Update plumber_hf.R
Browse files- plumber_hf.R +176 -176
plumber_hf.R
CHANGED
|
@@ -1,176 +1,176 @@
|
|
| 1 |
-
library(dampack)
|
| 2 |
-
library(readr)
|
| 3 |
-
library(assertthat)
|
| 4 |
-
library(jsonlite)
|
| 5 |
-
|
| 6 |
-
#* @apiTitle DARTH Health Economic Model API
|
| 7 |
-
#*
|
| 8 |
-
#* @apiDescription This API runs a health economic Markov model (Sick-Sicker)
|
| 9 |
-
#* for probabilistic sensitivity analysis. Hosted on Hugging Face Spaces.
|
| 10 |
-
#* Users send their own parameter CSV data from their local machine.
|
| 11 |
-
#* Supports multiple CSV inputs (e.g., base parameters + clinical inputs) that get merged.
|
| 12 |
-
|
| 13 |
-
#* Welcome page - API info
|
| 14 |
-
#* @get /
|
| 15 |
-
function() {
|
| 16 |
-
list(
|
| 17 |
-
name = "DARTH Health Economic Model API",
|
| 18 |
-
version = "1.0",
|
| 19 |
-
status = "running",
|
| 20 |
-
endpoints = list(
|
| 21 |
-
"GET /" = "This welcome page",
|
| 22 |
-
"GET /health" = "Health check",
|
| 23 |
-
"POST /runDARTHmodel" = "Run the DARTH model with your CSV data"
|
| 24 |
-
)
|
| 25 |
-
)
|
| 26 |
-
}
|
| 27 |
-
|
| 28 |
-
#* Health check endpoint
|
| 29 |
-
#* @get /health
|
| 30 |
-
function() {
|
| 31 |
-
list(status = "healthy", timestamp = Sys.time())
|
| 32 |
-
}
|
| 33 |
-
|
| 34 |
-
#* Run the DARTH model
|
| 35 |
-
#* @serializer csv
|
| 36 |
-
#* @param psa_inputs_json JSON string containing the base parameter distributions data from client's local CSV
|
| 37 |
-
#* @param clinical_inputs_json JSON string containing clinical input parameters (optional, merged with psa_inputs)
|
| 38 |
-
#* @param param_updates JSON string with optional parameter overrides (applied after merging all inputs)
|
| 39 |
-
#* @post /runDARTHmodel
|
| 40 |
-
function(psa_inputs_json = NULL, clinical_inputs_json = NULL, param_updates = NULL) {
|
| 41 |
-
|
| 42 |
-
# Source the model functions from Hugging Face (local) or SharePoint
|
| 43 |
-
# Local path = inside Docker container on Hugging Face after deployment
|
| 44 |
-
model_functions_path <- "/app/model/darth_funcs.R"
|
| 45 |
-
|
| 46 |
-
# SharePoint URL for model code (EXPERIMENT - will show roadblocks)
|
| 47 |
-
# Original sharing link: https://symmetronltd.sharepoint.com/:u:/s/EJP_Sandbox/IQAjXsgb1eL1Q5T2uBMT2-VRAQ8wm3t8w4HJqUlf11C2N74?e=rTpGVD
|
| 48 |
-
sharepoint_url <- "https://symmetronltd.sharepoint.com/
|
| 49 |
-
|
| 50 |
-
model_loaded <- FALSE
|
| 51 |
-
|
| 52 |
-
# Option 1: Try local file on Hugging Face first (fastest, most reliable)
|
| 53 |
-
if (file.exists(model_functions_path)) {
|
| 54 |
-
source(model_functions_path)
|
| 55 |
-
model_loaded <- TRUE
|
| 56 |
-
message("Model loaded from Hugging Face local file")
|
| 57 |
-
}
|
| 58 |
-
|
| 59 |
-
# Option 2: Try SharePoint (ROADBLOCK - requires authentication!)
|
| 60 |
-
if (!model_loaded) {
|
| 61 |
-
tryCatch({
|
| 62 |
-
# This will FAIL because:
|
| 63 |
-
# 1. SharePoint requires Microsoft 365 authentication
|
| 64 |
-
# 2. source() cannot pass authentication headers
|
| 65 |
-
# 3. The URL redirects to a login page, not the raw R file
|
| 66 |
-
source(sharepoint_url)
|
| 67 |
-
model_loaded <- TRUE
|
| 68 |
-
message("Model loaded from SharePoint")
|
| 69 |
-
}, error = function(e) {
|
| 70 |
-
message("SharePoint source failed: ", e$message)
|
| 71 |
-
message("ROADBLOCK: SharePoint requires Azure AD authentication")
|
| 72 |
-
})
|
| 73 |
-
}
|
| 74 |
-
|
| 75 |
-
# No fallback - fail if model cannot be loaded
|
| 76 |
-
if (!model_loaded) {
|
| 77 |
-
stop("ERROR: Could not load model. Neither local file nor SharePoint source available.")
|
| 78 |
-
}
|
| 79 |
-
|
| 80 |
-
# Helper function to parse and validate CSV JSON input
|
| 81 |
-
parse_csv_json <- function(json_input, input_name) {
|
| 82 |
-
if (is.null(json_input) || json_input == "") {
|
| 83 |
-
return(NULL)
|
| 84 |
-
}
|
| 85 |
-
|
| 86 |
-
tryCatch({
|
| 87 |
-
df <- jsonlite::fromJSON(json_input)
|
| 88 |
-
df <- as.data.frame(df)
|
| 89 |
-
|
| 90 |
-
# Validate required columns exist
|
| 91 |
-
required_cols <- c("parameter", "distribution", "v1", "v2")
|
| 92 |
-
if (!all(required_cols %in% names(df))) {
|
| 93 |
-
stop(paste("Error:", input_name, "must contain columns:", paste(required_cols, collapse = ", ")))
|
| 94 |
-
}
|
| 95 |
-
|
| 96 |
-
# Ensure v1 and v2 are numeric (handle NULL/NA values)
|
| 97 |
-
df$v1 <- as.numeric(df$v1)
|
| 98 |
-
df$v2 <- as.numeric(df$v2)
|
| 99 |
-
|
| 100 |
-
# Replace NA with 0 for v2 (needed for 'fixed' distributions)
|
| 101 |
-
df$v2[is.na(df$v2)] <- 0
|
| 102 |
-
|
| 103 |
-
return(df)
|
| 104 |
-
}, error = function(e) {
|
| 105 |
-
stop(paste("Error parsing", input_name, "JSON:", e$message))
|
| 106 |
-
})
|
| 107 |
-
}
|
| 108 |
-
|
| 109 |
-
# Parse the base PSA inputs (required)
|
| 110 |
-
if (is.null(psa_inputs_json) || psa_inputs_json == "") {
|
| 111 |
-
stop("Error: psa_inputs_json is required. Please send your local CSV data.")
|
| 112 |
-
}
|
| 113 |
-
|
| 114 |
-
psa_inputs <- parse_csv_json(psa_inputs_json, "psa_inputs_json")
|
| 115 |
-
|
| 116 |
-
# Parse clinical inputs if provided (optional, merged with psa_inputs)
|
| 117 |
-
clinical_inputs <- parse_csv_json(clinical_inputs_json, "clinical_inputs_json")
|
| 118 |
-
|
| 119 |
-
if (!is.null(clinical_inputs)) {
|
| 120 |
-
# Merge clinical inputs with base inputs (clinical inputs are appended)
|
| 121 |
-
# If duplicate parameters exist, clinical inputs take precedence
|
| 122 |
-
existing_params <- psa_inputs$parameter
|
| 123 |
-
new_params <- clinical_inputs$parameter
|
| 124 |
-
|
| 125 |
-
# Remove any duplicate parameters from psa_inputs (clinical takes precedence)
|
| 126 |
-
duplicates <- existing_params %in% new_params
|
| 127 |
-
if (any(duplicates)) {
|
| 128 |
-
psa_inputs <- psa_inputs[!duplicates, ]
|
| 129 |
-
}
|
| 130 |
-
|
| 131 |
-
# Combine the data frames
|
| 132 |
-
psa_inputs <- rbind(psa_inputs, clinical_inputs)
|
| 133 |
-
}
|
| 134 |
-
|
| 135 |
-
# Parse param_updates if provided (for additional overrides)
|
| 136 |
-
if (!is.null(param_updates) && param_updates != "") {
|
| 137 |
-
# Handle both JSON string and data frame input
|
| 138 |
-
if (is.character(param_updates)) {
|
| 139 |
-
param_updates <- jsonlite::fromJSON(param_updates)
|
| 140 |
-
}
|
| 141 |
-
|
| 142 |
-
# Convert to data frame if needed
|
| 143 |
-
if (!is.data.frame(param_updates)) {
|
| 144 |
-
param_updates <- as.data.frame(param_updates)
|
| 145 |
-
}
|
| 146 |
-
|
| 147 |
-
# For each row of the data-frame containing the variables to be changed...
|
| 148 |
-
for(n in 1:nrow(param_updates)) {
|
| 149 |
-
# Update parameters from API input
|
| 150 |
-
psa_inputs <- overwrite_parameter_value(
|
| 151 |
-
existing_df = psa_inputs,
|
| 152 |
-
parameter = param_updates[n, "parameter"],
|
| 153 |
-
distribution = param_updates[n, "distribution"],
|
| 154 |
-
v1 = param_updates[n, "v1"],
|
| 155 |
-
v2 = param_updates[n, "v2"]
|
| 156 |
-
)
|
| 157 |
-
}
|
| 158 |
-
}
|
| 159 |
-
|
| 160 |
-
# Run the model using the single run-model function
|
| 161 |
-
results <- run_model(psa_inputs)
|
| 162 |
-
|
| 163 |
-
# Check that the model results being returned are the correct dimensions
|
| 164 |
-
assertthat::assert_that(
|
| 165 |
-
all(dim(x = results) == c(1000, 6)),
|
| 166 |
-
class(results) == "data.frame",
|
| 167 |
-
msg = "Dimensions or type of data are incorrect, please check the model code."
|
| 168 |
-
)
|
| 169 |
-
|
| 170 |
-
# Check that no sensitive data is included in the output
|
| 171 |
-
assertthat::assert_that(
|
| 172 |
-
all(psa_inputs[, 1] %in% as.character(unlist(x = results, recursive = TRUE)) == FALSE)
|
| 173 |
-
)
|
| 174 |
-
|
| 175 |
-
return(results)
|
| 176 |
-
}
|
|
|
|
| 1 |
+
library(dampack)
|
| 2 |
+
library(readr)
|
| 3 |
+
library(assertthat)
|
| 4 |
+
library(jsonlite)
|
| 5 |
+
|
| 6 |
+
#* @apiTitle DARTH Health Economic Model API
|
| 7 |
+
#*
|
| 8 |
+
#* @apiDescription This API runs a health economic Markov model (Sick-Sicker)
|
| 9 |
+
#* for probabilistic sensitivity analysis. Hosted on Hugging Face Spaces.
|
| 10 |
+
#* Users send their own parameter CSV data from their local machine.
|
| 11 |
+
#* Supports multiple CSV inputs (e.g., base parameters + clinical inputs) that get merged.
|
| 12 |
+
|
| 13 |
+
#* Welcome page - API info
|
| 14 |
+
#* @get /
|
| 15 |
+
function() {
|
| 16 |
+
list(
|
| 17 |
+
name = "DARTH Health Economic Model API",
|
| 18 |
+
version = "1.0",
|
| 19 |
+
status = "running",
|
| 20 |
+
endpoints = list(
|
| 21 |
+
"GET /" = "This welcome page",
|
| 22 |
+
"GET /health" = "Health check",
|
| 23 |
+
"POST /runDARTHmodel" = "Run the DARTH model with your CSV data"
|
| 24 |
+
)
|
| 25 |
+
)
|
| 26 |
+
}
|
| 27 |
+
|
| 28 |
+
#* Health check endpoint
|
| 29 |
+
#* @get /health
|
| 30 |
+
function() {
|
| 31 |
+
list(status = "healthy", timestamp = Sys.time())
|
| 32 |
+
}
|
| 33 |
+
|
| 34 |
+
#* Run the DARTH model
|
| 35 |
+
#* @serializer csv
|
| 36 |
+
#* @param psa_inputs_json JSON string containing the base parameter distributions data from client's local CSV
|
| 37 |
+
#* @param clinical_inputs_json JSON string containing clinical input parameters (optional, merged with psa_inputs)
|
| 38 |
+
#* @param param_updates JSON string with optional parameter overrides (applied after merging all inputs)
|
| 39 |
+
#* @post /runDARTHmodel
|
| 40 |
+
function(psa_inputs_json = NULL, clinical_inputs_json = NULL, param_updates = NULL) {
|
| 41 |
+
|
| 42 |
+
# Source the model functions from Hugging Face (local) or SharePoint
|
| 43 |
+
# Local path = inside Docker container on Hugging Face after deployment
|
| 44 |
+
model_functions_path <- "/app/model/darth_funcs.R"
|
| 45 |
+
|
| 46 |
+
# SharePoint URL for model code (EXPERIMENT - will show roadblocks)
|
| 47 |
+
# Original sharing link: https://symmetronltd.sharepoint.com/:u:/s/EJP_Sandbox/IQAjXsgb1eL1Q5T2uBMT2-VRAQ8wm3t8w4HJqUlf11C2N74?e=rTpGVD
|
| 48 |
+
sharepoint_url <- "https://symmetronltd.sharepoint.com/sites/EJP_Sandbox/Shared%20Documents/API_Site2/model/darth_funcs.R"
|
| 49 |
+
|
| 50 |
+
model_loaded <- FALSE
|
| 51 |
+
|
| 52 |
+
# Option 1: Try local file on Hugging Face first (fastest, most reliable)
|
| 53 |
+
if (file.exists(model_functions_path)) {
|
| 54 |
+
source(model_functions_path)
|
| 55 |
+
model_loaded <- TRUE
|
| 56 |
+
message("Model loaded from Hugging Face local file")
|
| 57 |
+
}
|
| 58 |
+
|
| 59 |
+
# Option 2: Try SharePoint (ROADBLOCK - requires authentication!)
|
| 60 |
+
if (!model_loaded) {
|
| 61 |
+
tryCatch({
|
| 62 |
+
# This will FAIL because:
|
| 63 |
+
# 1. SharePoint requires Microsoft 365 authentication
|
| 64 |
+
# 2. source() cannot pass authentication headers
|
| 65 |
+
# 3. The URL redirects to a login page, not the raw R file
|
| 66 |
+
source(sharepoint_url)
|
| 67 |
+
model_loaded <- TRUE
|
| 68 |
+
message("Model loaded from SharePoint")
|
| 69 |
+
}, error = function(e) {
|
| 70 |
+
message("SharePoint source failed: ", e$message)
|
| 71 |
+
message("ROADBLOCK: SharePoint requires Azure AD authentication")
|
| 72 |
+
})
|
| 73 |
+
}
|
| 74 |
+
|
| 75 |
+
# No fallback - fail if model cannot be loaded
|
| 76 |
+
if (!model_loaded) {
|
| 77 |
+
stop("ERROR: Could not load model. Neither local file nor SharePoint source available.")
|
| 78 |
+
}
|
| 79 |
+
|
| 80 |
+
# Helper function to parse and validate CSV JSON input
|
| 81 |
+
parse_csv_json <- function(json_input, input_name) {
|
| 82 |
+
if (is.null(json_input) || json_input == "") {
|
| 83 |
+
return(NULL)
|
| 84 |
+
}
|
| 85 |
+
|
| 86 |
+
tryCatch({
|
| 87 |
+
df <- jsonlite::fromJSON(json_input)
|
| 88 |
+
df <- as.data.frame(df)
|
| 89 |
+
|
| 90 |
+
# Validate required columns exist
|
| 91 |
+
required_cols <- c("parameter", "distribution", "v1", "v2")
|
| 92 |
+
if (!all(required_cols %in% names(df))) {
|
| 93 |
+
stop(paste("Error:", input_name, "must contain columns:", paste(required_cols, collapse = ", ")))
|
| 94 |
+
}
|
| 95 |
+
|
| 96 |
+
# Ensure v1 and v2 are numeric (handle NULL/NA values)
|
| 97 |
+
df$v1 <- as.numeric(df$v1)
|
| 98 |
+
df$v2 <- as.numeric(df$v2)
|
| 99 |
+
|
| 100 |
+
# Replace NA with 0 for v2 (needed for 'fixed' distributions)
|
| 101 |
+
df$v2[is.na(df$v2)] <- 0
|
| 102 |
+
|
| 103 |
+
return(df)
|
| 104 |
+
}, error = function(e) {
|
| 105 |
+
stop(paste("Error parsing", input_name, "JSON:", e$message))
|
| 106 |
+
})
|
| 107 |
+
}
|
| 108 |
+
|
| 109 |
+
# Parse the base PSA inputs (required)
|
| 110 |
+
if (is.null(psa_inputs_json) || psa_inputs_json == "") {
|
| 111 |
+
stop("Error: psa_inputs_json is required. Please send your local CSV data.")
|
| 112 |
+
}
|
| 113 |
+
|
| 114 |
+
psa_inputs <- parse_csv_json(psa_inputs_json, "psa_inputs_json")
|
| 115 |
+
|
| 116 |
+
# Parse clinical inputs if provided (optional, merged with psa_inputs)
|
| 117 |
+
clinical_inputs <- parse_csv_json(clinical_inputs_json, "clinical_inputs_json")
|
| 118 |
+
|
| 119 |
+
if (!is.null(clinical_inputs)) {
|
| 120 |
+
# Merge clinical inputs with base inputs (clinical inputs are appended)
|
| 121 |
+
# If duplicate parameters exist, clinical inputs take precedence
|
| 122 |
+
existing_params <- psa_inputs$parameter
|
| 123 |
+
new_params <- clinical_inputs$parameter
|
| 124 |
+
|
| 125 |
+
# Remove any duplicate parameters from psa_inputs (clinical takes precedence)
|
| 126 |
+
duplicates <- existing_params %in% new_params
|
| 127 |
+
if (any(duplicates)) {
|
| 128 |
+
psa_inputs <- psa_inputs[!duplicates, ]
|
| 129 |
+
}
|
| 130 |
+
|
| 131 |
+
# Combine the data frames
|
| 132 |
+
psa_inputs <- rbind(psa_inputs, clinical_inputs)
|
| 133 |
+
}
|
| 134 |
+
|
| 135 |
+
# Parse param_updates if provided (for additional overrides)
|
| 136 |
+
if (!is.null(param_updates) && param_updates != "") {
|
| 137 |
+
# Handle both JSON string and data frame input
|
| 138 |
+
if (is.character(param_updates)) {
|
| 139 |
+
param_updates <- jsonlite::fromJSON(param_updates)
|
| 140 |
+
}
|
| 141 |
+
|
| 142 |
+
# Convert to data frame if needed
|
| 143 |
+
if (!is.data.frame(param_updates)) {
|
| 144 |
+
param_updates <- as.data.frame(param_updates)
|
| 145 |
+
}
|
| 146 |
+
|
| 147 |
+
# For each row of the data-frame containing the variables to be changed...
|
| 148 |
+
for(n in 1:nrow(param_updates)) {
|
| 149 |
+
# Update parameters from API input
|
| 150 |
+
psa_inputs <- overwrite_parameter_value(
|
| 151 |
+
existing_df = psa_inputs,
|
| 152 |
+
parameter = param_updates[n, "parameter"],
|
| 153 |
+
distribution = param_updates[n, "distribution"],
|
| 154 |
+
v1 = param_updates[n, "v1"],
|
| 155 |
+
v2 = param_updates[n, "v2"]
|
| 156 |
+
)
|
| 157 |
+
}
|
| 158 |
+
}
|
| 159 |
+
|
| 160 |
+
# Run the model using the single run-model function
|
| 161 |
+
results <- run_model(psa_inputs)
|
| 162 |
+
|
| 163 |
+
# Check that the model results being returned are the correct dimensions
|
| 164 |
+
assertthat::assert_that(
|
| 165 |
+
all(dim(x = results) == c(1000, 6)),
|
| 166 |
+
class(results) == "data.frame",
|
| 167 |
+
msg = "Dimensions or type of data are incorrect, please check the model code."
|
| 168 |
+
)
|
| 169 |
+
|
| 170 |
+
# Check that no sensitive data is included in the output
|
| 171 |
+
assertthat::assert_that(
|
| 172 |
+
all(psa_inputs[, 1] %in% as.character(unlist(x = results, recursive = TRUE)) == FALSE)
|
| 173 |
+
)
|
| 174 |
+
|
| 175 |
+
return(results)
|
| 176 |
+
}
|