Include a blank Approval Signature block at the end

Generate the complete draft document now."""

    response = client.chat.completions.create(
        model=os.environ["AZURE_OPENAI_DEPLOYMENT"],
        messages=[
            {"role": "system", "content": CDRL_SYSTEM_PROMPT},
            {"role": "user", "content": user_prompt}
        ],
        temperature=0.1,
        max_tokens=4000
    )

    header = f"""[DRAFT — AI GENERATED — REQUIRES PROGRAM MANAGER AND TECHNICAL LEAD REVIEW]
[CLASSIFICATION: {template['classification']}]
[DO NOT DISTRIBUTE TO GOVERNMENT CUSTOMER UNTIL APPROVED]

"""
    return header + response.choices[0].message.content


# Example usage:
if __name__ == "__main__":
    program_data = {
        "contract_number": "FA8621-24-C-0042",
        "program_name": "Advanced Sensor Integration System (ASIS)",
        "period_of_performance": "01 Oct 2024 – 30 Sep 2027",
        "current_phase": "System Development and Demonstration",
        "current_milestone": "Preliminary Design Review (PDR) — Completed 15 Jan 2025",
        "risk_register_summary": [
            {"risk_id": "R-001", "title": "Hardware supplier lead time", "likelihood": "Medium", "consequence": "High", "mitigation": "Dual-source qualification in progress"},
            {"risk_id": "R-002", "title": "Software integration complexity", "likelihood": "High", "consequence": "Medium", "mitigation": "Additional sprint capacity allocated Q2"}
        ],
        "schedule_status": "On schedule. Next milestone: CDR planned 15 Jun 2025.",
        "cost_status": "At EAC. BCWP: $4.2M, ACWP: $4.1M, CPI: 1.02",
        "top_issues": "Supplier for Component X has a 16-week lead time vs. 12-week plan. Mitigation in progress."
    }

    draft = generate_cdrl(
        did_number="DI-MGMT-81650",
        contract_number="FA8621-24-C-0042",
        program_name="Advanced Sensor Integration System (ASIS)",
        period="01 Oct 2024 – 30 Sep 2027",
        program_data=program_data,
        reporting_period="Q1 FY2025"
    )

    with open("RMP_Draft_Q1FY25.md", "w") as f:
        f.write(draft)
    print("Draft RMP saved.")

# Pulls EVM and financial data from Deltek Costpoint reporting API

# costpoint_integration.py
# Pulls EVM and financial data from Deltek Costpoint reporting API

import requests
import os
from datetime import datetime

COSTPOINT_URL = os.environ["COSTPOINT_URL"]
COSTPOINT_USER = os.environ["COSTPOINT_USER"]
COSTPOINT_PASS = os.environ["COSTPOINT_PASS"]

def get_evm_data(project_id: str, period: str) -> dict:
    """Retrieve EVM performance data from Costpoint for a given project and period."""
    # Costpoint REST API (version varies by installation — adjust endpoint paths)
    auth = (COSTPOINT_USER, COSTPOINT_PASS)
    headers = {"Accept": "application/json", "Content-Type": "application/json"}

    # Get EVM summary for project
    evm_url = f"{COSTPOINT_URL}/api/v1/projects/{project_id}/evm/summary"
    params = {"period": period}  # e.g., "2025-01"

    resp = requests.get(evm_url, auth=auth, headers=headers,
                        params=params, verify=True)
    resp.raise_for_status()
    evm_data = resp.json()

    # Structure data for CDRL generation
    return {
        "bcws": evm_data.get("budgetedCostWorkScheduled"),
        "bcwp": evm_data.get("budgetedCostWorkPerformed"),
        "acwp": evm_data.get("actualCostWorkPerformed"),
        "bac": evm_data.get("budgetAtCompletion"),
        "eac": evm_data.get("estimateAtCompletion"),
        "cpi": evm_data.get("costPerformanceIndex"),
        "spi": evm_data.get("schedulePerformanceIndex"),
        "cv": evm_data.get("costVariance"),
        "sv": evm_data.get("scheduleVariance"),
        "period": period
    }

def get_milestone_status(project_id: str) -> list:
    """Retrieve milestone status from Costpoint project schedule."""
    milestones_url = f"{COSTPOINT_URL}/api/v1/projects/{project_id}/milestones"
    auth = (COSTPOINT_USER, COSTPOINT_PASS)
    resp = requests.get(milestones_url, auth=auth, verify=True)
    resp.raise_for_status()

    milestones = []
    for m in resp.json().get("milestones", []):
        milestones.append({
            "name": m.get("name"),
            "planned_date": m.get("plannedDate"),
            "actual_date": m.get("actualDate"),
            "status": "Complete" if m.get("actualDate") else "Planned",
            "variance_days": m.get("varianceDays", 0)
        })
    return milestones

# omb_budget_generator.py

OMB_EXHIBIT_PROMPTS = {
    "Exhibit_53": """Generate an Exhibit 53 (IT Investment Portfolio Summary) narrative
for the following IT investment. Follow OMB Circular A-11 Section 53 format exactly.

INVESTMENT DATA:
{investment_data}

Generate:
1. Investment Summary (250 words max)
2. Business Case (500 words max) — quantify mission benefit and cost avoidance
3. Performance Goals — 3 SMART goals with baseline, target, and measurement method
4. Risk Summary — top 3 risks with likelihood, impact, and mitigation
5. Alignment Statement — tie to Agency Strategic Plan and Administration priorities

Flag any missing data with [DATA REQUIRED].
Mark output as DRAFT requiring CFO and CIO review before OMB submission.""",

    "CBJ_Narrative": """Generate a Congressional Budget Justification (CBJ) Program Activity
narrative for the following program. This will be submitted to Congress as part of the
President's Budget Request. Writing must be clear, factual, and defensible under
Congressional scrutiny.

PROGRAM DATA:
{program_data}

BUDGET YEAR: {budget_year}
PRIOR YEAR ENACTED: {py_amount}
CURRENT YEAR ESTIMATE: {cy_amount}
BUDGET YEAR REQUEST: {by_amount}

Generate:
1. Program Overview (300 words) — what the program does, who it serves, why it matters
2. Budget Year Request Justification — explain increases/decreases vs. prior year with specific rationale
3. Performance Summary — 3 key performance indicators with FY{prev_year} actuals and FY{budget_year} targets
4. Key Changes This Budget Year — bulleted list of major programmatic changes
5. Risks of Reduced Funding — factual statement of consequences if Congress does not appropriate requested amount

RULES:
- Do not make political arguments — factual, mission-focused only
- All funding figures must match the data provided — do not round or estimate
- CBJ language must survive Freedom of Information Act release — write accordingly
- Mark as DRAFT requiring PFM review, agency counsel review, and OMB passback"""
}

def generate_omb_document(
    exhibit_type: str,
    program_data: dict,
    budget_year: int
) -> str:
    prompt_template = OMB_EXHIBIT_PROMPTS.get(exhibit_type)
    if not prompt_template:
        raise ValueError(f"Unknown exhibit type: {exhibit_type}")

    formatted_prompt = prompt_template.format(
        program_data=json.dumps(program_data, indent=2),
        investment_data=json.dumps(program_data, indent=2),
        budget_year=budget_year,
        prev_year=budget_year - 1,
        py_amount=program_data.get("prior_year_enacted", "[REQUIRED]"),
        cy_amount=program_data.get("current_year_estimate", "[REQUIRED]"),
        by_amount=program_data.get("budget_year_request", "[REQUIRED]")
    )

    response = client.chat.completions.create(
        model=os.environ["AZURE_OPENAI_DEPLOYMENT"],
        messages=[
            {"role": "system", "content": "You are a federal budget analyst with expertise in OMB Circular A-11 and Congressional budget justification formats. Generate precise, defensible budget narrative."},
            {"role": "user", "content": formatted_prompt}
        ],
        temperature=0.1,
        max_tokens=3000
    )

    return f"[DRAFT — AI GENERATED — REQUIRES CFO REVIEW AND OMB CLEARANCE]\n\n{response.choices[0].message.content}"

{
  "did_number": "[DID]",
  "program_data": [assembled data],
  "period": "[current period]"
}

SYSTEM PROMPT:
You are a government CDRL acceptance reviewer. Check the following draft CDRL
against the applicable DID requirements and identify any deficiencies that
would cause the government customer to reject or return the deliverable.

DID REQUIREMENTS:
{did_sections_and_requirements}

CHECK FOR:
1. SECTION COMPLETENESS — is every required DID section present and substantively addressed?
2. DATA ACCURACY FLAGS — any [DATA REQUIRED] placeholders not yet filled in
3. FORMAT COMPLIANCE — correct document identification header, revision table, approval block
4. CLASSIFICATION MARKINGS — correct classification/CUI markings on every page
5. INTERNAL CONSISTENCY — do dates, values, and statuses align across sections?
6. GOVERNMENT COMMENT INCORPORATION — if prior review comments provided, are they addressed?

OUTPUT:
- Overall readiness: READY FOR PM REVIEW / NEEDS REWORK
- Deficiencies list (numbered, each with severity: Critical/Major/Minor)
- Estimated rework time
- Specific recommended edits for each deficiency

DRAFT CDRL:
{cdrl_draft}