feat(config): improve config hot reload handling and introduce mock sse endpoint for ui testing as not all providers support prompt caching

[JayGhiya]

User description

The changes in this commit improve the handling of configuration hot reloading in the config_hot_reload.py module. Specifically:

• The EnvironmentSettings import is now wrapped in a conditional block to avoid circular imports.
• The settings parameter in the _model_factory.build() call is now properly handled, falling back to creating a new EnvironmentSettings instance if settings is None.

These changes help ensure the configuration hot reload functionality works as expected and avoids potential issues with circular imports.

feat(experiments): add mock SSE server

This commit introduces a new standalone FastAPI application in the experiments directory that provides a mock Server-Sent Events (SSE) server. The purpose of this mock server is to replay events from a captured log file (docs/sse-agent-results.txt) to enable UI/development testing without running the full agent pipeline or external services.

The key features of the mock SSE server include:

• Parsing the reference SSE log file and caching the events in memory on startup.
• Exposing an API endpoint at /v1/codebase-agent-rules that streams the cached events to clients.
• Allowing an optional delay between event emissions to simulate a real-time streaming experience.
• Adding CORS middleware to enable cross-origin requests, matching the main application configuration.

This mock server provides a convenient way to test the UI and other components that consume the codebase agent rules SSE stream without the need for a fully running system.

fix(models): correct typo in app_interfaces field

This commit fixes a minor typo in the AgentMDOutput model, where the app_interaces field was misspelled as app_interaces. The correct field name is now app_interfaces.

This change ensures the model definition matches the expected field name and avoids potential issues with data serialization/deserialization.

chore(docker): add unoplat-code-confluence-query-engine service

This commit adds a new service definition for the unoplat-code-confluence-query-engine application in the local-docker-compose.yml file. The new service includes the following configuration:

• Builds the application image using the Dockerfile in the unoplat-code-confluence-query-engine directory.
• Sets the necessary environment variables for connecting to the PostgreSQL and Neo4j databases.
• Exposes the application on port 8001.
• Depends on the postgresql and neo4j services, ensuring they are healthy before starting the application.

This change makes it easier to run the full application stack locally using Docker Compose, including the query engine service.

refactor(codebase-agent-rules): improve SSE event naming

The changes in this commit improve the naming of the Server-Sent Events (SSE) emitted by the /v1/codebase-agent-rules endpoint. Specifically:

• A new build_repo_agent_event_name() function is introduced to construct a namespaced event name in the format {repository_qualified_name}:{agent_name}:{event_type}.
• This namespaced event name is used when emitting the initial "status" event and the final "aggregated_final_summary_agent" event.

Using the repository-qualified name as the first part of the event name ensures uniqueness across concurrent SSE connections for different repositories, improving the ability to track and identify events.

---

PR Type

Enhancement, Tests, Bug fix

---

Description

• Add mock SSE server for UI testing without full pipeline

• Improve SSE event naming with repository-qualified namespacing

• Fix typo in app_interfaces field name

• Add query engine service to Docker Compose configurations

---

Diagram Walkthrough

flowchart LR
  A["Mock SSE Server"] --> B["Parse SSE Log File"]
  B --> C["Stream Events to UI"]
  D["SSE Event Generator"] --> E["Build Namespaced Event Names"]
  E --> F["Repository:Agent:EventType Format"]
  G["Docker Compose"] --> H["Query Engine Service"]

Loading

File Walkthrough

	Relevant files
Tests	2 files mock_sse_server.py Add standalone mock SSE server for testing                              +161/-0  yaak.rq_ZCT8QECJLF.yaml Add HTTP request configuration for mock testing                    +30/-0
Enhancement	1 files codebase_agent_rules.py Implement namespaced SSE event naming system                          +37/-6
Bug fix	2 files agent_md_output.py Fix typo in app_interfaces field name                                        +1/-1      config_hot_reload.py Improve import handling for circular dependency avoidance +4/-1
Configuration changes	3 files local-docker-compose.yml Add query engine service configuration                                      +25/-0    prod-docker-compose.yml Add query engine service for production                                    +22/-0    Taskfile.yml Add mock SSE server task command                                                  +6/-0
Additional files	1 files __init__.py [link]

---

[claude]

Claude encountered an error —— View job

---

I'll analyze this and get back to you.

[huly-for-github]

Connected to Huly®: UNOPL-951

[qodo-code-review]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 No relevant tests

🔒 Security concerns CORS configuration: Mock SSE server allows all origins, credentials, methods, and headers. Safe for local experiments but risky if deployed inadvertently; restrict origins for non-dev environments.

⚡ Recommended focus areas for review Robustness The SSE log parser assumes each event is on single-line fields and strips lines; multi-line data fields or leading spaces in the source log would be lost. Consider supporting multi-line "data:" fields and preserving exact content where needed. def parse_sse_log_file(file_path: Path) -> List[Dict[str, str]]: """Parse a newline-delimited SSE capture file into event dicts. The expected format mirrors what `curl -N` would show: id: 1\n event: repo:agent:event_type\n data: {"json": "object"}\n \n - Blank line separates events - Lines starting with "::" are treated as comments/keepalive pings and skipped Returns a list of dicts each containing keys: "id", "event", "data". Data is preserved as a raw JSON string to match the real endpoint behavior. """ events: List[Dict[str, str]] = [] current: Dict[str, str] = {} try: raw_text: str = file_path.read_text(encoding="utf-8") except FileNotFoundError as exc: logger.error("SSE log file not found at {}", file_path) raise exc for raw_line in raw_text.splitlines(): line: str = raw_line.strip() if not line: if "event" in current and "data" in current: if "id" not in current: current["id"] = str(len(events)) events.append(current) current = {} continue if line.startswith("::"): # Skip comment/ping lines continue if line.startswith("id:"): current["id"] = line.partition("id:")[2].strip() continue if line.startswith("event:"): current["event"] = line.partition("event:")[2].strip() continue if line.startswith("data:"): # Preserve raw JSON string as-is current["data"] = line.partition("data:")[2].strip() continue # Finalize trailing event if file doesn't end with a blank line if current and "event" in current and "data" in current: if "id" not in current: current["id"] = str(len(events)) events.append(current) return events Compatibility Event names changed to namespaced values; ensure all SSE consumers are updated accordingly and that filtering/handlers expect the new "repo:agent:event_type" format to avoid breaking the UI. def build_repo_agent_event_name( repository_qualified_name: str, agent_name: str, event_type: str ) -> str: """Construct a namespaced SSE event string as repo:agent:event_type. Using repository-qualified name first ensures uniqueness across concurrent SSE connections for different repositories. """ return f"{repository_qualified_name}:{agent_name}:{event_type}" Networking The new service exposes port 8001:8000 while your Taskfile dev runs on 8001; verify no port conflicts and that dependent services reach the correct internal hostname and port. unoplat-code-confluence-query-engine: # container_name: unoplat-code-confluence-query-engine build: context: ./unoplat-code-confluence-query-engine dockerfile: Dockerfile environment: - NEO4J_HOST=neo4j - NEO4J_PORT=7687 - NEO4J_USERNAME=neo4j - NEO4J_PASSWORD=password - DB_HOST=postgresql - DB_PORT=5432 - DB_USER=postgres - DB_PASSWORD=postgres - DB_NAME=code_confluence ports: - "8001:8000" networks: - temporal-network depends_on: postgresql: condition: service_healthy neo4j: condition: service_healthy

[qodo-code-review]

CI Feedback 🧐

A test triggered by this PR failed. Here is an AI-generated analysis of the failure:

Action: claude-review

Failed stage: Run Claude Code Review [❌]

Failure summary: The action failed because the Claude run hit a usage limit and returned an error: - The model response contained the message 5-hour limit reached ∙ resets 3pm, indicating a rate/usage window limit was reached. - The run result was marked with "subtype": "success", "is_error": true, leading the workflow to exit with code 1. - Environment shows CLAUDE_SUCCESS: false, confirming the Claude step failed. - Additionally, copying slash commands failed with Slash commands directory not found or error copying: ShellError: Failed with exit code 1, but the primary failure was the usage limit error.

Relevant error logs: 1: ##[group]Runner Image Provisioner 2: Hosted Compute Agent ... 549: ANTHROPIC_BEDROCK_BASE_URL: 550: ANTHROPIC_VERTEX_PROJECT_ID: 551: CLOUD_ML_REGION: 552: GOOGLE_APPLICATION_CREDENTIALS: 553: ANTHROPIC_VERTEX_BASE_URL: 554: VERTEX_REGION_CLAUDE_3_5_HAIKU: 555: VERTEX_REGION_CLAUDE_3_5_SONNET: 556: VERTEX_REGION_CLAUDE_3_7_SONNET: 557: ##[endgroup] 558: Setting up Claude settings at: /home/runner/.claude/settings.json 559: Creating .claude directory... 560: No existing settings file found, creating new one 561: Updated settings with enableAllProjectMcpServers: true 562: Settings saved successfully 563: Copying slash commands from /home/runner/work/_actions/anthropics/claude-code-action/beta/slash-commands to /home/runner/.claude/ 564: Slash commands directory not found or error copying: ShellError: Failed with exit code 1 565: Prompt file size: 17016 bytes ... 657: }, 658: "content": [ 659: { 660: "type": "text", 661: "text": "5-hour limit reached ∙ resets 3pm" 662: } 663: ] 664: }, 665: "parent_tool_use_id": null, 666: "session_id": "c3c065ed-8805-4634-ae06-d6840862c099", 667: "uuid": "da349fb2-7b8c-47a1-9e2c-24ea2704da92" 668: } 669: { 670: "type": "result", 671: "subtype": "success", 672: "is_error": true, 673: "duration_ms": 397, ... 677: "session_id": "c3c065ed-8805-4634-ae06-d6840862c099", 678: "total_cost_usd": 0, 679: "usage": { 680: "input_tokens": 0, 681: "cache_creation_input_tokens": 0, 682: "cache_read_input_tokens": 0, 683: "output_tokens": 0, 684: "server_tool_use": { 685: "web_search_requests": 0 686: }, 687: "service_tier": "standard" 688: }, 689: "permission_denials": [], 690: "uuid": "57a0a22e-86ba-432a-8223-1c1d87cf0354" 691: } 692: ##[error]Process completed with exit code 1. 693: ##[group]Run bun run ${GITHUB_ACTION_PATH}/src/entrypoints/update-comment-link.ts ... 698: DISALLOWED_TOOLS: WebSearch,WebFetch 699: REPOSITORY: unoplat/unoplat-code-confluence 700: PR_NUMBER: 814 701: CLAUDE_COMMENT_ID: 3292186639 702: GITHUB_RUN_ID: 17734846014 703: GITHUB_TOKEN: *** 704: GITHUB_EVENT_NAME: pull_request 705: TRIGGER_COMMENT_ID: 706: CLAUDE_BRANCH: 707: IS_PR: false 708: BASE_BRANCH: 709: CLAUDE_SUCCESS: false 710: OUTPUT_FILE: /home/runner/work/_temp/claude-execution-output.json 711: TRIGGER_USERNAME: JayGhiya 712: PREPARE_SUCCESS: true 713: PREPARE_ERROR: 714: USE_STICKY_COMMENT: false

[qodo-code-review]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
High-level	Integrate the mock server into the main application Merge the new standalone mock SSE server into the main application, activating it with an environment variable or query parameter. This change would simplify the development setup and help keep the mock and real APIs synchronized. Examples: unoplat-code-confluence-query-engine/experiments/mock_sse_server.py [1-161] """Mock SSE server that replays events from docs/sse-agent-results.txt. This standalone FastAPI app exposes an endpoint compatible with the real `/v1/codebase-agent-rules` route and streams Server-Sent Events in the exact order captured in the reference log file. Useful for UI/dev testing without running the full agent pipeline or external services. """ from __future__ import annotations ... (clipped 151 lines) unoplat-code-confluence-query-engine/Taskfile.yml [57-61] mock-sse: desc: Run mock SSE server (replays docs/sse-agent-results.txt) on port 8002 dir: experiments cmds: - uv run fastapi dev mock_sse_server.py --port 8002 Solution Walkthrough: Before: # file: experiments/mock_sse_server.py # A new, separate FastAPI application is created for mocking. app: FastAPI = FastAPI(title="Mock ... SSE Server") @app.on_event("startup") async def on_startup() -> None: # Loads events from a file ... @app.get("/v1/codebase-agent-rules") async def get_mock_codebase_agent_rules(...) -> EventSourceResponse: # Streams the pre-loaded mock events return EventSourceResponse(...) # file: Taskfile.yml # A separate command is needed to run this mock server. # mock-sse: # cmds: # - uv run fastapi dev mock_sse_server.py --port 8002 After: # file: src/.../main.py (conceptual) # The main application is modified to include the mock logic. app: FastAPI = FastAPI(...) # A configuration setting controls the mock behavior. USE_MOCK_SSE = os.getenv("USE_MOCK_SSE", "false").lower() == "true" @app.get("/v1/codebase-agent-rules") async def get_codebase_agent_rules( request: Request, # ... other params ) -> EventSourceResponse: if USE_MOCK_SSE: # Use the same logic from the mock server, but integrated. return EventSourceResponse(stream_mock_events(request, ...)) else: # The original logic for generating real events. return EventSourceResponse(generate_sse_events(...)) Suggestion importance[1-10]: 7 __ Why: The suggestion proposes a valid architectural improvement by integrating the mock server into the main application, which would simplify the development workflow and reduce maintenance overhead.	Medium
Possible issue	Fix incorrect parsing of multi-line SSE data Modify the SSE parser to correctly handle multi-line data fields by appending subsequent data lines with a newline, instead of overwriting the field. unoplat-code-confluence-query-engine/experiments/mock_sse_server.py [78-81] if line.startswith("data:"): # Preserve raw JSON string as-is - current["data"] = line.partition("data:")[2].strip() + new_data = line.partition("data:")[2].strip() + if "data" in current: + current["data"] += "\n" + new_data + else: + current["data"] = new_data continue Apply / Chat Suggestion importance[1-10]: 6 __ Why: The suggestion correctly identifies a bug in the new SSE parser where it fails to handle multi-line data fields, improving the mock server's robustness and compliance with the SSE standard.	Low
General	Avoid blocking event loop on startup Make the on_startup function synchronous by removing async def to avoid blocking the event loop with synchronous file I/O. unoplat-code-confluence-query-engine/experiments/mock_sse_server.py [125-130] @app.on_event("startup") -async def on_startup() -> None: +def on_startup() -> None: """Load and cache SSE events from the reference file on startup.""" global LOADED_EVENTS LOADED_EVENTS = parse_sse_log_file(SSE_LOG_PATH) logger.info("Loaded {} mock SSE events from {}", len(LOADED_EVENTS), SSE_LOG_PATH) Apply / Chat Suggestion importance[1-10]: 5 __ Why: The suggestion correctly identifies that a synchronous, blocking I/O call is made within an async function, which is an anti-pattern. The proposed fix to make the function synchronous is appropriate and improves code quality.	Low
More

[chatgpt-codex-connector]

Codex Review: Here are some suggestions.

Reply with @codex fix comments to fix any unresolved comments.

About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you open a pull request for review, mark a draft as ready, or comment "@codex review". If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex fix this CI failure" or "@codex address that feedback".

[chatgpt-codex-connector]

[P1] Handle missing SSE log file in mock server

The mock SSE app loads REPO_ROOT / "docs" / "sse-agent-results.txt" during startup, but no such file exists anywhere in the repository (see find . -name "sse-agent-results.txt" returning nothing). Because parse_sse_log_file immediately raises FileNotFoundError when the file is missing, the FastAPI application never finishes starting and the new task mock-sse or Docker usage will crash before serving any events. Either ship the reference file with the repo or guard the startup logic so the server can run without it.

Useful? React with 👍 / 👎.