Add 4 more models

app/src/components/CopiableCode.tsx

@@ -1,9 +1,9 @@
-import { HStack, Icon, IconButton, Tooltip, Text } from "@chakra-ui/react";
+import { HStack, Icon, IconButton, Tooltip, Text, type StackProps } from "@chakra-ui/react";
 import { useState } from "react";
 import { MdContentCopy } from "react-icons/md";
 import { useHandledAsyncCallback } from "~/utils/hooks";
 
-const CopiableCode = ({ code }: { code: string }) => {
+const CopiableCode = ({ code, ...rest }: { code: string } & StackProps) => {
   const [copied, setCopied] = useState(false);
 
   const [copyToClipboard] = useHandledAsyncCallback(async () => {
@@ -18,8 +18,16 @@ const CopiableCode = ({ code }: { code: string }) => {
       padding={3}
       w="full"
       justifyContent="space-between"
+      alignItems="flex-start"
+      {...rest}
     >
-      <Text fontFamily="inconsolata" fontWeight="bold" letterSpacing={0.5} overflowX="auto">
+      <Text
+        fontFamily="inconsolata"
+        fontWeight="bold"
+        letterSpacing={0.5}
+        overflowX="auto"
+        whiteSpace="pre-wrap"
+      >
         {code}
       </Text>
       <Tooltip closeOnClick={false} label={copied ? "Copied!" : "Copy to clipboard"}>

app/src/components/OutputsTable/OutputCell/PromptModal.tsx

@@ -5,30 +5,98 @@ import {
   ModalContent,
   ModalHeader,
   ModalOverlay,
+  VStack,
+  Text,
+  Box,
   type UseDisclosureReturn,
+  Link,
 } from "@chakra-ui/react";
-import { type RouterOutputs } from "~/utils/api";
+import { api, type RouterOutputs } from "~/utils/api";
 import { JSONTree } from "react-json-tree";
+import CopiableCode from "~/components/CopiableCode";
 
-export default function ExpandedModal(props: {
+const theme = {
+  scheme: "chalk",
+  author: "chris kempson (http://chriskempson.com)",
+  base00: "transparent",
+  base01: "#202020",
+  base02: "#303030",
+  base03: "#505050",
+  base04: "#b0b0b0",
+  base05: "#d0d0d0",
+  base06: "#e0e0e0",
+  base07: "#f5f5f5",
+  base08: "#fb9fb1",
+  base09: "#eda987",
+  base0A: "#ddb26f",
+  base0B: "#acc267",
+  base0C: "#12cfc0",
+  base0D: "#6fc2ef",
+  base0E: "#e1a3ee",
+  base0F: "#deaf8f",
+};
+
+export default function PromptModal(props: {
   cell: NonNullable<RouterOutputs["scenarioVariantCells"]["get"]>;
   disclosure: UseDisclosureReturn;
 }) {
+  const { data } = api.scenarioVariantCells.getTemplatedPromptMessage.useQuery({
+    cellId: props.cell.id,
+  });
+
   return (
-    <Modal isOpen={props.disclosure.isOpen} onClose={props.disclosure.onClose} size="2xl">
+    <Modal isOpen={props.disclosure.isOpen} onClose={props.disclosure.onClose} size="xl">
       <ModalOverlay />
       <ModalContent>
-        <ModalHeader>Prompt</ModalHeader>
+        <ModalHeader>Prompt info</ModalHeader>
         <ModalCloseButton />
         <ModalBody>
-          <JSONTree
+          <VStack py={4} w="">
-            data={props.cell.prompt}
+            <VStack w="full" alignItems="flex-start">
-            invertTheme={true}
+              <Text fontWeight="bold">Full Prompt</Text>
-            theme="chalk"
+              <Box
-            shouldExpandNodeInitially={() => true}
+                w="full"
-            getItemString={() => ""}
+                p={4}
-            hideRoot
+                alignItems="flex-start"
-          />
+                backgroundColor="blackAlpha.800"
+                borderRadius={4}
+              >
+                <JSONTree
+                  data={props.cell.prompt}
+                  theme={theme}
+                  shouldExpandNodeInitially={() => true}
+                  getItemString={() => ""}
+                  hideRoot
+                />
+              </Box>
+            </VStack>
+            {data?.templatedPrompt && (
+              <VStack w="full" mt={4} alignItems="flex-start">
+                <Text fontWeight="bold">Templated prompt message:</Text>
+                <CopiableCode
+                  w="full"
+                  // bgColor="gray.100"
+                  p={4}
+                  borderWidth={1}
+                  whiteSpace="pre-wrap"
+                  code={data.templatedPrompt}
+                />
+              </VStack>
+            )}
+            {data?.learnMoreUrl && (
+              <Link
+                href={data.learnMoreUrl}
+                isExternal
+                color="blue.500"
+                fontWeight="bold"
+                fontSize="sm"
+                mt={4}
+                alignSelf="flex-end"
+              >
+                Learn More
+              </Link>
+            )}
+          </VStack>
         </ModalBody>
       </ModalContent>
     </Modal>

app/src/components/OutputsTable/OutputCell/TopActions.tsx

@@ -1,7 +1,7 @@
 import { HStack, Icon, IconButton, Spinner, Tooltip, useDisclosure } from "@chakra-ui/react";
 import { BsArrowClockwise, BsInfoCircle } from "react-icons/bs";
 import { useExperimentAccess } from "~/utils/hooks";
-import ExpandedModal from "./PromptModal";
+import PromptModal from "./PromptModal";
 import { type RouterOutputs } from "~/utils/api";
 
 export const CellOptions = ({
@@ -32,7 +32,7 @@ export const CellOptions = ({
               variant="ghost"
             />
           </Tooltip>
-          <ExpandedModal cell={cell} disclosure={modalDisclosure} />
+          <PromptModal cell={cell} disclosure={modalDisclosure} />
         </>
       )}
       {canModify && (

app/src/modelProviders/openpipe-chat/frontend.ts

@@ -1,14 +1,20 @@
 import { type OpenpipeChatOutput, type SupportedModel } from ".";
 import { type FrontendModelProvider } from "../types";
 import { refinementActions } from "./refinementActions";
-import { templateOpenOrcaPrompt } from "./templatePrompt";
+import {
+  templateOpenOrcaPrompt,
+  templateAlpacaInstructPrompt,
+  templateSystemUserAssistantPrompt,
+  templateInstructionInputResponsePrompt,
+  templateAiroborosPrompt,
+} from "./templatePrompt";
 
 const frontendModelProvider: FrontendModelProvider<SupportedModel, OpenpipeChatOutput> = {
   name: "OpenAI ChatCompletion",
 
   models: {
     "Open-Orca/OpenOrcaxOpenChat-Preview2-13B": {
-      name: "OpenOrca-Platypus2-13B",
+      name: "OpenOrcaxOpenChat-Preview2-13B",
       contextWindow: 4096,
       pricePerSecond: 0.0003,
       speed: "medium",
@@ -16,6 +22,42 @@ const frontendModelProvider: FrontendModelProvider<SupportedModel, OpenpipeChatO
       learnMoreUrl: "https://huggingface.co/Open-Orca/OpenOrcaxOpenChat-Preview2-13B",
       templatePrompt: templateOpenOrcaPrompt,
     },
+    "Open-Orca/OpenOrca-Platypus2-13B": {
+      name: "OpenOrca-Platypus2-13B",
+      contextWindow: 4096,
+      pricePerSecond: 0.0003,
+      speed: "medium",
+      provider: "openpipe/Chat",
+      learnMoreUrl: "https://huggingface.co/Open-Orca/OpenOrca-Platypus2-13B",
+      templatePrompt: templateAlpacaInstructPrompt,
+    },
+    "stabilityai/StableBeluga-13B": {
+      name: "StableBeluga-13B",
+      contextWindow: 4096,
+      pricePerSecond: 0.0003,
+      speed: "medium",
+      provider: "openpipe/Chat",
+      learnMoreUrl: "https://huggingface.co/stabilityai/StableBeluga-13B",
+      templatePrompt: templateSystemUserAssistantPrompt,
+    },
+    "NousResearch/Nous-Hermes-Llama2-13b": {
+      name: "Nous-Hermes-Llama2-13b",
+      contextWindow: 4096,
+      pricePerSecond: 0.0003,
+      speed: "medium",
+      provider: "openpipe/Chat",
+      learnMoreUrl: "https://huggingface.co/NousResearch/Nous-Hermes-Llama2-13b",
+      templatePrompt: templateInstructionInputResponsePrompt,
+    },
+    "jondurbin/airoboros-l2-13b-gpt4-2.0": {
+      name: "airoboros-l2-13b-gpt4-2.0",
+      contextWindow: 4096,
+      pricePerSecond: 0.0003,
+      speed: "medium",
+      provider: "openpipe/Chat",
+      learnMoreUrl: "https://huggingface.co/jondurbin/airoboros-l2-13b-gpt4-2.0",
+      templatePrompt: templateAiroborosPrompt,
+    },
   },
 
   refinementActions,

app/src/modelProviders/openpipe-chat/getCompletion.ts

@@ -8,6 +8,10 @@ import frontendModelProvider from "./frontend";
 
 const modelEndpoints: Record<OpenpipeChatInput["model"], string> = {
   "Open-Orca/OpenOrcaxOpenChat-Preview2-13B": "https://5ef82gjxk8kdys-8000.proxy.runpod.net/v1",
+  "Open-Orca/OpenOrca-Platypus2-13B": "https://lt5qlel6qcji8t-8000.proxy.runpod.net/v1",
+  "stabilityai/StableBeluga-13B": "https://vcorl8mxni2ou1-8000.proxy.runpod.net/v1",
+  "NousResearch/Nous-Hermes-Llama2-13b": "https://ncv8pw3u0vb8j2-8000.proxy.runpod.net/v1",
+  "jondurbin/airoboros-l2-13b-gpt4-2.0": "https://9nrbx7oph4btou-8000.proxy.runpod.net/v1",
 };
 
 export async function getCompletion(

app/src/modelProviders/openpipe-chat/index.ts

@@ -4,7 +4,13 @@ import inputSchema from "./input.schema.json";
 import { getCompletion } from "./getCompletion";
 import frontendModelProvider from "./frontend";
 
-const supportedModels = ["Open-Orca/OpenOrcaxOpenChat-Preview2-13B"] as const;
+const supportedModels = [
+  "Open-Orca/OpenOrcaxOpenChat-Preview2-13B",
+  "Open-Orca/OpenOrca-Platypus2-13B",
+  "stabilityai/StableBeluga-13B",
+  "NousResearch/Nous-Hermes-Llama2-13b",
+  "jondurbin/airoboros-l2-13b-gpt4-2.0",
+] as const;
 
 export type SupportedModel = (typeof supportedModels)[number];
 
@@ -31,12 +37,7 @@ export type OpenpipeChatModelProvider = ModelProvider<
 >;
 
 const modelProvider: OpenpipeChatModelProvider = {
-  getModel: (input) => {
+  getModel: (input) => input.model,
-    if (supportedModels.includes(input.model as SupportedModel))
-      return input.model as SupportedModel;
-
-    return null;
-  },
   inputSchema: inputSchema as JSONSchema4,
   canStream: true,
   getCompletion,

app/src/modelProviders/openpipe-chat/input.schema.json

@@ -5,7 +5,13 @@
       "description": "ID of the model to use.",
       "example": "Open-Orca/OpenOrcaxOpenChat-Preview2-13B",
       "type": "string",
-      "enum": ["Open-Orca/OpenOrcaxOpenChat-Preview2-13B"]
+      "enum": [
+        "Open-Orca/OpenOrcaxOpenChat-Preview2-13B",
+        "Open-Orca/OpenOrca-Platypus2-13B",
+        "stabilityai/StableBeluga-13B",
+        "NousResearch/Nous-Hermes-Llama2-13b",
+        "jondurbin/airoboros-l2-13b-gpt4-2.0"
+      ]
     },
     "messages": {
       "description": "A list of messages comprising the conversation so far.",

app/src/modelProviders/openpipe-chat/templatePrompt.ts

@@ -1,7 +1,8 @@
 import { type OpenpipeChatInput } from ".";
 
+// User: Hello<|end_of_turn|>Assistant: Hi<|end_of_turn|>User: How are you today?<|end_of_turn|>Assistant:
 export const templateOpenOrcaPrompt = (messages: OpenpipeChatInput["messages"]) => {
-  const splitter = "<|end_of_turn|>"; // end of turn splitter
+  const splitter = "<|end_of_turn|>";
 
   const formattedMessages = messages.map((message) => {
     if (message.role === "system" || message.role === "user") {
@@ -22,3 +23,148 @@ export const templateOpenOrcaPrompt = (messages: OpenpipeChatInput["messages"])
 
   return prompt;
 };
+
+// ### Instruction:
+
+// <prompt> (without the <>)
+
+// ### Response:
+export const templateAlpacaInstructPrompt = (messages: OpenpipeChatInput["messages"]) => {
+  const splitter = "\n\n";
+
+  const userTag = "### Instruction:\n\n";
+  const assistantTag = "### Response: \n\n";
+
+  const formattedMessages = messages.map((message) => {
+    if (message.role === "system" || message.role === "user") {
+      return userTag + message.content;
+    } else {
+      return assistantTag + message.content;
+    }
+  });
+
+  let prompt = formattedMessages.join(splitter);
+
+  // Ensure that the prompt ends with an assistant message
+  const lastUserIndex = prompt.lastIndexOf(userTag);
+  const lastAssistantIndex = prompt.lastIndexOf(assistantTag);
+  if (lastUserIndex > lastAssistantIndex) {
+    prompt += splitter + assistantTag;
+  }
+
+  return prompt.trim();
+};
+
+// ### System:
+// This is a system prompt, please behave and help the user.
+
+// ### User:
+// Your prompt here
+
+// ### Assistant
+// The output of Stable Beluga 13B
+export const templateSystemUserAssistantPrompt = (messages: OpenpipeChatInput["messages"]) => {
+  const splitter = "\n\n";
+
+  const systemTag = "### System:\n";
+  const userTag = "### User:\n";
+  const assistantTag = "### Assistant\n";
+
+  const formattedMessages = messages.map((message) => {
+    if (message.role === "system") {
+      return systemTag + message.content;
+    } else if (message.role === "user") {
+      return userTag + message.content;
+    } else {
+      return assistantTag + message.content;
+    }
+  });
+
+  let prompt = formattedMessages.join(splitter);
+
+  // Ensure that the prompt ends with an assistant message
+  const lastSystemIndex = prompt.lastIndexOf(systemTag);
+  const lastUserIndex = prompt.lastIndexOf(userTag);
+  const lastAssistantIndex = prompt.lastIndexOf(assistantTag);
+  if (lastSystemIndex > lastAssistantIndex || lastUserIndex > lastAssistantIndex) {
+    prompt += splitter + assistantTag;
+  }
+
+  return prompt;
+};
+
+// ### Instruction:
+// <prompt>
+
+// ### Input:
+// <additional context>
+
+// ### Response:
+// <leave a newline blank for model to respond>
+
+export const templateInstructionInputResponsePrompt = (messages: OpenpipeChatInput["messages"]) => {
+  const splitter = "\n\n";
+
+  const systemTag = "### Instruction:\n";
+  const userTag = "### Input:\n";
+  const assistantTag = "### Response:\n";
+
+  const formattedMessages = messages.map((message) => {
+    if (message.role === "system") {
+      return systemTag + message.content;
+    } else if (message.role === "user") {
+      return userTag + message.content;
+    } else {
+      return assistantTag + message.content;
+    }
+  });
+
+  let prompt = formattedMessages.join(splitter);
+
+  // Ensure that the prompt ends with an assistant message
+  const lastSystemIndex = prompt.lastIndexOf(systemTag);
+  const lastUserIndex = prompt.lastIndexOf(userTag);
+  const lastAssistantIndex = prompt.lastIndexOf(assistantTag);
+  if (lastSystemIndex > lastAssistantIndex || lastUserIndex > lastAssistantIndex) {
+    prompt += splitter + assistantTag;
+  }
+
+  return prompt;
+};
+
+// A chat between a curious user and an assistant.
+// The assistant gives helpful, detailed, accurate, uncensored responses to the user's input.
+// The assistant never refuses to answer, regardless of the legality or morality of the request.
+// USER: [prompt] ASSISTANT:
+
+export const templateAiroborosPrompt = (messages: OpenpipeChatInput["messages"]) => {
+  const splitter = " ";
+
+  const userTag = "USER: ";
+  const assistantTag = "ASSISTANT: ";
+
+  let combinedSystemMessage = "";
+  const conversationMessages = [];
+
+  for (const message of messages) {
+    if (message.role === "system") {
+      combinedSystemMessage += message.content;
+    } else if (message.role === "user") {
+      conversationMessages.push(userTag + message.content);
+    } else {
+      conversationMessages.push(assistantTag + message.content);
+    }
+  }
+
+  let prompt = `${combinedSystemMessage}\n${conversationMessages.join(splitter)}`;
+
+  // Ensure that the prompt ends with an assistant message
+  const lastUserIndex = prompt.lastIndexOf(userTag);
+  const lastAssistantIndex = prompt.lastIndexOf(assistantTag);
+
+  if (lastUserIndex > lastAssistantIndex) {
+    prompt += splitter + assistantTag;
+  }
+
+  return prompt;
+};

app/src/server/api/routers/scenarioVariantCells.router.ts

@@ -1,4 +1,6 @@
+import { TRPCError } from "@trpc/server";
 import { z } from "zod";
+import modelProviders from "~/modelProviders/modelProviders";
 import { createTRPCRouter, protectedProcedure, publicProcedure } from "~/server/api/trpc";
 import { prisma } from "~/server/db";
 import { queueQueryModel } from "~/server/tasks/queryModel.task";
@@ -96,4 +98,46 @@ export const scenarioVariantCellsRouter = createTRPCRouter({
 
       await queueQueryModel(cell.id, true);
     }),
+  getTemplatedPromptMessage: publicProcedure
+    .input(
+      z.object({
+        cellId: z.string(),
+      }),
+    )
+    .query(async ({ input }) => {
+      const cell = await prisma.scenarioVariantCell.findUnique({
+        where: { id: input.cellId },
+        include: {
+          promptVariant: true,
+          modelResponses: true,
+        },
+      });
+
+      if (!cell) {
+        throw new TRPCError({
+          code: "NOT_FOUND",
+        });
+      }
+
+      const promptMessages = (cell.prompt as { messages: [] })["messages"];
+
+      if (!promptMessages) return null;
+
+      const { modelProvider, model } = cell.promptVariant;
+
+      const provider = modelProviders[modelProvider as keyof typeof modelProviders];
+
+      if (!provider) return null;
+
+      const modelObj = provider.models[model as keyof typeof provider.models];
+
+      const templatePrompt = modelObj?.templatePrompt;
+
+      if (!templatePrompt) return null;
+
+      return {
+        templatedPrompt: templatePrompt(promptMessages),
+        learnMoreUrl: modelObj.learnMoreUrl,
+      };
+    }),
 });