Merge pull request #87 from zamm-dev/lightweight-llm-calls

Return lightweight LLM call info back to frontend
zamm-dev · May 16, 2024 · 06bdcc6 · 06bdcc6
2 parents 8f36971 + 6ced415
commit 06bdcc6
Show file tree

Hide file tree

Showing 21 changed files with 700 additions and 2,447 deletions.
diff --git a/src-svelte/src/lib/bindings.ts b/src-svelte/src/lib/bindings.ts
@@ -35,30 +35,31 @@ export function getSystemInfo() {
 }
 
 export function chat(provider: Service, llm: string, temperature: number | null, prompt: ChatMessage[]) {
-    return invoke()<LlmCall>("chat", { provider,llm,temperature,prompt })
+    return invoke()<LightweightLlmCall>("chat", { provider,llm,temperature,prompt })
 }
 
 export function getApiCall(id: string) {
     return invoke()<LlmCall>("get_api_call", { id })
 }
 
 export function getApiCalls(offset: number) {
-    return invoke()<LlmCall[]>("get_api_calls", { offset })
+    return invoke()<LightweightLlmCall[]>("get_api_calls", { offset })
 }
 
-export type Request = { prompt: Prompt; temperature: number }
 export type ChatMessage = { role: "System"; text: string } | { role: "Human"; text: string } | { role: "AI"; text: string }
-export type Llm = { name: string; requested: string; provider: Service }
-export type Service = "OpenAI"
-export type EntityId = { uuid: string }
+export type Prompt = ({ type: "Chat" } & ChatPrompt)
 export type Response = { completion: ChatMessage }
+export type Service = "OpenAI"
+export type Request = { prompt: Prompt; temperature: number }
 export type Preferences = { animations_on?: boolean | null; background_animation?: boolean | null; animation_speed?: number | null; transparency_on?: boolean | null; sound_on?: boolean | null; volume?: number | null }
+export type Llm = { name: string; requested: string; provider: Service }
+export type LightweightLlmCall = { id: EntityId; timestamp: string; response_message: ChatMessage }
+export type LlmCall = { id: EntityId; timestamp: string; llm: Llm; request: Request; response: Response; tokens: TokenMetadata }
 export type ApiKeys = { openai: string | null }
 export type OS = "Mac" | "Linux" | "Windows"
 export type Shell = "Bash" | "Zsh" | "PowerShell"
 export type SystemInfo = { zamm_version: string; os: OS | null; shell: Shell | null; shell_init_file: string | null }
-export type LlmCall = { id: EntityId; timestamp: string; llm: Llm; request: Request; response: Response; tokens: TokenMetadata }
+export type ChatPrompt = { messages: ChatMessage[] }
 export type TokenMetadata = { prompt: number | null; response: number | null; total: number | null }
 export type Sound = "Switch" | "Whoosh"
-export type Prompt = ({ type: "Chat" } & ChatPrompt)
-export type ChatPrompt = { messages: ChatMessage[] }
+export type EntityId = { uuid: string }
diff --git a/src-svelte/src/routes/api-calls/ApiCalls.svelte b/src-svelte/src/routes/api-calls/ApiCalls.svelte
@@ -1,5 +1,5 @@
 <script lang="ts">
-  import { getApiCalls, type LlmCall } from "$lib/bindings";
+  import { getApiCalls, type LightweightLlmCall } from "$lib/bindings";
   import { snackbarError } from "$lib/snackbar/Snackbar.svelte";
   import InfoBox from "$lib/InfoBox.svelte";
   import Scrollable from "$lib/Scrollable.svelte";
@@ -12,7 +12,7 @@
 
   export let dateTimeLocale: string | undefined = undefined;
   export let timeZone: string | undefined = undefined;
-  let llmCalls: LlmCall[] = [];
+  let llmCalls: LightweightLlmCall[] = [];
   let llmCallsPromise: Promise<void> | undefined = undefined;
   let allCallsLoaded = false;
   let messageWidth = MIN_MESSAGE_WIDTH;
@@ -111,7 +111,7 @@
             <a href={`/api-calls/${call.id}`}>
               <div class="message instance">
                 <div class="text-container">
-                  <div class="text">{call.response.completion.text}</div>
+                  <div class="text">{call.response_message.text}</div>
                 </div>
                 <div class="time">{formatTimestamp(call.timestamp)}</div>
               </div>

diff --git a/src-svelte/src/routes/chat/Chat.svelte b/src-svelte/src/routes/chat/Chat.svelte
@@ -79,7 +79,7 @@
 
     try {
       let llmCall = await chat("OpenAI", "gpt-4", null, $conversation);
-      appendMessage(llmCall.response.completion);
+      appendMessage(llmCall.response_message);
     } catch (err) {
       snackbarError(err as string);
     } finally {

diff --git a/src-svelte/src/routes/chat/Chat.test.ts b/src-svelte/src/routes/chat/Chat.test.ts
@@ -7,7 +7,7 @@ import PersistentChatView from "./PersistentChatView.svelte";
 import userEvent from "@testing-library/user-event";
 import { TauriInvokePlayback, type ParsedCall } from "$lib/sample-call-testing";
 import { animationSpeed } from "$lib/preferences";
-import type { ChatMessage, LlmCall } from "$lib/bindings";
+import type { ChatMessage, LightweightLlmCall } from "$lib/bindings";
 
 describe("Chat conversation", () => {
   let tauriInvokeMock: Mock;
@@ -81,8 +81,9 @@ describe("Chat conversation", () => {
     expect(screen.getByText(nextExpectedHumanPrompt)).toBeInTheDocument();
 
     expect(tauriInvokeMock).toHaveReturnedTimes(1);
-    const lastResult: LlmCall = tauriInvokeMock.mock.results[0].value;
-    const aiResponse = lastResult.response.completion.text;
+    const lastResult: LightweightLlmCall =
+      tauriInvokeMock.mock.results[0].value;
+    const aiResponse = lastResult.response_message.text;
     const lastSentence = aiResponse.split("\n").slice(-1)[0];
     await waitFor(() => {
       expect(

diff --git a/src-tauri/api/sample-calls/chat-continue-conversation.yaml b/src-tauri/api/sample-calls/chat-continue-conversation.yaml
@@ -29,45 +29,9 @@ response:
     {
       "id": "c13c1e67-2de3-48de-a34c-a32079c03316",
       "timestamp": "2024-01-16T09:50:19.738093890",
-      "llm": {
-        "name": "gpt-4-0613",
-        "requested": "gpt-4",
-        "provider": "OpenAI"
-      },
-      "request": {
-        "prompt": {
-          "type": "Chat",
-          "messages": [
-            {
-              "role": "System",
-              "text": "You are ZAMM, a chat program. Respond in first person."
-            },
-            {
-              "role": "Human",
-              "text": "Hello, does this work?"
-            },
-            {
-              "role": "AI",
-              "text": "Yes, it works. How can I assist you today?"
-            },
-            {
-              "role": "Human",
-              "text": "Tell me something funny."
-            }
-          ]
-        },
-        "temperature": 1.0
-      },
-      "response": {
-        "completion": {
-          "role": "AI",
-          "text": "Sure, here's a joke for you: Why don't scientists trust atoms? Because they make up everything!"
-        }
-      },
-      "tokens": {
-        "prompt": 57,
-        "response": 22,
-        "total": 79
+      "response_message": {
+        "role": "AI",
+        "text": "Sure, here's a joke for you: Why don't scientists trust atoms? Because they make up everything!"
       }
     }
 sideEffects:

diff --git a/src-tauri/api/sample-calls/chat-start-conversation.yaml b/src-tauri/api/sample-calls/chat-start-conversation.yaml
@@ -21,37 +21,9 @@ response:
     {
       "id": "d5ad1e49-f57f-4481-84fb-4d70ba8a7a74",
       "timestamp": "2024-01-16T08:50:19.738093890",
-      "llm": {
-        "name": "gpt-4-0613",
-        "requested": "gpt-4",
-        "provider": "OpenAI"
-      },
-      "request": {
-        "prompt": {
-          "type": "Chat",
-          "messages": [
-            {
-              "role": "System",
-              "text": "You are ZAMM, a chat program. Respond in first person."
-            },
-            {
-              "role": "Human",
-              "text": "Hello, does this work?"
-            }
-          ]
-        },
-        "temperature": 1.0
-      },
-      "response": {
-        "completion": {
-          "role": "AI",
-          "text": "Yes, it works. How can I assist you today?"
-        }
-      },
-      "tokens": {
-        "prompt": 32,
-        "response": 12,
-        "total": 44
+      "response_message": {
+        "role": "AI",
+        "text": "Yes, it works. How can I assist you today?"
       }
     }
 sideEffects: