fix: add exponential retry logic to stackit embedder

[a-klos]

This pull request introduces robust retry logic for embedding requests in the StackitEmbedder and refactors configuration management for external services in the RAG infrastructure. The main changes are the addition of exponential backoff for embedding failures, new retry-related settings, and a more organized approach to service configuration in values.yaml.

Embedding robustness and configuration enhancements:

Retry logic improvements:

• Added exponential backoff retry logic to the embed_documents method in StackitEmbedder, allowing up to a configurable number of retries with delays on transient failures. This improves reliability when the embedding API is temporarily unavailable.
• Introduced logging for failed embedding attempts and retries in stackit_embedder.py for better observability.

Embedder settings:

• Added new configuration fields to StackitEmbedderSettings for max_retries, retry_base_delay, and retry_max_delay, allowing fine-grained control of retry behavior. [1] [2]

Infrastructure/service configuration refactor:

• Refactored infrastructure/rag/values.yaml to move service deployment and connection details (PostgreSQL, Redis, ClickHouse, S3/MinIO) under dedicated configuration sections, clarifying which services are external and improving maintainability. [1] [2]
• Enhanced the langfuse configuration block with explicit image, web, worker, authentication, and environment variable settings for more granular control over deployment.

Embedder deployment configuration:

• Added the new retry parameters (STACKIT_EMBEDDER_MAX_RETRIES, STACKIT_EMBEDDER_RETRY_BASE_DELAY, STACKIT_EMBEDDER_RETRY_MAX_DELAY) to the stackitEmbedder section in values.yaml to enable configuration via environment variables.

[Copilot]

Pull Request Overview

This pull request enhances the StackitEmbedder with robust retry logic using exponential backoff to handle transient embedding API failures. The changes improve reliability and observability while making retry behavior configurable.

Key changes:

• Added exponential backoff retry logic to the embed_documents method with configurable parameters
• Introduced new settings for retry behavior: max retries, base delay, and maximum delay
• Enhanced error handling and logging for better observability of embedding failures

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
stackit_embedder_settings.py	Added three new configuration fields for retry behavior with defaults
stackit_embedder.py	Implemented exponential backoff retry logic with comprehensive error handling and logging
values.yaml	Added environment variable configuration for the new retry parameters

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
You can also share your feedback on Copilot code review for a chance to win a $100 gift card. Take the survey.}

[Copilot]

Catching all exceptions with a bare Exception is too broad. Consider catching specific exceptions like openai.APIError, openai.RateLimitError, or requests.RequestException to avoid retrying on non-recoverable errors like authentication failures.

Suggested change

	except Exception as e:
	except (APIError, RateLimitError, Timeout, APIConnectionError) as e:

[manu-hoffmann]

Looks good with one suggestion on handling rate limited requests.

[manu-hoffmann]

suggestion(rate-limits): you could investigate the response code and use the x-ratelimit-reset-requests and x-ratelimit-reset-tokens header on 429 responses to set the delay. This would reduce client errors and handle rate limits as expected.

To get the headers, you could do something like:

raw_response = openai_client.embeddings.with_raw_response.create(..)
response_headers = raw_response.headers