Skip to content

Analyze Failed Build Logs #156

Analyze Failed Build Logs

Analyze Failed Build Logs #156

# .github/workflows/analyze-failed-build-logs.yml
name: Analyze Failed Build Logs
on:
workflow_run:
workflows: ["CI"] # Ensure this matches the exact name of your primary workflow
types:
- completed
jobs:
analyze-logs:
name: Analyze Logs on Failure
runs-on: ubuntu-latest
if: ${{ github.event.workflow_run.conclusion == 'failure' }}
steps:
- name: Set up environment
run: |
sudo apt-get update
sudo apt-get install -y jq unzip curl
- name: Fetch Failed Jobs
id: fetch_failed_jobs
env:
GH_PAT: ${{ secrets.LOG_ACCESS_TOKEN }}
run: |
# Extract repository owner and name
REPO_OWNER=$(echo "${GITHUB_REPOSITORY}" | cut -d'/' -f1)
REPO_NAME=$(echo "${GITHUB_REPOSITORY}" | cut -d'/' -f2)
RUN_ID=${{ github.event.workflow_run.id }}
echo "Repository Owner: $REPO_OWNER"
echo "Repository Name: $REPO_NAME"
echo "Run ID: $RUN_ID"
# Fetch list of jobs for the run
JOBS_JSON=$(curl -s -H "Authorization: token $GH_PAT" \
-H "Accept: application/vnd.github.v3+json" \
"https://api.github.com/repos/$REPO_OWNER/$REPO_NAME/actions/runs/$RUN_ID/jobs")
# Extract failed jobs' IDs and names
FAILED_JOBS=$(echo "$JOBS_JSON" | jq -r '.jobs[] | select(.conclusion == "failure") | "\(.id) \(.name)"')
if [ -z "$FAILED_JOBS" ]; then
echo "No failed jobs found."
exit 1
fi
echo "Failed Jobs:"
echo "$FAILED_JOBS"
# Save failed jobs info to a temporary file for later use
echo "$FAILED_JOBS" > failed_jobs.txt
- name: Download Logs for Failed Jobs
id: download_failed_logs
env:
GH_PAT: ${{ secrets.LOG_ACCESS_TOKEN }}
run: |
# Re-extract repository owner and name
REPO_OWNER=$(echo "${GITHUB_REPOSITORY}" | cut -d'/' -f1)
REPO_NAME=$(echo "${GITHUB_REPOSITORY}" | cut -d'/' -f2)
# Read failed jobs from previous step
FAILED_JOBS=$(cat failed_jobs.txt)
# Initialize log files
> full_build_log.txt
> failed_jobs_summary.txt
while read -r job_id job_name; do
# Sanitize job name by removing or replacing special characters
SANITIZED_JOB_NAME=$(echo "$job_name" | tr -cd '[:alnum:]_')
echo "- $job_name (Job ID: $job_id)" >> failed_jobs_summary.txt
echo "--- Logs for $job_name ---" >> full_build_log.txt
# Fetch the log URL for the specific job
LOG_URL=$(curl -s -H "Authorization: token $GH_PAT" \
-H "Accept: application/vnd.github.v3+json" \
"https://api.github.com/repos/$REPO_OWNER/$REPO_NAME/actions/jobs/$job_id" | jq -r '.logs_url')
echo "Job ID: $job_id"
echo "Job Name: $job_name"
echo "Log URL: $LOG_URL"
if [ "$LOG_URL" == "null" ] || [ -z "$LOG_URL" ]; then
echo "Failed to get logs URL for job $job_name (ID: $job_id)" >> full_build_log.txt
continue
fi
echo "Downloading logs for job '$job_name' from $LOG_URL"
# Download the job logs using job ID to ensure filename uniqueness
LOG_FILENAME="${job_id}.log"
curl -s -L -H "Authorization: token $GH_PAT" "$LOG_URL" -o "$LOG_FILENAME"
if [ -f "$LOG_FILENAME" ]; then
echo "Logs downloaded for job '$job_name' as $LOG_FILENAME."
echo "Contents of $LOG_FILENAME:" >> full_build_log.txt
cat "$LOG_FILENAME" >> full_build_log.txt
echo "--- End of logs for $job_name ---" >> full_build_log.txt
# Clean up the individual log file
rm "$LOG_FILENAME"
else
echo "Failed to download logs for job '$job_name'." >> full_build_log.txt
fi
done <<< "$FAILED_JOBS"
# Check if any logs were downloaded (excluding headers)
if grep -q "Contents of" full_build_log.txt; then
echo "Logs were successfully downloaded."
else
echo "No logs were downloaded for failed jobs."
exit 1
fi
# For debugging: Output the contents of full_build_log.txt
echo "----- Full Build Log -----"
cat full_build_log.txt
echo "----- End of Full Build Log -----"
- name: Send Logs to DeepSeek API
env:
DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
run: |
# Read the full build log and summary
BUILD_LOG=$(cat full_build_log.txt)
FAILED_JOBS_SUMMARY=$(cat failed_jobs_summary.txt)
# Truncate the log if necessary
MAX_LENGTH=50000
if [ ${#BUILD_LOG} -gt $MAX_LENGTH ]; then
BUILD_LOG=$(echo "$BUILD_LOG" | head -c $MAX_LENGTH)
BUILD_LOG="${BUILD_LOG}... [Truncated]"
fi
# Prepare the payload for DeepSeek
SYSTEM_PROMPT="You are a helpful assistant that analyzes CI build logs to find errors, summarize them, and propose solutions if confident. Focus on the failed jobs and their specific errors."
USER_MESSAGE="Here's a summary of failed jobs:\n\n$FAILED_JOBS_SUMMARY\n\nHere are the CI build logs from the failed jobs (truncated if too long):\n\n$BUILD_LOG"
# Debugging: Output the messages being sent to DeepSeek
echo "## DeepSeek Payload"
echo "System Prompt: $SYSTEM_PROMPT"
echo "User Message: $USER_MESSAGE"
# Send the logs to DeepSeek API
RESPONSE=$(curl -s https://api.deepseek.com/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $DEEPSEEK_API_KEY" \
-d @- <<EOF
{
"model": "deepseek-chat",
"messages": [
{"role": "system", "content": "$SYSTEM_PROMPT"},
{"role": "user", "content": "$USER_MESSAGE"}
],
"stream": false
}
EOF
)
# Extract DeepSeek's response
ASSISTANT_REPLY=$(echo "$RESPONSE" | jq -r '.choices[0].message.content // "No response from DeepSeek."')
# Output the analysis to the step summary
echo "## DeepSeek Analysis" >> $GITHUB_STEP_SUMMARY
echo "$ASSISTANT_REPLY" >> $GITHUB_STEP_SUMMARY
# Fallback: Also output the analysis to the standard log
echo "## DeepSeek Analysis"
echo "$ASSISTANT_REPLY"
# For debugging: Output the full response from DeepSeek
echo "## DeepSeek API Response"
echo "$RESPONSE"
# Optional: Upload the full_build_log.txt as an artifact for further inspection
- name: Upload Full Build Log
if: failure() # Upload only if previous steps failed
uses: actions/upload-artifact@v3
with:
name: full_build_log
path: full_build_log.txt