Fix embedding model not found and missing GitHub token in SST config

docs/comprehensive_tech_spec.md

@@ -57,7 +57,7 @@ export interface IntentRouterOutput {
 ```
 
 ### **Step 2: Define "Intent Router" System Prompt**
-This prompt will be used by a `gemini-2.5-flash` model to process and classify all incoming data.
+This prompt will be used by a `gemini-1.5-flash` model to process and classify all incoming data.
 
 ```text
 # System Prompt: Constellation Engine Intent Router

docs/implementation_plan.md

@@ -187,7 +187,7 @@ This phase implements the logic for getting new data into the Unified Lake.
 
 **2.1. Technical Specification: Intent Router**
 
-*   **Objective:** Create a system prompt for `gemini-2.5-flash` to classify and structure incoming data.
+*   **Objective:** Create a system prompt for `gemini-1.5-flash` to classify and structure incoming data.
 *   **Content:**
     ```text
     # System Prompt: Constellation Engine Intent Router
@@ -276,7 +276,7 @@ This phase focuses on migrating existing data from the GitHub repository into th
      * Uses the Intent Router prompt to process raw text content.
      */
     async function generateUnifiedMetadata(content: string): Promise<IntentRouterOutput> {
-      const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" }); // Or your fine-tuned model
+      const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" }); // Or your fine-tuned model
       const prompt = `# System Prompt: Constellation Engine Intent Router
     ... The Intent Router System Prompt ...\n\nINPUT:\n${content}`; // Replace with actual prompt from phase 2
 

scripts/migrateLegacy.ts

@@ -90,7 +90,7 @@ function extractDateFromMetadata(filePath: string, content: string): string | nu
 }
 
 async function generateUnifiedMetadata(content: string): Promise<ExtendedIntentRouterOutput> {
-  const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+  const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
   const result = await model.generateContent(`${INTENT_ROUTER_PROMPT}\n\nINPUT:\n${content}`);
   const jsonText = result.response.text().replace(/```json\n?|\n?```/g, '').trim();
   return JSON.parse(jsonText) as ExtendedIntentRouterOutput;

src/biographerAsync.ts

@@ -184,7 +184,7 @@ export async function handler(event: AsyncPayload) {
     - Do not use markdown code blocks (\`\`\`markdown).
     `;
 
-    const generativeModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+    const generativeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
     const result = await generativeModel.generateContent(systemPrompt);
     const newLifeLogContent = sanitizeMarkdown(result.response.text());
 

src/dreams.ts

@@ -75,7 +75,7 @@ export async function handler(event: APIGatewayProxyEvent): Promise<APIGatewayPr
       - Output RAW markdown only. Do not wrap the output in markdown code blocks.
     `;
 
-    const generativeModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+    const generativeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
     const result = await generativeModel.generateContent(systemPrompt);
     let newAnalysis = result.response.text();
 

src/fiction.ts

@@ -160,7 +160,7 @@ export async function handler(event: APIGatewayProxyEventV2): Promise<APIGateway
       `;
     }
 
-    const generativeModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+    const generativeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
     const result = await generativeModel.generateContent(systemPrompt);
     let newBibleContent = result.response.text();
 

src/ingest.ts

@@ -35,7 +35,7 @@ export async function handler(event: APIGatewayProxyEvent): Promise<APIGatewayPr
     }
 
     // 2. Intent Router (Classification & Extraction)
-    const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" }); // Fast model for routing
+    const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" }); // Fast model for routing
     const result = await model.generateContent(`${INTENT_ROUTER_SYSTEM_PROMPT}\n\nINPUT:\n${rawInput}`);
     const responseText = result.response.text().replace(/```json\n?|\n?```/g, '').trim();
     const routerOutput = JSON.parse(responseText) as IntentRouterOutput;
@@ -127,7 +127,7 @@ export async function handler(event: APIGatewayProxyEvent): Promise<APIGatewayPr
       const contextText = contextEntries.join("\n\n");
 
       // 4. Synthesize Answer with Gemini
-      const ragModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+      const ragModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
       const ragPrompt = `${RAG_SYSTEM_PROMPT}\n\nUSER QUESTION:\n${routerOutput.content}\n\nRETRIEVED CONTEXT:\n${contextText || "No relevant context found."}`;
 
       const ragResult = await ragModel.generateContent(ragPrompt);

src/librarian/dreamer.ts

@@ -106,7 +106,7 @@ export const handler = async (event?: APIGatewayProxyEventV2): Promise<APIGatewa
               - Output RAW markdown only. Do not wrap the output in markdown code blocks.
             `;
 
-            const generativeModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" }); // Using 2.0 Flash for consistency
+            const generativeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" }); // Using 2.0 Flash for consistency
             const result = await generativeModel.generateContent(systemPrompt);
             let newAnalysis = result.response.text();
 
@@ -221,7 +221,7 @@ export const handler = async (event?: APIGatewayProxyEventV2): Promise<APIGatewa
         }
 
         // 3. Synthesize Connection (The Spark)
-        const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" }); 
+        const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
         const prompt = `
         You are a serendipity engine. I will present two seemingly disparate entries from the user's second brain.
         Your task is to find a creative, insightful, or surprising connection between them.

src/librarian/logBook.ts

@@ -78,7 +78,7 @@ export async function updateReadingList(newLog: string) {
         - Do not use markdown code blocks.
         `;
 
-        const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+        const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
         const result = await model.generateContent(systemPrompt);
         const newContent = sanitizeMarkdown(result.response.text());
 

src/librarian/retrieveAndCurate.ts

@@ -107,7 +107,7 @@ export const handler = async (event: HandlerInput): Promise<Book | null> => {
   }
 
   // 2. Curate & Dedupe: Call Gemini
-  const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+  const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
   const prompt = `You are a curator. Review this list of 20 books. Your task is to select the single best book that fits the provided rationale.
 
 **Rationale:** ${rationale}

src/librarian/strategicAnalysis.ts

@@ -53,7 +53,7 @@ export const handler = async (event: { recentEntries: ConstellationRecord[] }):
   }).join("\n\n---\n\n");
 
   const model = genAI.getGenerativeModel({
-    model: "gemini-2.5-flash",
+    model: "gemini-1.5-flash",
     systemInstruction: `Analyze the user's recent entries (notes, saved articles, thoughts). Identify the core topics and interests.
     Pay attention to the metadata (tags, media types).
 

src/librarian/synthesizeInsights.ts

@@ -83,7 +83,7 @@ export const handler = async (event: HandlerInput): Promise<string> => {
     return `## No new recommendations were found in this run.`;
   }
 
-  const model = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+  const model = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
 
   // Format entries for the prompt
   const entriesContext = recentEntries.map(entry => {

src/lyrics.ts

@@ -118,7 +118,7 @@ INSTRUCTIONS:
 CONSTRAINT: Do not alter the raw text of the lyrics. Only group and arrange them.
 IMPORTANT: Output RAW markdown only. Do not wrap the output in markdown code blocks. Do not include any conversational text.`;
 
-    const generativeModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+    const generativeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
     const result = await generativeModel.generateContent(systemPrompt);
     let newSongSeeds = result.response.text();
 

src/philosopher.ts

@@ -133,7 +133,7 @@ ${contextEntries}
     - Do not use markdown code blocks (
     `;
 
-    const generativeModel = genAI.getGenerativeModel({ model: "gemini-2.5-flash" });
+    const generativeModel = genAI.getGenerativeModel({ model: "gemini-1.5-flash" });
     const result = await generativeModel.generateContent(systemPrompt);
     const newGardenContent = sanitizeMarkdown(result.response.text());
 

src/utils.ts

@@ -17,7 +17,7 @@ const GITHUB_REPO = Resource.GITHUB_REPO.value;
  * @param model The model to use for embedding.
  * @returns The embedding vector.
  */
-export async function getEmbedding(content: string, model = "text-embedding-004"): Promise<number[]> {
+export async function getEmbedding(content: string, model = "embedding-001"): Promise<number[]> {
   const embeddingModel = genAI.getGenerativeModel({ model });
   const embeddingResult = await embeddingModel.embedContent(content);
   return embeddingResult.embedding.values;

sst.config.ts

@@ -232,7 +232,7 @@ export default $config({
 
     api.route("POST /dream", {
       handler: "src/librarian/dreamer.handler",
-      link: [GEMINI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX_HOST, table, auth],
+      link: [GEMINI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX_HOST, table, auth, GITHUB_TOKEN, GITHUB_OWNER, GITHUB_REPO],
       timeout: "90 seconds",
     }, {
       auth: {
@@ -332,7 +332,7 @@ export default $config({
       schedule: "rate(1 day)",
       job: {
         handler: "src/librarian/dreamer.handler",
-        link: [GEMINI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX_HOST, table],
+        link: [GEMINI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX_HOST, table, GITHUB_TOKEN, GITHUB_OWNER, GITHUB_REPO],
         timeout: "90 seconds",
       }
     });