fix: resolve CI test failure and pages deployment error

[OriNachum]

Summary

• CI tests: KnowledgeBaseIntegrationService.__init__ now catches KB service init failures gracefully (sets kb_service = None) instead of crashing with ConfigurationError when AWS_REGION is unset. This fixes test_chat_completion_openai_format returning 500 in CI.
• Pages deployment: Regenerated docs/Gemfile.lock so its DEPENDENCIES section matches the Gemfile's just-the-docs (~> 0.12) constraint (was ~> 0.10), fixing the frozen-mode bundle install failure.

Test plan

• uv run pytest -m "unit" -v — 26 passed
• uv run pytest -m "integration and not real_api" -v — 4 passed, 1 skipped (previously 1 failed)
• CI workflows pass on this PR

🤖 Generated with Claude Code

• Claude

[qodo-code-review]

Review Summary by Qodo

Fix CI test failures and pages deployment errors

🐞 Bug fix

Walkthroughs

Description

• KB service initialization now gracefully handles missing AWS credentials
  - Catches exceptions and sets kb_service = None instead of crashing
  - Prevents 500 errors in chat completion endpoints during CI tests
• Added null check for kb_service in RAG decision logic
• Regenerated docs/Gemfile.lock to match just-the-docs ~> 0.12 constraint

Diagram

flowchart LR
  A["KB Service Init"] -->|Exception Caught| B["Log Warning"]
  B --> C["Set kb_service = None"]
  C --> D["Chat Completion Works"]
  E["RAG Decision Logic"] -->|Check kb_service| F["Graceful Fallback"]
  G["Gemfile.lock"] -->|Regenerated| H["Match Gemfile Constraint"]

Loading

File Changes

1. src/open_bedrock_server/services/knowledge_base_integration_service.py 🐞 Bug fix +6/-2

Add graceful KB service initialization error handling

• Wrapped get_knowledge_base_service() call in try-except block to gracefully handle
 initialization failures
• Logs warning and sets kb_service = None when service initialization fails
• Added null check for self.kb_service in should_use_direct_rag() method to prevent errors when
 KB service is unavailable

src/open_bedrock_server/services/knowledge_base_integration_service.py

---

2. docs/Gemfile.lock ⚙️ Configuration changes +0/-0

Regenerate Gemfile.lock for just-the-docs version

• Regenerated lock file to match Gemfile's just-the-docs (~> 0.12) dependency constraint
• Updates DEPENDENCIES section to reflect correct version specification
• Fixes frozen-mode bundle install failure in pages deployment

docs/Gemfile.lock

---

[qodo-code-review]

Code Review by Qodo

🐞 Bugs (0) 📘 Rule violations (0) 📎 Requirement gaps (0) 🎨 UX Issues (0)

1. KB None triggers exceptions ☑ 🐞 Bug ☼ Reliability

Description

When __init__ sets self.kb_service=None, enhance_chat_request() can still call
self.kb_service.retrieve(...), raising AttributeError and falling into the generic exception handler
(logging an error and skipping KB augmentation). This will cause repeated exception-driven flow and
noisy error logs in environments where KB init fails but requests still trigger KB detection.

Code

src/open_bedrock_server/services/knowledge_base_integration_service.py[R34-38]

+        try:
+            self.kb_service = get_knowledge_base_service()
+        except Exception as e:
+            logger.warning(f"Knowledge Base service unavailable: {e}")
+            self.kb_service = None

Evidence

The PR change explicitly sets kb_service to None on initialization failure; enhance_chat_request()
later calls self.kb_service.retrieve(...) without a None guard, and will then hit its broad
exception handler (logger.error + return original request). The main chat route always instantiates
this integration service and calls enhance_chat_request() regardless of whether direct RAG was used,
so this path can be exercised frequently when KB is unavailable.

src/open_bedrock_server/services/knowledge_base_integration_service.py[33-39]
src/open_bedrock_server/services/knowledge_base_integration_service.py[56-114]
src/open_bedrock_server/api/routes/chat.py[213-257]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`KnowledgeBaseIntegrationService` now allows `self.kb_service` to be `None`, but `enhance_chat_request()` still calls `self.kb_service.retrieve(...)` unconditionally once KB usage is detected. This leads to `AttributeError` (caught later), causing exception-driven control flow and error logs when KB is simply unavailable.

## Issue Context
Direct RAG is guarded via `should_use_direct_rag()` now checking `not self.kb_service`, but the fallback path still calls `enhance_chat_request()` and can attempt retrieval.

## Fix Focus Areas
- src/open_bedrock_server/services/knowledge_base_integration_service.py[41-115]

## Suggested fix
- Add an early return in `enhance_chat_request()` when `self.kb_service is None` (ideally after computing `should_use_kb` / `kb_id`, or even earlier), e.g.:
 - if KB would be used but service is unavailable: log at `debug`/`warning` once and return the original request.
- Optionally add a similar guard in `process_rag_request()` for defense-in-depth (raise a `ServiceUnavailableError` or fall back cleanly).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

2. Init warning lacks traceback ☑ 🐞 Bug ✧ Quality

Description

The __init__ handler logs only the exception message when KB service initialization fails, which
makes diagnosing deployment/config failures harder. Without the stack trace, root-causing
intermittent boto/session/client init issues is significantly more difficult.

Code

src/open_bedrock_server/services/knowledge_base_integration_service.py[R36-37]

+        except Exception as e:
+            logger.warning(f"Knowledge Base service unavailable: {e}")

Evidence

On KB init failure, the code logs a warning with the exception string only and does not include
exc_info=True, so the traceback is dropped.

src/open_bedrock_server/services/knowledge_base_integration_service.py[33-38]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
Initialization failure logging drops the traceback, reducing debuggability.

## Issue Context
`logger.warning(f"Knowledge Base service unavailable: {e}")` only logs the exception message.

## Fix Focus Areas
- src/open_bedrock_server/services/knowledge_base_integration_service.py[33-38]

## Suggested fix
- Switch to structured logging with traceback, e.g.:
 - `logger.warning("Knowledge Base service unavailable", exc_info=True)`
 - or `logger.exception(...)` if appropriate (note: `logger.exception` logs at ERROR by default).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

---

ⓘ The new review experience is currently in Beta. Learn more

[Copilot]

Pull request overview

This PR addresses CI instability in the chat completions endpoint by making Knowledge Base integration initialization non-fatal when AWS configuration is missing, and fixes GitHub Pages deployment by aligning the docs lockfile dependency constraint with the Gemfile.

Changes:

• Make KnowledgeBaseIntegrationService tolerate Knowledge Base service initialization failures by falling back to kb_service = None.
• Prevent direct RAG selection when the KB service is unavailable.
• Update docs/Gemfile.lock to match the Gemfile’s just-the-docs (~> 0.12) constraint.

Reviewed changes

Copilot reviewed 1 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/open_bedrock_server/services/knowledge_base_integration_service.py	Adds graceful KB service init fallback and avoids direct RAG when KB is unavailable.
docs/Gemfile.lock	Updates dependency constraint to just-the-docs (~> 0.12) for bundler frozen-mode compatibility.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.