Add integration tests and CI workflow

[Copilot]

No CI existed to validate the RAG pipeline end-to-end across code check-ins. This adds a full integration test suite wired to GitHub Actions, covering Cosmos DB retrieval, Azure OpenAI endpoints, and the decomposed RAG pipeline.

New files

• tests/test_questions.json – Sample Mars-themed questions driving the RAG pipeline tests

• tests/conftest.py – Module-scoped fixtures (event_loop, config, llm_client, retriever); all tests auto-skip when config.yaml is absent

• tests/test_integration.py – 15 integration tests across 5 classes:

  Class	Coverage
  TestEmbeddingEndpoint	Non-empty float vectors, correct dimension
  TestLLMEndpoint	Non-empty, coherent LLM response
  TestVectorSearch	Non-null results per configured source
  TestFullTextSearch	Non-null results for sources with fulltext_k > 0
  TestRAGPipeline	initial_answer, rounds (reasoning traces), non-empty final_answer for all test questions

• scripts/generate_config.py – Writes config.yaml from config.yaml.example by substituting environment variables sourced from GitHub Secrets; avoids committing credentials

• .github/workflows/ci.yml – Runs on every PR and push to main: installs deps, generates config.yaml from secrets, runs pytest tests/ -v --tb=short; GITHUB_TOKEN scoped to contents: read

Required GitHub Secrets

Secret	Purpose
COSMOS_URI / COSMOS_KEY	Cosmos DB account
COSMOS_DATABASE_NAME	Database (default: divdet)
COSMOS_SOURCE1_CONTAINER / COSMOS_SOURCE2_CONTAINER	Container names
LLM_ENDPOINT / LLM_API_KEY / LLM_MODEL	Azure OpenAI LLM
EMBED_ENDPOINT / EMBED_API_KEY / EMBED_MODEL / EMBED_DIMENSIONS	Azure OpenAI embeddings

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Add unit tests + CI</issue_title>
<issue_description>We need to add unit tests to ensure this code is working end to end between code check ins. The unit tests should be run against every PR.

Guidelines for copilot

They tests should utilize the following assets:

• 2 small prebuilt Cosmos DB collections under 1 account with full text and vector index. We will store its details and credentials in github secrets.
• A seperate Azure OpenAI LLM and inference endpoint with moderate rate limit. The details and credentials will be in github secrets.
• Th CI test should securely generate config.yaml from config.yaml.example by inserting GH secrets (if this is advisable, if not advise on best practice)
• Use sample questions stored in a test_file in the repo

The tests should check that:

• vector search and full text search to both cosmos DB accounts shoudl work and give non-null results
• that the LLM and embedding endpoints are functional.
• python rag_divdet.py with the test questions should generate reasoning traces and final answers </issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Add unit tests + CI #10

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.