1
import chromadb
2
import google.generativeai as genai
3
import os
4

5
genai.configure(api_key=os.environ["GEMINI_API_KEY"])
6
client = chromadb.PersistentClient(path="./vault_data")
7
collection = client.get_collection(name="campaign_memory")
8

9
def sovereign_query(user_query):
10
    # 1. Embed query (Task type 'retrieval_query' is optimized for questions)
11
    res = genai.embed_content(
12
        model="models/text-embedding-004",
13
        content=user_query,
14
        task_type="retrieval_query"
15
    )
16
    query_vec = res['embedding']
17

18
    # 2. Hunt in Vault
19
    # NOTE: query_embeddings must be a list of lists [[vec]]
20
    results = collection.query(
21
        query_embeddings=[query_vec],
22
        n_results=3
23
    )
24

25
    # Extract documents from the results dictionary
26
    context = "\n\n".join(results['documents'][0])
27

28
    # 3. Grounded Execution
29
    prompt = f"""
30
    <role>Sovereign Project Lead</role>
31
    <context>{context}</context>
32
    <task>{user_query}</task>
33
    <constraint>Answer ONLY using provided context. If unknown, say so.</constraint>
34
    """
35
    model = genai.GenerativeModel('gemini-1.5-pro')
36
    return model.generate_content(prompt).text

1
import { GoogleGenerativeAI } from "@google/generative-ai";
2
import { ChromaClient } from "chromadb";
3

4
const genAI = new GoogleGenerativeAI(process.env.GEMINI_API_KEY);
5
const chroma = new ChromaClient(); // Assumes local Chroma server is running
6

7
async function sovereignQuery(query) {
8
    const collection = await chroma.getCollection({ name: "campaign_memory" });
9
    const embedModel = genAI.getGenerativeModel({ model: "text-embedding-004" });
10

11
    // 1. Vectorize Query
12
    const embedRes = await embedModel.embedContent(query);
13

14
    // 2. Hunt in Vault
15
    // queryEmbeddings expects an array of arrays
16
    const results = await collection.query({
17
        queryEmbeddings: [embedRes.embedding.values],
18
        nResults: 3
19
    });
20
    const contextText = results.documents[0].join("\n\n");
21

22
    // 3. Grounded Response
23
    const prompt = `
24
    <role>Sovereign Project Lead</role>
25
    <context>${contextText}</context>
26
    <task>${query}</task>
27
    <constraint>Answer strictly using context. No hallucinations.</constraint>`;
28

29
    const model = genAI.getGenerativeModel({ model: "gemini-1.5-pro" });
30
    const chatRes = await model.generateContent(prompt);
31
    return chatRes.response.text();
32
}
33

34
sovereignQuery("How do I perform a neural handshake?").then(console.log);

1
import dev.langchain4j.model.googleai.GoogleAiGeminiChatModel;
2
import dev.langchain4j.store.embedding.chroma.ChromaEmbeddingStore;
3
import dev.langchain4j.data.segment.TextSegment;
4
import dev.langchain4j.store.embedding.EmbeddingMatch;
5
import java.util.List;
6
import java.util.stream.Collectors;
7

8
public class Orchestrator {
9
    public static String sovereignQuery(String userQuery) {
10
        // 1. Setup Models & Vault
11
        // Ensure your Chroma container is running on port 8000
12
        ChromaEmbeddingStore store = ChromaEmbeddingStore.builder()
13
            .baseUrl("http://localhost:8000")
14
            .collectionName("campaign_memory")
15
            .build();
16

17
        GoogleAiGeminiChatModel model = GoogleAiGeminiChatModel.builder()
18
            .apiKey(System.getenv("GEMINI_API_KEY"))
19
            .modelName("gemini-1.5-pro")
20
            .build();
21

22
        // 2. Search
23
        // LangChain4j findRelevant handles vectorizing the userQuery automatically
24
        // provided the store is configured with an internal embedding model or
25
        // using the manual match pattern.
26
        List<EmbeddingMatch<TextSegment>> results = store.findRelevant(userQuery, 3);
27

28
        String context = results.stream()
29
            .map(match -> match.embedded().text())
30
            .collect(Collectors.joining("\n\n"));
31

32
        // 3. Execute Grounded Prompt
33
        String prompt = String.format("""
34
            <role>Sovereign Project Lead</role>
35
            <context>%s</context>
36
            <task>%s</task>
37
            <constraint>Strictly ground your answer in the context provided.</constraint>
38
            """, context, userQuery);
39

40
        return model.generate(prompt);
41
    }
42

43
    public static void main(String[] args) {
44
        String answer = sovereignQuery("How do I perform a neural handshake?");
45
        System.out.println("Architect Insight: " + answer);
46
    }
47
}