diff --git a/chainforge/react-server/src/App.js b/chainforge/react-server/src/App.tsx
similarity index 93%
rename from chainforge/react-server/src/App.js
rename to chainforge/react-server/src/App.tsx
index 3f1e872..c08c5ed 100644
--- a/chainforge/react-server/src/App.js
+++ b/chainforge/react-server/src/App.tsx
@@ -1,5 +1,5 @@
 import React, { useState, useCallback, useRef, useEffect } from "react";
-import ReactFlow, { Controls, Background } from "reactflow";
+import ReactFlow, { Controls, Background, ReactFlowInstance } from "reactflow";
 import {
   Button,
   Menu,
@@ -31,7 +31,7 @@ import CodeEvaluatorNode from "./CodeEvaluatorNode";
 import VisNode from "./VisNode";
 import InspectNode from "./InspectorNode";
 import ScriptNode from "./ScriptNode";
-import AlertModal from "./AlertModal";
+import AlertModal, { AlertModalHandles } from "./AlertModal";
 import ItemsNode from "./ItemsNode";
 import TabularDataNode from "./TabularDataNode";
 import JoinNode from "./JoinNode";
@@ -56,7 +56,7 @@ import "./text-fields-node.css"; // project
 
 // State management (from https://reactflow.dev/docs/guides/state-management/)
 import { shallow } from "zustand/shallow";
-import useStore from "./store";
+import useStore, { StoreHandles } from "./store";
 import fetch_from_backend from "./fetch_from_backend";
 import StorageCache from "./backend/cache";
 import { APP_IS_RUNNING_LOCALLY, browserTabIsActive } from "./backend/utils";
@@ -69,19 +69,20 @@ import {
   isEdgeChromium,
   isChromium,
 } from "react-device-detect";
+import { Dict, LLMSpec } from "./backend/typing";
 const IS_ACCEPTED_BROWSER =
   (isChrome ||
     isChromium ||
     isEdgeChromium ||
     isFirefox ||
-    navigator?.brave !== undefined) &&
+    (navigator as any)?.brave !== undefined) &&
   !isMobile;
 
 // Whether we are running on localhost or not, and hence whether
 // we have access to the Flask backend for, e.g., Python code evaluation.
 const IS_RUNNING_LOCALLY = APP_IS_RUNNING_LOCALLY();
 
-const selector = (state) => ({
+const selector = (state: StoreHandles) => ({
   nodes: state.nodes,
   edges: state.edges,
   onNodesChange: state.onNodesChange,
@@ -108,7 +109,7 @@ const INITIAL_LLM = () => {
       temp: 1.0,
       settings: getDefaultModelSettings("hf"),
       formData: getDefaultModelFormData("hf"),
-    };
+    } satisfies LLMSpec;
     falcon7b.formData.shortname = falcon7b.name;
     falcon7b.formData.model = falcon7b.model;
     return falcon7b;
@@ -123,7 +124,7 @@ const INITIAL_LLM = () => {
       temp: 1.0,
       settings: getDefaultModelSettings("gpt-3.5-turbo"),
       formData: getDefaultModelFormData("gpt-3.5-turbo"),
-    };
+    } satisfies LLMSpec;
     chatgpt.formData.shortname = chatgpt.name;
     chatgpt.formData.model = chatgpt.model;
     return chatgpt;
@@ -173,7 +174,13 @@ const getSharedFlowURLParam = () => {
   return undefined;
 };
 
-const MenuTooltip = ({ label, children }) => {
+const MenuTooltip = ({
+  label,
+  children,
+}: {
+  label: string;
+  children: React.ReactNode;
+}) => {
   return (
     <Tooltip
       label={label}
@@ -208,7 +215,7 @@ const App = () => {
   } = useStore(selector, shallow);
 
   // For saving / loading
-  const [rfInstance, setRfInstance] = useState(null);
+  const [rfInstance, setRfInstance] = useState<ReactFlowInstance | null>(null);
   const [autosavingInterval, setAutosavingInterval] = useState(null);
 
   // For 'share' button
@@ -236,7 +243,7 @@ const App = () => {
   const { hideContextMenu } = useContextMenu();
 
   // For displaying error messages to user
-  const alertModal = useRef(null);
+  const alertModal = useRef<AlertModalHandles>(null);
 
   // For displaying a pending 'loading' status
   const [isLoading, setIsLoading] = useState(true);
@@ -252,6 +259,7 @@ const App = () => {
   };
   const getViewportCenter = () => {
     const { centerX, centerY } = getWindowCenter();
+    if (rfInstance === null) return { x: centerX, y: centerY };
     // Support Zoom
     const { x, y, zoom } = rfInstance.getViewport();
     return { x: -(x / zoom) + centerX / zoom, y: -(y / zoom) + centerY / zoom };
@@ -293,7 +301,7 @@ const App = () => {
       position: { x: x - 200, y: y - 100 },
     });
   };
-  const addEvalNode = (progLang) => {
+  const addEvalNode = (progLang: string) => {
     const { x, y } = getViewportCenter();
     let code = "";
     if (progLang === "python")
@@ -388,7 +396,7 @@ const App = () => {
       position: { x: x - 200, y: y - 100 },
     });
   };
-  const addProcessorNode = (progLang) => {
+  const addProcessorNode = (progLang: string) => {
     const { x, y } = getViewportCenter();
     let code = "";
     if (progLang === "python")
@@ -410,11 +418,12 @@ const App = () => {
     if (settingsModal && settingsModal.current) settingsModal.current.trigger();
   };
 
-  const handleError = (err) => {
+  const handleError = (err: Error | string) => {
+    const msg = typeof err === "string" ? err : err.message;
     setIsLoading(false);
     setWaitingForShare(false);
-    if (alertModal.current) alertModal.current.trigger(err.message);
-    console.error(err.message);
+    if (alertModal.current) alertModal.current.trigger(msg);
+    console.error(msg);
   };
 
   /**
@@ -468,7 +477,7 @@ const App = () => {
   const resetFlow = useCallback(() => {
     resetLLMColors();
 
-    const uid = (id) => `${id}-${Date.now()}`;
+    const uid = (id: string) => `${id}-${Date.now()}`;
     const starting_nodes = [
       {
         id: uid("prompt"),
@@ -493,39 +502,38 @@ const App = () => {
     if (rfInstance) rfInstance.setViewport({ x: 200, y: 80, zoom: 1 });
   }, [setNodes, setEdges, resetLLMColors, rfInstance]);
 
-  const loadFlow = async (flow, rf_inst) => {
-    if (flow) {
-      if (rf_inst) {
-        if (flow.viewport)
-          rf_inst.setViewport({
-            x: flow.viewport.x || 0,
-            y: flow.viewport.y || 0,
-            zoom: flow.viewport.zoom || 1,
-          });
-        else rf_inst.setViewport({ x: 0, y: 0, zoom: 1 });
-      }
-      resetLLMColors();
-
-      // First, clear the ReactFlow state entirely
-      // NOTE: We need to do this so it forgets any node/edge ids, which might have cross-over in the loaded flow.
-      setNodes([]);
-      setEdges([]);
-
-      // After a delay, load in the new state.
-      setTimeout(() => {
-        setNodes(flow.nodes || []);
-        setEdges(flow.edges || []);
-
-        // Save flow that user loaded to autosave cache, in case they refresh the browser
-        StorageCache.saveToLocalStorage("chainforge-flow", flow);
-
-        // Cancel loading spinner
-        setIsLoading(false);
-      }, 10);
-
-      // Start auto-saving, if it's not already enabled
-      if (rf_inst) initAutosaving(rf_inst);
+  const loadFlow = async (flow?: Dict, rf_inst?: ReactFlowInstance | null) => {
+    if (flow === undefined) return;
+    if (rf_inst) {
+      if (flow.viewport)
+        rf_inst.setViewport({
+          x: flow.viewport.x || 0,
+          y: flow.viewport.y || 0,
+          zoom: flow.viewport.zoom || 1,
+        });
+      else rf_inst.setViewport({ x: 0, y: 0, zoom: 1 });
     }
+    resetLLMColors();
+
+    // First, clear the ReactFlow state entirely
+    // NOTE: We need to do this so it forgets any node/edge ids, which might have cross-over in the loaded flow.
+    setNodes([]);
+    setEdges([]);
+
+    // After a delay, load in the new state.
+    setTimeout(() => {
+      setNodes(flow.nodes || []);
+      setEdges(flow.edges || []);
+
+      // Save flow that user loaded to autosave cache, in case they refresh the browser
+      StorageCache.saveToLocalStorage("chainforge-flow", flow);
+
+      // Cancel loading spinner
+      setIsLoading(false);
+    }, 10);
+
+    // Start auto-saving, if it's not already enabled
+    if (rf_inst) initAutosaving(rf_inst);
   };
 
   const importGlobalStateFromCache = useCallback(() => {
@@ -535,7 +543,7 @@ const App = () => {
   const autosavedFlowExists = () => {
     return window.localStorage.getItem("chainforge-flow") !== null;
   };
-  const loadFlowFromAutosave = async (rf_inst) => {
+  const loadFlowFromAutosave = async (rf_inst: ReactFlowInstance) => {
     const saved_flow = StorageCache.loadFromLocalStorage(
       "chainforge-flow",
       false,
@@ -602,8 +610,8 @@ const App = () => {
   );
 
   const importFlowFromJSON = useCallback(
-    (flowJSON, rf_inst) => {
-      const rf = rf_inst || rfInstance;
+    (flowJSON: Dict, rf_inst?: ReactFlowInstance | null) => {
+      const rf = rf_inst ?? rfInstance;
 
       setIsLoading(true);
 
@@ -656,13 +664,16 @@ const App = () => {
       // Handle file load event
       reader.addEventListener("load", function () {
         try {
+          if (typeof reader.result !== "string")
+            throw new Error("File could not be read: Unknown format or empty.");
+
           // We try to parse the JSON response
           const flow_and_cache = JSON.parse(reader.result);
 
           // Import it to React Flow and import cache data on the backend
           importFlowFromJSON(flow_and_cache);
         } catch (error) {
-          handleError(error);
+          handleError(error as Error);
         }
       });
 
@@ -675,7 +686,7 @@ const App = () => {
   };
 
   // Downloads the selected OpenAI eval file (preconverted to a .cforge flow)
-  const importFlowFromOpenAIEval = (evalname) => {
+  const importFlowFromOpenAIEval = (evalname: string) => {
     setIsLoading(true);
 
     fetch_from_backend(
@@ -707,7 +718,7 @@ const App = () => {
   };
 
   // Load flow from examples modal
-  const onSelectExampleFlow = (name, example_category) => {
+  const onSelectExampleFlow = (name: string, example_category: string) => {
     // Trigger the 'loading' modal
     setIsLoading(true);
 
@@ -772,9 +783,9 @@ const App = () => {
     }
 
     // Helper function
-    function isFileSizeLessThan5MB(str) {
+    function isFileSizeLessThan5MB(json_str: string) {
       const encoder = new TextEncoder();
-      const encodedString = encoder.encode(str);
+      const encodedString = encoder.encode(json_str);
       const fileSizeInBytes = encodedString.length;
       const fileSizeInMB = fileSizeInBytes / (1024 * 1024); // Convert bytes to megabytes
       return fileSizeInMB < 5;
@@ -894,7 +905,7 @@ const App = () => {
   };
 
   // Run once upon ReactFlow initialization
-  const onInit = (rf_inst) => {
+  const onInit = (rf_inst: ReactFlowInstance) => {
     setRfInstance(rf_inst);
 
     if (IS_RUNNING_LOCALLY) {
diff --git a/chainforge/react-server/src/ModelSettingSchemas.js b/chainforge/react-server/src/ModelSettingSchemas.tsx
similarity index 95%
rename from chainforge/react-server/src/ModelSettingSchemas.js
rename to chainforge/react-server/src/ModelSettingSchemas.tsx
index 5f2d062..a78612c 100644
--- a/chainforge/react-server/src/ModelSettingSchemas.js
+++ b/chainforge/react-server/src/ModelSettingSchemas.tsx
@@ -10,7 +10,14 @@
  * Descriptions of OpenAI model parameters copied from OpenAI's official chat completions documentation: https://platform.openai.com/docs/models/model-endpoint-compatibility
  */
 
-import { LLMProvider, RATE_LIMITS, getProvider } from "./backend/models";
+import { LLM, LLMProvider, RATE_LIMITS, getProvider } from "./backend/models";
+import {
+  Dict,
+  JSONCompatible,
+  CustomLLMProviderSpec,
+  ModelSettingsDict,
+  LLMSpec,
+} from "./backend/typing";
 import { transformDict } from "./backend/utils";
 import useStore from "./store";
 
@@ -21,9 +28,9 @@ const UI_SUBMIT_BUTTON_SPEC = {
   },
   norender: false,
   submitText: "Submit",
-};
+} satisfies Dict;
 
-const ChatGPTSettings = {
+const ChatGPTSettings: ModelSettingsDict = {
   fullName: "GPT-3.5+ (OpenAI)",
   schema: {
     type: "object",
@@ -224,29 +231,32 @@ const ChatGPTSettings = {
   },
 
   postprocessors: {
-    functions: (str) => {
+    functions: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       if (str.trim().length === 0) return [];
       return JSON.parse(str); // parse the JSON schema
     },
-    function_call: (str) => {
+    function_call: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       const s = str.trim();
       if (s.length === 0) return "";
       if (s === "auto" || s === "none") return s;
       else return { name: s };
     },
-    stop: (str) => {
+    stop: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       if (str.trim().length === 0) return [];
       return str
         .match(/"((?:[^"\\]|\\.)*)"/g)
-        .map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
+        ?.map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
     },
-    response_format: (str) => {
+    response_format: (str: string | number | boolean) => {
       return { type: str };
     },
   },
 };
 
-const GPT4Settings = {
+const GPT4Settings: ModelSettingsDict = {
   fullName: ChatGPTSettings.fullName,
   schema: {
     type: "object",
@@ -275,7 +285,7 @@ const GPT4Settings = {
   postprocessors: ChatGPTSettings.postprocessors,
 };
 
-const ClaudeSettings = {
+const ClaudeSettings: ModelSettingsDict = {
   fullName: "Claude (Anthropic)",
   schema: {
     type: "object",
@@ -420,16 +430,17 @@ const ClaudeSettings = {
   },
 
   postprocessors: {
-    stop_sequences: (str) => {
+    stop_sequences: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       if (str.trim().length === 0) return ["\n\nHuman:"];
       return str
         .match(/"((?:[^"\\]|\\.)*)"/g)
-        .map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
+        ?.map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
     },
   },
 };
 
-const PaLM2Settings = {
+const PaLM2Settings: ModelSettingsDict = {
   fullName: "Google AI Models (Gemini & PaLM)",
   schema: {
     type: "object",
@@ -531,16 +542,17 @@ const PaLM2Settings = {
   },
 
   postprocessors: {
-    stop_sequences: (str) => {
+    stop_sequences: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       if (str.trim().length === 0) return [];
       return str
         .match(/"((?:[^"\\]|\\.)*)"/g)
-        .map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
+        ?.map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
     },
   },
 };
 
-const DalaiModelSettings = {
+const DalaiModelSettings: ModelSettingsDict = {
   fullName: "Dalai-hosted local model (Alpaca, Llama)",
   schema: {
     type: "object",
@@ -691,7 +703,7 @@ const DalaiModelSettings = {
   postprocessors: {},
 };
 
-const AzureOpenAISettings = {
+const AzureOpenAISettings: ModelSettingsDict = {
   fullName: "Azure OpenAI Model",
   schema: {
     type: "object",
@@ -738,7 +750,7 @@ const AzureOpenAISettings = {
   postprocessors: ChatGPTSettings.postprocessors,
 };
 
-const HuggingFaceTextInferenceSettings = {
+const HuggingFaceTextInferenceSettings: ModelSettingsDict = {
   fullName: "HuggingFace-hosted text generation models",
   schema: {
     type: "object",
@@ -910,7 +922,7 @@ const HuggingFaceTextInferenceSettings = {
   postprocessors: {},
 };
 
-const AlephAlphaLuminousSettings = {
+const AlephAlphaLuminousSettings: ModelSettingsDict = {
   fullName: "Aleph Alpha Luminous",
   schema: {
     type: "object",
@@ -1111,22 +1123,25 @@ const AlephAlphaLuminousSettings = {
     },
   },
   postprocessors: {
-    stop_sequences: (str) => {
+    stop_sequences: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       if (str.trim().length === 0) return [];
       return str
         .match(/"((?:[^"\\]|\\.)*)"/g)
-        .map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
+        ?.map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
     },
-    log_probs: (bool) => {
+    log_probs: (bool: boolean | number | string) => {
+      if (typeof bool !== "boolean") return bool;
       return bool ? 3 : null;
     },
-    best_of: (a) => {
+    best_of: (a: number | string | boolean) => {
+      if (typeof a !== "number") return a;
       return a === 1 ? null : a;
     },
   },
 };
 
-const OllamaSettings = {
+const OllamaSettings: ModelSettingsDict = {
   fullName: "Ollama",
   schema: {
     type: "object",
@@ -1234,11 +1249,12 @@ const OllamaSettings = {
     },
   },
   postprocessors: {
-    stop_sequences: (str) => {
+    stop_sequences: (str: string | number | boolean) => {
+      if (typeof str !== "string") return str;
       if (str.trim().length === 0) return [];
       return str
         .match(/"((?:[^"\\]|\\.)*)"/g)
-        .map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
+        ?.map((s) => s.substring(1, s.length - 1)); // split on double-quotes but exclude escaped double-quotes inside the group
     },
   },
 };
@@ -1929,7 +1945,7 @@ const MetaLlama2ChatSettings = {
 };
 
 // A lookup table indexed by base_model.
-export const ModelSettings = {
+export const ModelSettings: Dict<ModelSettingsDict> = {
   "gpt-3.5-turbo": ChatGPTSettings,
   "gpt-4": GPT4Settings,
   "claude-v1": ClaudeSettings,
@@ -1948,7 +1964,9 @@ export const ModelSettings = {
   "br.meta.llama2": MetaLlama2ChatSettings,
 };
 
-export function getSettingsSchemaForLLM(llm_name) {
+export function getSettingsSchemaForLLM(
+  llm_name: string,
+): ModelSettingsDict | undefined {
   const llm_provider = getProvider(llm_name);
 
   const provider_to_settings_schema = {
@@ -1963,13 +1981,13 @@ export function getSettingsSchemaForLLM(llm_name) {
   };
 
   if (llm_provider === LLMProvider.Custom) return ModelSettings[llm_name];
-  else if (llm_provider in provider_to_settings_schema)
+  else if (llm_provider && llm_provider in provider_to_settings_schema)
     return provider_to_settings_schema[llm_provider];
   else if (llm_provider === LLMProvider.Bedrock) {
     return ModelSettings[llm_name.split("-")[0]];
   } else {
     console.error(`Could not find provider for llm ${llm_name}`);
-    return {};
+    return undefined;
   }
 }
 
@@ -1978,7 +1996,10 @@ export function getSettingsSchemaForLLM(llm_name) {
  * @param {*} settings_dict A dict of form setting_name: value (string: string)
  * @param {*} llm A string of the name of the model to query.
  */
-export function typecastSettingsDict(settings_dict, llm) {
+export function typecastSettingsDict(
+  settings_dict: ModelSettingsDict,
+  llm: string,
+) {
   const settings = getSettingsSchemaForLLM(llm);
   const schema = settings?.schema?.properties ?? {};
   const postprocessors = settings?.postprocessors ?? {};
@@ -2014,16 +2035,16 @@ export function typecastSettingsDict(settings_dict, llm) {
  * @param {*} settings_schema
  */
 export const setCustomProvider = (
-  name,
-  emoji,
-  models,
-  rate_limit,
-  settings_schema,
+  name: string,
+  emoji: string,
+  models?: string[],
+  rate_limit?: number,
+  settings_schema?: CustomLLMProviderSpec["settings_schema"],
 ) => {
   if (typeof emoji === "string" && (emoji.length === 0 || emoji.length > 2))
     throw new Error(`Emoji for a custom provider must have a character.`);
 
-  const new_provider = { name };
+  const new_provider: Dict<JSONCompatible> = { name };
   new_provider.emoji = emoji || "✨";
 
   // Each LLM *model* must have a unique name. To avoid name collisions, for custom providers,
@@ -2036,7 +2057,7 @@ export const setCustomProvider = (
     (Array.isArray(models) && models.length > 0 ? `${models[0]}` : "");
 
   // Build the settings form schema for this new custom provider
-  const compiled_schema = {
+  const compiled_schema: ModelSettingsDict = {
     fullName: `${name} (custom provider)`,
     schema: {
       type: "object",
@@ -2057,6 +2078,7 @@ export const setCustomProvider = (
         "ui:autofocus": true,
       },
     },
+    postprocessors: {},
   };
 
   // Add a models selector if there's multiple models
@@ -2093,8 +2115,9 @@ export const setCustomProvider = (
   // Add the built provider and its settings to the global lookups:
   const AvailableLLMs = useStore.getState().AvailableLLMs;
   const prev_provider_idx = AvailableLLMs.findIndex((d) => d.name === name);
-  if (prev_provider_idx > -1) AvailableLLMs[prev_provider_idx] = new_provider;
-  else AvailableLLMs.push(new_provider);
+  if (prev_provider_idx > -1)
+    AvailableLLMs[prev_provider_idx] = new_provider as LLMSpec;
+  else AvailableLLMs.push(new_provider as LLMSpec);
   ModelSettings[base_model] = compiled_schema;
 
   // Add rate limit info, if specified
@@ -2113,7 +2136,7 @@ export const setCustomProvider = (
   useStore.getState().setAvailableLLMs(AvailableLLMs);
 };
 
-export const setCustomProviders = (providers) => {
+export const setCustomProviders = (providers: CustomLLMProviderSpec[]) => {
   for (const p of providers)
     setCustomProvider(
       p.name,
@@ -2124,7 +2147,7 @@ export const setCustomProviders = (providers) => {
     );
 };
 
-export const getTemperatureSpecForModel = (modelName) => {
+export const getTemperatureSpecForModel = (modelName: string) => {
   if (modelName in ModelSettings) {
     const temperature_property =
       ModelSettings[modelName].schema?.properties?.temperature;
@@ -2139,11 +2162,14 @@ export const getTemperatureSpecForModel = (modelName) => {
   return null;
 };
 
-export const postProcessFormData = (settingsSpec, formData) => {
+export const postProcessFormData = (
+  settingsSpec: ModelSettingsDict,
+  formData: Dict<JSONCompatible>,
+) => {
   // Strip all 'model' and 'shortname' props in the submitted form, as these are passed elsewhere or unecessary for the backend
   const skip_keys = { model: true, shortname: true };
 
-  const new_data = {};
+  const new_data: Dict<JSONCompatible> = {};
   const postprocessors = settingsSpec?.postprocessors
     ? settingsSpec.postprocessors
     : {};
@@ -2151,28 +2177,32 @@ export const postProcessFormData = (settingsSpec, formData) => {
   Object.keys(formData).forEach((key) => {
     if (key in skip_keys) return;
     if (key in postprocessors)
-      new_data[key] = postprocessors[key](formData[key]);
+      new_data[key] = postprocessors[key](
+        formData[key] as string | number | boolean,
+      );
     else new_data[key] = formData[key];
   });
 
   return new_data;
 };
 
-export const getDefaultModelFormData = (settingsSpec) => {
+export const getDefaultModelFormData = (
+  settingsSpec: string | ModelSettingsDict,
+) => {
   if (typeof settingsSpec === "string")
     settingsSpec = ModelSettings[settingsSpec];
-  const default_formdata = {};
+  const default_formdata: Dict<JSONCompatible> = {};
   const schema = settingsSpec.schema;
   Object.keys(schema.properties).forEach((key) => {
     default_formdata[key] =
       "default" in schema.properties[key]
         ? schema.properties[key].default
-        : undefined;
+        : null;
   });
   return default_formdata;
 };
 
-export const getDefaultModelSettings = (modelName) => {
+export const getDefaultModelSettings = (modelName: string) => {
   if (!(modelName in ModelSettings)) {
     console.warn(
       `Model ${modelName} not found in list of available model settings.`,
diff --git a/chainforge/react-server/src/backend/query.ts b/chainforge/react-server/src/backend/query.ts
index 9bbeddb..1174bf2 100644
--- a/chainforge/react-server/src/backend/query.ts
+++ b/chainforge/react-server/src/backend/query.ts
@@ -6,6 +6,7 @@ import {
   RawLLMResponseObject,
   isEqualChatHistory,
   ChatHistoryInfo,
+  ModelSettingsDict,
 } from "./typing";
 import {
   extract_responses,
@@ -288,7 +289,13 @@ export class PromptPipeline {
               num_queries_sent,
               max_req,
               wait_secs,
-              { ...llm_params, ...typecastSettingsDict(settings_params, llm) },
+              {
+                ...llm_params,
+                ...typecastSettingsDict(
+                  settings_params as ModelSettingsDict,
+                  llm,
+                ),
+              },
               chat_history,
               should_cancel,
             ),
@@ -305,7 +312,13 @@ export class PromptPipeline {
             undefined,
             undefined,
             undefined,
-            { ...llm_params, ...typecastSettingsDict(settings_params, llm) },
+            {
+              ...llm_params,
+              ...typecastSettingsDict(
+                settings_params as ModelSettingsDict,
+                llm,
+              ),
+            },
             chat_history,
             should_cancel,
           );
diff --git a/chainforge/react-server/src/backend/typing.ts b/chainforge/react-server/src/backend/typing.ts
index e6da410..de15253 100644
--- a/chainforge/react-server/src/backend/typing.ts
+++ b/chainforge/react-server/src/backend/typing.ts
@@ -16,6 +16,13 @@ export interface Dict<T = any> {
 // Function types
 export type Func<T = void> = (...args: any[]) => T;
 
+// JSON-compatible leaf types
+export type JSONLeaf = string | number | boolean | null;
+export type JSONCompatible<T = JSONLeaf> =
+  | T
+  | JSONCompatible<T>[]
+  | Dict<JSONCompatible<T>>;
+
 /** OpenAI function call format */
 export interface OpenAIFunctionCall {
   name: string;
@@ -110,6 +117,30 @@ export type LLMSpec = {
   progress?: Dict<number>; // only used for front-end to display progress collecting responses for this LLM
 };
 
+/** A spec for a user-defined custom LLM provider */
+export type CustomLLMProviderSpec = {
+  name: string;
+  emoji: string;
+  models?: string[];
+  rate_limit?: number;
+  settings_schema?: {
+    settings: Dict<Dict<JSONCompatible>>;
+    ui: Dict<Dict<JSONCompatible>>;
+  };
+};
+
+/** Internal description of model settings, passed to react-json-schema */
+export interface ModelSettingsDict {
+  fullName: string;
+  schema: {
+    type: "object";
+    required: string[];
+    properties: Dict<Dict<JSONCompatible>>;
+  };
+  uiSchema: Dict<JSONCompatible>;
+  postprocessors: Dict<(val: string | number | boolean) => any>;
+}
+
 export type ResponseUID = string;
 
 /** Standard properties that every LLM response object must have. */
diff --git a/chainforge/react-server/src/store.tsx b/chainforge/react-server/src/store.tsx
index bdbb93c..b3d6152 100644
--- a/chainforge/react-server/src/store.tsx
+++ b/chainforge/react-server/src/store.tsx
@@ -22,8 +22,9 @@ import {
   PromptVarType,
   PromptVarsDict,
   TemplateVarInfo,
+  TabularDataColType,
+  TabularDataRowType,
 } from "./backend/typing";
-import { TabularDataColType, TabularDataRowType } from "./TabularDataNode";
 
 // Initial project settings
 const initialAPIKeys = {};
@@ -227,7 +228,7 @@ export const initLLMProviders = initLLMProviderMenu
   .map((item) => ("group" in item && "items" in item ? item.items : item))
   .flat();
 
-interface StoreHandles {
+export interface StoreHandles {
   // Nodes and edges
   nodes: Node[];
   edges: Edge[];