Conversation
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
📝 WalkthroughSummary by CodeRabbitRelease Notes
WalkthroughThis pull request introduces a complete Cloud Governance AI agent feature. It adds a Streamlit application that discovers OpenSearch indices, runs a Gemini-powered agent loop with tool-calling enabled, and uses a custom MCP server to expose OpenSearch query tools. The PR includes full deployment infrastructure (Docker, shell bootstrap), dependency manifest, configuration templates, comprehensive documentation, and all application code. ChangesCloud Governance MCP AI Agent
Sequence Diagram(s)sequenceDiagram
participant User as User
participant UI as Streamlit<br/>Chat UI
participant Agent as Gemini<br/>Agent Loop
participant MCP as MCP<br/>Tools
participant OS as OpenSearch
User->>UI: Enter natural<br/>language question
UI->>Agent: run_agent_loop_gemini<br/>(question, tools, history)
Agent->>Agent: Load prior<br/>conversation
Agent->>Agent: Force<br/>tool_choice=ANY
Agent->>Agent: Call Gemini<br/>with tools
loop Up to 10 turns (while tool_calls)
Gemini-->>Agent: Function<br/>call(s)
Agent->>Agent: Extract function<br/>names & args
Agent->>MCP: execute_mcp_tool
MCP->>OS: search_documents /<br/>count_by_field / etc.
OS-->>MCP: Query results
MCP-->>Agent: Result text<br/>(markdown)
Agent->>Agent: Append to<br/>message history
Agent->>Agent: Call Gemini<br/>again
end
Agent-->>Agent: No more<br/>tool calls
Agent-->>UI: Final answer<br/>(with citations)
UI->>UI: Append to<br/>chat history
UI->>UI: Save to<br/>conversations.json
UI-->>User: Display in<br/>chat interface
🎯 4 (Complex) | ⏱️ ~60 minutes 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
Comment |
ebattat
reviewed
May 18, 2026
ebattat
left a comment
ebattat
left a comment
Member
There was a problem hiding this comment.
Maybe we can limit it only to the indexes that we are using, instead of all the existing indexes. Also, it would be great if you could add a filter to the index list.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 10
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@cloud-governance-mcp/app.py`:
- Line 251: The st.error call uses an unnecessary f-string: change the call to
st.error("❌ Failed to connect to MCP server") (remove the leading f) so the
literal isn't treated as an f-string; update the string in the existing st.error
invocation where the message "❌ Failed to connect to MCP server" is used.
- Around line 336-345: The current bare "except:" around the json.loads(result)
block should be replaced with a specific exception handler for JSON parsing
(e.g., catch json.JSONDecodeError or ValueError) so non-JSON parsing failures
aren't swallowed; update the try/except that surrounds json.loads(result) to
"except json.JSONDecodeError:" (import JSONDecodeError if needed) and keep the
existing fallback behavior that truncates and calls st.text(result) when parsing
fails.
- Line 350: The function signature for run_agent_loop_gemini uses an implicit
optional type for previous_messages ("list = None"); update the parameter
annotation to explicitly show nullability by changing previous_messages to use
an explicit union with None (e.g., previous_messages: list | None = None) so the
type hint conforms to PEP 484; modify the function definition of
run_agent_loop_gemini accordingly and update any callers or imports if they rely
on a different typing style.
In `@cloud-governance-mcp/Dockerfile`:
- Line 6: The Dockerfile currently writes a Streamlit config with
showErrorDetails enabled via the printf command that contains
"[client]\nshowErrorDetails = true"; change that default to false by updating
the printf invocation (the line that writes to /root/.streamlit/config.toml) so
the file contains "showErrorDetails = false" instead of "true".
- Around line 1-15: The Dockerfile currently runs as root; create and switch to
a non-root user before CMD by adding a user/group (e.g., appuser), chowning
WORKDIR and any config files (like /root/.streamlit or move config to
/home/appuser/.streamlit) and then set USER appuser so that the final CMD
["streamlit", "run", "app.py", ...] executes unprivileged; update any file paths
or ownership for files copied (app.py, mcp_server.py, requirements) and ensure
the home/config directory exists and is owned by the new user before switching.
In `@cloud-governance-mcp/mcp_server.py`:
- Around line 619-622: The _tool_raw_search function accepts arbitrary Query
DSL; validate and sanitize query_body before calling _get_client().search by (1)
enforcing a capped size: if query_body.get("size") is missing or >1000, set it
to 1000; (2) reject or remove dangerous keys such as "script", "script_fields",
"scroll", and any top-level "stored_fields" that could cause heavy work; (3)
block requests containing "from" values greater than a safe threshold (e.g.,
10000) or convert them to use safe pagination; and (4) pass a client.search
timeout (e.g., timeout="30s" or request_timeout) to the call in _tool_raw_search
to bound execution time. Apply these checks in _tool_raw_search before calling
client.search and return a clear error message when input is rejected.
In `@cloud-governance-mcp/README.md`:
- Around line 147-149: The README currently suggests running the literal command
"cat .env | grep OPENSEARCH" which can expose secrets; replace that step with a
safer approach that only reveals variable names or redacts values—i.e., locate
lines matching OPENSEARCH but then remove or mask the value portion instead of
printing the raw .env contents (use standard text-processing tools such as grep
combined with a value-strip or redaction step, or output only the variable names
with cut), and update the README to show that safer command in place of the
existing one.
In `@cloud-governance-mcp/requirements.txt`:
- Line 3: Replace the requirements entry that currently reads "mcp[all]>=1.0.0"
with a pinned stdio-only dependency "mcp>=1.23.0,<2"; specifically update the
token mcp[all]>=1.0.0 in the requirements file to mcp>=1.23.0,<2 to remove
extras used only by non-stdio transports, tighten the minimum version to 1.23.0
for security fixes, and add the <2 upper bound for reproducible installs.
In `@cloud-governance-mcp/run_agent.sh`:
- Around line 41-43: The configuration snippet in run_agent.sh currently sets
showErrorDetails = true which exposes internal traces; change this default to
showErrorDetails = false in the [client] block (replace the true value with
false) so detailed error traces are disabled by default; ensure any comments or
downstream logic that relies on showErrorDetails are updated to expect false as
the safe default.
- Line 33: Replace the unsafe hard-kill line `kill -9 $(lsof -ti tcp:8501)` with
a graceful shutdown sequence: use `lsof -ti tcp:8501` to capture PIDs, send
SIGTERM (kill -15) to those PIDs, wait a short interval and re-check lsof; only
if any PIDs remain, escalate to SIGKILL (kill -9) for those remaining PIDs. Also
consider filtering the PIDs by expected command/user before sending signals to
avoid killing unrelated processes on port 8501.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: d1b2ebb5-f92c-4ad4-b4d2-9530bc506420
📒 Files selected for processing (9)
cloud-governance-mcp/.dockerignorecloud-governance-mcp/.env.examplecloud-governance-mcp/.gitignorecloud-governance-mcp/Dockerfilecloud-governance-mcp/README.mdcloud-governance-mcp/app.pycloud-governance-mcp/mcp_server.pycloud-governance-mcp/requirements.txtcloud-governance-mcp/run_agent.sh
Member
|
/approved |
Member
|
@pragya811, did u review coderabbitai comments ? |
There was a problem hiding this comment.
♻️ Duplicate comments (1)
cloud-governance-mcp/run_agent.sh (1)
33-37:⚠️ Potential issue | 🟠 Major | ⚡ Quick winRestrict port cleanup to this app’s Streamlit process before signaling.
This still signals any process on port 8501, so it can stop unrelated services on shared machines. Filter PIDs by command (e.g.,
streamlit run app.py) and only escalate to-9for survivors after re-check.Suggested fix
-pids="$(lsof -ti tcp:8501 || true)" -if [ -n "$pids" ]; then - kill $pids 2>/dev/null || true - sleep 1 - kill -9 $pids 2>/dev/null || true -fi +pids="$(lsof -ti tcp:8501 -sTCP:LISTEN || true)" +target_pids="" +for pid in $pids; do + cmd="$(ps -p "$pid" -o args= 2>/dev/null || true)" + case "$cmd" in + *"streamlit run app.py"*) target_pids="$target_pids $pid" ;; + esac +done +if [ -n "$target_pids" ]; then + kill $target_pids 2>/dev/null || true + sleep 1 + still_running="$(lsof -ti tcp:8501 -sTCP:LISTEN || true)" + [ -n "$still_running" ] && kill -9 $still_running 2>/dev/null || true +fi🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@cloud-governance-mcp/run_agent.sh` around lines 33 - 37, The current port cleanup grabs any PID listening on tcp:8501; change the logic in run_agent.sh so you first collect candidate PIDs from lsof -ti tcp:8501 into the pids variable, then filter that list to only include processes whose command matches the Streamlit invocation (e.g., "streamlit run app.py") by checking each PID's command (ps -o cmd= -p <pid> or similar) before signaling; send a graceful SIGTERM to the filtered list, wait a short period, re-check which of those same PIDs are still running and only then escalate to kill -9 for survivors. Ensure variable names (pids and the filtered list) and the ordering (terminate, sleep, re-check, force-kill) match the existing script structure.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Duplicate comments:
In `@cloud-governance-mcp/run_agent.sh`:
- Around line 33-37: The current port cleanup grabs any PID listening on
tcp:8501; change the logic in run_agent.sh so you first collect candidate PIDs
from lsof -ti tcp:8501 into the pids variable, then filter that list to only
include processes whose command matches the Streamlit invocation (e.g.,
"streamlit run app.py") by checking each PID's command (ps -o cmd= -p <pid> or
similar) before signaling; send a graceful SIGTERM to the filtered list, wait a
short period, re-check which of those same PIDs are still running and only then
escalate to kill -9 for survivors. Ensure variable names (pids and the filtered
list) and the ordering (terminate, sleep, re-check, force-kill) match the
existing script structure.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Enterprise
Run ID: a876e573-4888-40f9-ae24-a5bd341c418f
📒 Files selected for processing (6)
cloud-governance-mcp/.env.examplecloud-governance-mcp/Dockerfilecloud-governance-mcp/README.mdcloud-governance-mcp/app.pycloud-governance-mcp/mcp_server.pycloud-governance-mcp/run_agent.sh
✅ Files skipped from review due to trivial changes (2)
- cloud-governance-mcp/README.md
- cloud-governance-mcp/.env.example
🚧 Files skipped from review as they are similar to previous changes (2)
- cloud-governance-mcp/app.py
- cloud-governance-mcp/mcp_server.py
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
cloud-governance-mcp/app.py (1)
534-546:⚠️ Potential issue | 🟠 Major | ⚡ Quick winMissing check for empty
candidatescan crash the agent loop.If Gemini blocks or filters the response (e.g., safety filters triggered),
response.candidatesmay be empty. Accessingcandidates[0]will raiseIndexError, which gets caught by the generic exception handler but provides a poor user experience.🛡️ Proposed fix
+ # Check for valid response + if not response.candidates: + return "❌ No response from Gemini. The request may have been filtered or blocked." + # Add assistant response to history history.append(response.candidates[0].content)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@cloud-governance-mcp/app.py` around lines 534 - 546, The code assumes response.candidates[0] exists and will IndexError when Gemini returns no candidates; before appending to history or accessing .content, add a guard like if not response.candidates: log/warn about the empty response (include any response.status/error if available), optionally set final_answer = response.text or continue the agent loop, and skip extracting function calls; update the blocks that reference response.candidates[0] (history.append(...), fcalls = [...]) to run only after the guard so you never access candidates[0] when the list is empty.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@cloud-governance-mcp/app.py`:
- Around line 534-546: The code assumes response.candidates[0] exists and will
IndexError when Gemini returns no candidates; before appending to history or
accessing .content, add a guard like if not response.candidates: log/warn about
the empty response (include any response.status/error if available), optionally
set final_answer = response.text or continue the agent loop, and skip extracting
function calls; update the blocks that reference response.candidates[0]
(history.append(...), fcalls = [...]) to run only after the guard so you never
access candidates[0] when the list is empty.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Enterprise
Run ID: 7f79fdbe-520c-42bc-8ebf-3775b585cf49
📒 Files selected for processing (1)
cloud-governance-mcp/app.py
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Type of change
Note: Fill x in []
Summary
search_documents(filters=[...])andcount_by_field(group_by="policy")instead of constructing complex JSON queries.keywordfield resolution, type coercion, and case-insensitive field matchingDockerfilefor containerized deployment and a sidebar index selector dropdownArchitecture
Streamlit UI → Gemini AI → MCP stdio → mcp_server.py subprocess → OpenSearch
Tools Available
list_indicesget_fieldssearch_documents.keywordhandlingcount_by_fieldaggregatedate_range_searchraw_searchSetup