export GITHUB_TOKEN="ghp_..."            # your GitHub PAT
export OPENAI_API_KEY="sk-..."           # your OpenAI API key
export POSTGRES_CONNECTION_STRING="host=localhost dbname=github user=postgres password=..."
# optional overrides
export EMBEDDING_MODEL="text-embedding-3-small"
export COMPLETION_MODEL="gpt-4o-mini"

kind: source
spec:
  name: github
  path: cloudquery/github
  registry: cloudquery
  version: "v14.1.0"
  tables: ["github_issues"]
  destinations: ["postgresql"]
  spec:
    access_token: "${GITHUB_TOKEN}" # Personal Access Token, required if not using App Authentication.
    repos: ["cloudquery/cloudquery"]
    table_options:
      github_issues:
        state: "open" # e.g. "open, all, closed", defaults to `all`

---
kind: destination
spec:
  name: "postgresql"
  path: "cloudquery/postgresql"
  registry: "cloudquery"
  version: "v8.9.0"
  write_mode: "overwrite-delete-stale"
  spec:
    connection_string: ${POSTGRES_CONNECTION_STRING}
    pgvector_config:
      tables: 
        - source_table_name: github_issues
          target_table_name: github_issues_embeddings
          embed_columns: ["title", "body"] # list of columns to embed
        #   metadata_columns: [] # list of columns to copy from the source table as metadata
      text_splitter:
        recursive_text: # default text splitter
          chunk_size: 1000
          chunk_overlap: 500
      openai_embedding:
        api_key: ${OPENAI_API_KEY}
        model_name: text-embedding-3-small # the model to use for embedding
        dimensions: 1536 # the number of dimensions the embedding model is generating

cloudquery sync github_to_postgresql.yaml

-- Latest issues
SELECT number, title, state, updated_at
FROM github_issues
ORDER BY updated_at DESC
LIMIT 5;

-- Embeddings: confirm chunking and vector dimensions
SELECT LEFT(chunk, 120) AS chunk_preview,
       vector_dims(embedding) AS dims
FROM github_issues_embeddings
LIMIT 3;

python3 -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install openai psycopg2-binary

import psycopg2
import sys
from openai import OpenAI
import os

# ----------------------------
# CONFIG
# ----------------------------
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "sk-proj-.....")
POSTGRES_CONNECTION_STRING = os.getenv("POSTGRES_CONNECTION_STRING", "host=localhost dbname=github user=postgres password=...")
EMBEDDING_MODEL = os.getenv("EMBEDDING_MODEL", "text-embedding-3-small")
COMPLETION_MODEL = os.getenv("COMPLETION_MODEL", "gpt-4o-mini")
CONTEXT_CHUNK_COUNT = 5
# ----------------------------

client = OpenAI(api_key=OPENAI_API_KEY)

def get_query_embedding(query: str):
    resp = client.embeddings.create(
        input=query,
        model=EMBEDDING_MODEL,
    )
    return resp.data[0].embedding

def get_most_similar_chunks(query_embedding, limit=5):
    """
    Searches pgvector table for the most similar chunks.
    """
    conn = psycopg2.connect(POSTGRES_CONNECTION_STRING)
    cur = conn.cursor()

    # Using <-> operator for cosine similarity in pgvector
    sql = """
    SELECT chunk
    FROM github_issues_embeddings
    ORDER BY embedding <-> %s::vector
    LIMIT %s
    """
    # Convert Python list to pgvector literal string: "[v1,v2,...]"
    embedding_literal = "[" + ",".join(str(v) for v in query_embedding) + "]"
    cur.execute(sql, (embedding_literal, limit))
    results = [r[0] for r in cur.fetchall()]
    cur.close()
    conn.close()
    return results

def answer_question(query):
    # Step 1: Embed the query
    query_emb = get_query_embedding(query)

    # Step 2: Retrieve top-k similar chunks
    chunks = get_most_similar_chunks(query_emb, limit=CONTEXT_CHUNK_COUNT)
    context = "\n\n".join(chunks)

    # Step 3: Ask GPT with context
    prompt = f"""
You are an assistant. Use the following content to answer the question.
Content:
{context}

Question:
{query}

Answer concisely:
"""
    resp = client.chat.completions.create(
        model=COMPLETION_MODEL,
        messages=[{"role": "user", "content": prompt}],
        temperature=0,
    )
    return resp.choices[0].message.content

# ----------------------------
# USAGE
# ----------------------------
if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("Usage: python main.py <question>")
        print("Example: python main.py 'What are people requesting for Azure support?'")
        sys.exit(1)
    
    question = sys.argv[1]
    answer = answer_question(question)
    print(answer)

./.venv/bin/python main.py "What are people requesting for Azure support?"

People are requesting support for the following Azure features:

1. Programmatic management of Azure Service Quotas through Azure Quota Rest APIs.
2. Update of CQ's policy support to align with the latest CIS Microsoft Azure Foundations Benchmark v1.5.0.
3. Addition of Defender endpoints data to the Azure source plugin for detecting security vulnerabilities in machines.
4. Correct documentation for the `cloud_name` options in CloudQuery configuration, specifically for AzurePublic, AzureGovernment, and AzureChina.