Anesthesia Co-Pilot

Welcome, Doctor.

Click \"New Command\", use the 🧙‍♂️ Wizard, or press Ctrl + K.

Example: preop for Josef Smith 75yo 80kg ASA 3 with CAD for CABG

\n \n \n \n \n\n \n `;\n};\n\n// --- CORE BACKEND LOGIC ---\nconst llm = async (sysPrompt, userPrompt, log, tid, model = \"gpt-4o\") => {\n log(\"DEBUG\", \"LLM\", `Calling OpenAI for TID ${tid}`);\n try {\n const oa = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });\n const completion = await oa.chat.completions.create({\n model,\n messages: [{ role: \"system\", content: sysPrompt }, { role: \"user\", content: userPrompt }],\n response_format: { type: \"json_object\" },\n temperature: 0.1,\n });\n const content = completion.choices[0]?.message?.content;\n if (!content) throw new Error(\"OpenAI returned no content.\");\n return JSON.parse(content);\n } catch (err) {\n log(\"ERROR\", \"LLM\", `OpenAI API call failed for TID ${tid}`, { error: err.message });\n throw new Error(`AI model error: ${err.message}`);\n }\n};\n\nconst helpData = {\n \"disclaimer\":\n \"AI generated for informational purposes only. This is not medical advice. Verify all information and exercise independent clinical judgment. Adhere to all institutional protocols and standards of care.\",\n \"title\": \"Anesthesia Co-Pilot Help\",\n \"description\":\n \"This tool uses AI to provide clinical decision support. Enter a command in natural language. The Co-Pilot maintains a 'patient context' for your current session. Start by loading a patient with a 'preop' command.\",\n \"commands\": [\n {\n \"name\": \"Pre-Op Assessment\",\n \"usage\": \"Loads patient data and provides a full pre-operative workup.\",\n \"example\": \"preop for 80kg 65yo M with HTN for a CABG\",\n },\n {\n \"name\": \"Anesthetic Plan\",\n \"usage\": \"Generates a phased anesthetic plan for the active patient.\",\n \"example\": \"plan for GA with difficult airway risk\",\n },\n {\n \"name\": \"Drug Information\",\n \"usage\": \"Provides detailed info for a drug, tailored to the active patient.\",\n \"example\": \"drug info for rocuronium\",\n },\n {\n \"name\": \"Differential Diagnosis\",\n \"usage\": \"Initiates an interactive diagnostic pathway for a clinical sign.\",\n \"example\": \"diff for intraoperative hypotension\",\n },\n {\n \"name\": \"Crisis Protocol\",\n \"usage\": \"IMMEDIATE access to critical event management protocols.\",\n \"example\": \"crisis malignant hyperthermia\",\n },\n {\n \"name\": \"Checklist Generation\",\n \"usage\": \"Creates an interactive safety checklist.\",\n \"example\": \"checklist for central line insertion\",\n },\n {\n \"name\": \"Guideline Summary\",\n \"usage\": \"Summarizes key points from major clinical guidelines.\",\n \"example\": \"guideline on ASA fasting\",\n },\n ],\n};\n\n// --- MAIN HTTP HANDLER ---\nexport default async function(req: Request): Promise {\n const reqId = `req-${Date.now()}`;\n const logger = new Logger(reqId);\n const headers = { \"Content-Type\": \"application/json\" };\n const url = new URL(req.url);\n\n if (req.method === \"GET\" && !url.searchParams.has(\"endpoint\")) {\n return new Response(generateHtml(req.url), { headers: { \"Content-Type\": \"text/html\" } });\n }\n\n if (req.method !== \"POST\") {\n return new Response(JSON.stringify({ error: \"Method Not Allowed\" }), { status: 405, headers });\n }\n\n try {\n const body = await req.json();\n const { endpoint } = body;\n\n if (endpoint === \"full-process\") {\n const { command, patientContext } = body;\n const parsedData = await llm(COMMAND_PARSER_PROMPT, command, logger.log.bind(logger), \"parser-task\");\n const task = parsedData.task;\n\n if (task === \"HELP\") {\n return new Response(JSON.stringify({ task, result: HelpOutSchema.parse(helpData), patientContext }), {\n status: 200,\n headers,\n });\n }\n\n const agentPayload = {\n task,\n data: parsedData.data,\n patientContext: { ...patientContext, ...parsedData.data },\n conversationState: null,\n };\n UserInputSchema.parse(agentPayload);\n\n const agentPrompt = getAgentPrompt(agentPayload, `task-${Date.now()}`);\n const aiResult = await llm(agentPrompt.sys, agentPrompt.user, logger.log.bind(logger), `task-${Date.now()}`);\n\n const validatedResult = await AllOutputSchemas[task].parseAsync(aiResult);\n const updatedPatientContext = task === \"PREOP\" ? agentPayload.patientContext : patientContext;\n\n return new Response(JSON.stringify({ task, result: validatedResult, patientContext: updatedPatientContext }), {\n status: 200,\n headers,\n });\n }\n\n if (endpoint === \"follow-up\") {\n const { task, patientContext, conversationState } = body;\n const agentPayload = { task, data: {}, patientContext, conversationState };\n UserInputSchema.parse(agentPayload);\n\n const agentPrompt = getAgentPrompt(agentPayload, `followup-${Date.now()}`);\n const aiResult = await llm(agentPrompt.sys, agentPrompt.user, logger.log.bind(logger), `followup-${Date.now()}`);\n\n const validatedResult = await AllOutputSchemas[task].parseAsync(aiResult);\n\n return new Response(JSON.stringify({ task, result: validatedResult }), { status: 200, headers });\n }\n\n return new Response(JSON.stringify({ error: \"Invalid endpoint\" }), { status: 404, headers });\n } catch (err) {\n const errorMsg = (err instanceof z.ZodError)\n ? `Validation Error: ${err.issues.map(i => `'${i.path.join(\".\")}' - ${i.message}`).join(\", \")}`\n : err.message;\n logger.log(\"ERROR\", \"Handler\", \"Request failed\", { error: errorMsg, stack: err.stack });\n return new Response(JSON.stringify({ error: errorMsg }), { status: 500, headers });\n }\n}","deleted","label",[98],"permalink","https://www.val.town/x/svc/med/code/main.tsx","extension","actionData","errors"]