Underwriter's Co-Pilot

[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_152":153,"_154":155},"routes/_app.x.$handle.$name",{"_11":12,"_13":14,"_15":16,"_18":19,"_17":-7,"_115":-5,"_20":116,"_22":117,"_118":119,"_120":121,"_25":122,"_24":-5,"_123":-5,"_124":-5,"_125":53,"_126":127,"_128":44,"_129":-5,"_130":-5,"_131":132},"routes/_app.x.$handle.$name.branch.$branch.code.$",{"_11":12,"_13":14,"_15":16,"_17":-7,"_18":19,"_20":21,"_22":23,"_24":-7,"_25":26,"_27":-5,"_28":-5,"_29":-5,"_30":-5,"_31":-5},"handle","legal","name","underwriting","branchName","main","version","path","","project",{"_33":61,"_13":14,"_45":62,"_63":-5,"_64":53,"_65":66,"_67":66,"_68":-5,"_69":-5,"_70":71,"_72":37,"_73":44,"_74":44,"_75":-5,"_76":77,"_78":79,"_80":-5,"_81":-5,"_82":-5,"_83":-5,"_84":85,"_86":85,"_87":88,"_89":88,"_90":88,"_91":85,"_92":85,"_93":94,"_95":96,"_97":98,"_99":53},"tree",[56],"file","contents",[32],"upstream","pull","ahead","behind","envVars",{"_33":34,"_13":35,"_36":37,"_38":-5,"_39":40,"_41":42,"_43":44,"_45":46,"_17":47,"_48":49,"_50":51,"_52":53,"_54":-5,"_18":55},"id","a7612a46-48f1-11f0-8aae-76b3cceeab13","main.tsx","fileId","65f93bb2-48f0-11f0-8aa6-76b3cceeab13","parentId","branchId","65f754d2-48f0-11f0-8aa6-76b3cceeab13","branchVersion",6,"latest",true,"createdAt",["D",1749886318086],5,"fileType","http","content","// --- [SYSTEM PROMPT] Underwriter Task Suggestion ---\nconst underwriterTaskSuggestionSystemPromptTemplate = `\nYou are an AI assistant for a professional loan underwriter. Your goal is to suggest a comprehensive list of essential underwriting tasks based on the provided loan document package text. These tasks should cover the core pillars of underwriting: Capital, Capacity, Collateral, and Credit. Return your suggestions STRICTLY as a JSON object. The JSON object's keys are category names. Values are arrays of objects, each with \"task\" (a string describing the task) and \"is_high_priority\" (boolean). Foundational tasks that are critical for an initial assessment should be marked as \"is_high_priority: true\".\n\nExample Categories & Tasks:\n- \"Borrower Profile & Capacity\": \"Calculate the Debt-to-Income (DTI) ratio for all borrowers.\", \"Verify employment history and income stability.\", \"Identify all sources of qualifying income.\"\n- \"Asset & Capital Analysis\": \"Verify and sum all liquid assets for down payment and reserves.\", \"Identify any large, unusual, or unsourced deposits in bank statements.\", \"Confirm the source of down payment funds.\"\n- \"Credit & Liability Review\": \"Identify all listed liabilities and compare against the credit report.\", \"Flag any undisclosed debts or liabilities found in bank statements.\", \"Analyze the credit report for derogatory events (late payments, collections, bankruptcies).\"\n- \"Collateral & Loan Terms\": \"Extract the property address and appraised value.\", \"Identify the loan type (e.g., Conventional, FHA, VA), term, and interest rate.\", \"Verify the Loan-to-Value (LTV) ratio.\"\n- \"Red Flag & Compliance Checks\": \"Scan for inconsistencies across all provided documents (e.g., names, addresses, income figures).\", \"Identify potential signs of fraud or misrepresentation.\", \"Check for compliance with basic TRID (TILA-RESPA Integrated Disclosure) requirements if a Loan Estimate is present.\"\n\nDOCUMENT TEXT:\n==============\n%%DOCUMENT_TEXT%%\n`;\n\n// --- [SYSTEM PROMPT] Underwriter AI Analysis ---\nconst underwriterAIAnalysisSystemPromptTemplate = `\nYou are a precision AI system for Loan Underwriting Analysis. Your purpose is to execute the user-specified underwriting task(s) on the provided loan document text. Your output MUST be a single JSON object (no markdown formatting) that strictly follows the schema below. When identifying specific items, use a corresponding \"finding_type\" (e.g., \"DTI Calculation\", \"Income Verification\", \"Asset Analysis\", \"Credit Anomaly\", \"Collateral Review\", \"Red Flag\"). For \"Red Flag\" or \"Guideline Compliance\" findings, provide a clear explanation and a recommended action for the underwriter.\n\nLOAN ANALYSIS TASK(S) / QUERY: %%UNDERWRITER_TASK_QUERY%%\nDOCUMENT SOURCE(S): %%INPUT_SOURCE_DESCRIPTION%%\n\nAnalyze the document text provided. Meticulously map ALL findings to the JSON schema. Present information quantitatively and objectively. If the text is insufficient to perform a task (e.g., a missing page from a bank statement), note this clearly in the 'limitations_or_warnings' field.\n\nOutput JSON Schema:\n{\n \"request_details\": { \"operation_performed\": \"string\", \"input_source_description\": \"string\" },\n \"loan_summary\": {\n \"borrowers\": \"string[]\",\n \"property_address\": \"string\",\n \"loan_type\": \"string\",\n \"loan_amount\": \"number\",\n \"appraised_value\": \"number\",\n \"ltv_ratio\": \"number\",\n \"dti_ratio\": \"number\",\n \"total_qualifying_income\": \"number\",\n \"total_verified_assets\": \"number\"\n },\n \"detailed_findings\": [\n {\n \"finding_id\": \"string\",\n \"finding_type\": \"string\",\n \"description\": \"string\",\n \"severity_or_importance\": \"string\",\n \"relevant_input_excerpt\": \"string\",\n \"page_or_section_reference\": \"string\",\n \"implications\": \"string\",\n \"underwriter_action_needed\": \"string\",\n \"confidence_score\": \"float\"\n }\n ],\n \"execution_metadata\": {\n \"status\": \"string\",\n \"timestamp\": \"string\",\n \"limitations_or_warnings\": \"string\",\n \"disclaimer\": \"string\"\n }\n}\n`;\n\n// --- [SYSTEM PROMPT] Loan Scenario & Guideline Co-Pilot ---\nconst loanScenarioCoPilotSystemPromptTemplate = `\nYou are an expert AI Loan Underwriting Strategist. You have been provided with a structured JSON object containing the results of a preliminary analysis of a loan file and a knowledge base of the lender's current guidelines. Your goal is to act as a co-pilot for the human underwriter. You must:\n1. Analyze the provided JSON, focusing on the 'loan_summary' and 'detailed_findings'.\n2. Cross-reference these findings against the provided lending guidelines in the KNOWLEDGE BASE.\n3. Identify any findings that represent a \"deal-breaker\" or \"guideline exception\".\n4. For each issue, propose a concrete, actionable remediation step for the underwriter to request.\n5. If remediation is not straightforward, propose at least one viable alternative loan scenario or product that the borrower might qualify for.\n6. Return your response STRICTLY as a JSON object following the schema below. If there are no compliance issues, state this clearly in the 'overall_assessment'.\n\nANALYSIS RESULTS JSON:\n=======================\n%%ANALYSIS_JSON%%\n\nKNOWLEDGE BASE (Lender's Custom Guidelines):\n=============================================\n%%GUIDELINES%%\n\nOutput JSON Schema:\n{\n \"co_pilot_assessment\": {\n \"overall_assessment\": \"string\",\n \"compliance_issues\": [\n {\n \"issue_id\": \"string\",\n \"issue_description\": \"string\",\n \"guideline_reference\": \"string\",\n \"related_finding_ids\": [\"string\"]\n }\n ],\n \"recommended_remediations\": [\n {\n \"remediation_id\": \"string\",\n \"description\": \"string\",\n \"impact_simulation\": \"string\",\n \"required_documentation\": \"string\"\n }\n ],\n \"alternative_scenarios\": [\n {\n \"scenario_id\": \"string\",\n \"scenario_name\": \"string\",\n \"justification\": \"string\",\n \"next_steps\": \"string\"\n }\n ]\n }\n}\n`;\n\n// --- Type Definitions ---\ninterface LogEntry {\n agent: string;\n message: string | object;\n type: \"step\" | \"result\" | \"error\" | \"input\" | \"final\" | \"info\";\n}\ninterface UnderwriterAIResponse {\n request_details: { operation_performed: string; input_source_description: string };\n loan_summary: {\n borrowers: string[];\n property_address: string;\n loan_type: string;\n loan_amount: number;\n appraised_value: number;\n ltv_ratio: number;\n dti_ratio: number;\n total_qualifying_income: number;\n total_verified_assets: number;\n };\n detailed_findings: Array<{\n finding_id: string;\n finding_type: string;\n description: string;\n severity_or_importance?: string;\n relevant_input_excerpt?: string;\n page_or_section_reference?: string;\n implications?: string;\n underwriter_action_needed?: string;\n confidence_score?: number;\n }>;\n execution_metadata: { status: string; timestamp: string; limitations_or_warnings: string; disclaimer: string };\n}\ninterface CoPilotResponse {\n co_pilot_assessment: {\n overall_assessment: string;\n compliance_issues: any[];\n recommended_remediations: any[];\n alternative_scenarios: any[];\n };\n}\ninterface UnderwritingAnalysis {\n id: string;\n docId: string;\n docName?: string;\n query: string;\n result: UnderwriterAIResponse;\n coPilotResult?: CoPilotResponse;\n datePerformed: string;\n type: \"analysis\";\n}\ninterface GroupedSuggestions {\n [category: string]: { task: string; is_high_priority: boolean }[];\n}\n\nfunction generateHtmlShell(initialQuery, sourceUrl) {\n const escapedQuery = initialQuery.replace(//g, \">\");\n return `\n\n\n\n\n\nUnderwriter's Co-Pilot\n\n\n\n\n

\n\n

\n${\n [\"dashboard\", \"new-analysis\", \"history\", \"settings\", \"analysis-detail\"].map(v =>\n `

`\n ).join(\"\")\n }\n

\n\n\n`;\n}\n\nexport default async function(req: Request) {\n const { OpenAI } = await import(\"https://esm.town/v/std/openai\");\n const { PDFExtract } = await import(\"npm:pdf.js-extract\");\n\n const CORS_HEADERS = {\n \"Access-Control-Allow-Origin\": \"*\",\n \"Access-Control-Allow-Methods\": \"POST, GET, OPTIONS\",\n \"Access-Control-Allow-Headers\": \"Content-Type\",\n };\n if (req.method === \"OPTIONS\") return new Response(null, { status: 204, headers: CORS_HEADERS });\n\n const url = new URL(req.url);\n const format = url.searchParams.get(\"format\");\n const action = url.searchParams.get(\"action\");\n const sourceUrl = import.meta.url.replace(\"esm.town\", \"val.town\");\n const openai = new OpenAI();\n const MAX_TEXT_SUGGEST = 20000;\n const MAX_TEXT_ANALYZE = 30000;\n\n async function callAI(\n systemPrompt: string,\n userMessage: string,\n log: LogEntry[],\n agentName: string,\n ): Promise

Code