Orphaned fixes (#165)

Co-authored-by: justinthelaw <justin.law@defenseunicorns.com>

chart/templates/ui/deployment.yaml

@@ -39,8 +39,8 @@ spec:
               value: "###ZARF_VAR_SYSTEM_PROMPT###"
             - name: FINAL_SUMMARIZATION_PROMPT
               value: "###ZARF_VAR_FINAL_SUMMARIZATION_PROMPT###"
-            - name: INTERMEDIATE_SUMMARY_PROMPT
-              value: "###ZARF_VAR_INTERMEDIATE_SUMMARY_PROMPT###"
+            - name: INTERMEDIATE_SUMMARIZATION_PROMPT
+              value: "###ZARF_VAR_INTERMEDIATE_SUMMARIZATION_PROMPT###"
             - name: PUBLIC_DEFAULT_TEMPERATURE
               value: "###ZARF_VAR_TEMPERATURE###"
             - name: MAX_TOKENS

chart/templates/ui/service.yaml

@@ -19,4 +19,4 @@ spec:
     - name: http
       port: 3000
       targetPort: 3000
-      protocol: TCP
+      protocol: TCP

src/lib/components/chatPanel.svelte

@@ -1,5 +1,5 @@
 <script lang="ts">
-    import { ArrowRightSolid, RotateOutline } from "flowbite-svelte-icons";
+    import { ArrowRightSolid } from "flowbite-svelte-icons";
     import SvelteMarkdown from "svelte-markdown";
     import codeblock from "$lib/components/codeblock.svelte";
     import codespan from "$lib/components/codespan.svelte";
@@ -92,7 +92,7 @@
         if (ragEndpointActive && agentSettings.rag_enabled) {
             // Construct the RAG message that will be inserted before the user's message
             let ragResponse = {
-                role: "system",
+                role: "user",
                 content: await queryRag(lastMessage.content),
             };
 

src/routes/upload/+page.server.ts

@@ -10,7 +10,6 @@ import OpenAI from "openai";
 import { env } from "$env/dynamic/private";
 import { PUBLIC_TRANSCRIPTION_MODEL } from "$env/static/public";
 import { batchTranscript, tokenize } from "$lib/tokenizer";
-import { generateSummarizationPrompt } from "$lib/prompt";
 import { clearTmp } from "$lib/cleanup";
 
 const TEMPORARY_DIRECTORY = tmpdir();

zarf.yaml

@@ -44,7 +44,7 @@ variables:
     sensitive: false
   - name: MODEL
     description: The default LLM model to use for chat and summarization
-    default: llama-cpp-python
+    default: vllm
     prompt: true
     sensitive: false
   - name: TRANSCRIPTION_MODEL
@@ -54,15 +54,15 @@ variables:
     sensitive: false
   - name: SYSTEM_PROMPT
     description: The default system prompt to use for the LLM
-    default: "You are a helpful AI assistant created by Defense Unicorns."
+    default: "You are a helpful AI assistant."
     prompt: true
     sensitive: false
   - name: FINAL_SUMMARIZATION_PROMPT
     description: The default system summarization prompt to use for the LLM
     default: "You are a summarizer tasked with creating summaries. You will return an coherent and concise summary using 3 concise sections that are each separated by a newline character: 1) BOTTOM LINE UP FRONT: this section will be a concise paragraph containing an overarching, executive summary of all the notes. 2) NOTES: this section will be bullet points highlighting and summarizing key points, risks, issues, and opportunities. 3) ACTION ITEMS: this section will focus on listing any action items, unanswered questions, or issues present in the text; if there are none that can be identified from the notes, just return 'None' for ACTION ITEMS; if possible, also include the individual or team assigned to each item in ACTION ITEMS."
     prompt: true
     sensitive: false
-  - name: INTERMEDIATE_SUMMARY_PROMPT
+  - name: INTERMEDIATE_SUMMARIZATION_PROMPT
     description: The default system summarization prompt to use for the LLM when summary batching activates
     default: "You are a summarizer tasked with creating summaries. Your key activities include identifying the main points and key details in the given text, and condensing the information into a concise summary that accurately reflects the original text. It is important to avoid any risks such as misinterpreting the text, omitting crucial information, or distorting the original meaning. Use clear and specific language, ensuring that the summary is coherent, well-organized, and effectively communicates the main ideas of the original text."
     prompt: true