Research workflow already fixed - no changes needed, requires validation

[Copilot]

Investigation Results

Research workflow failures (2026-01-08 to 2026-01-16) were caused by ubuntu-slim runner instability during preview period. Issue already resolved by:

• ubuntu-slim GA release (2026-01-22)
• TAVILY_API_KEY secret addition (2026-01-22)
• Workflow recompilation incorporating both fixes (2026-01-24, commit 1bc18eb)

Current State

# research.lock.yml line 380 - TAVILY_API_KEY properly configured
env:
  TAVILY_API_KEY: ${{ secrets.TAVILY_API_KEY }}

Both research.lock.yml and daily-news.lock.yml last compiled 2026-01-24 16:17:22 UTC, after both fixes applied.

Why Unfixed Appearance

Zero workflow runs since 2026-01-24 recompilation. Research workflow is workflow_dispatch only - requires manual trigger for validation.

Recommendation

Close this PR. Validate fix with:

gh workflow run research.lock.yml -f topic="Validation test"

If successful, close original issue. If failure persists, investigate new error messages.

Timeline

Date	Event
2026-01-08	Last successful run (preview period)
2026-01-16	Failures (preview instability)
2026-01-22	ubuntu-slim GA + TAVILY_API_KEY added
2026-01-24	Workflows recompiled (fix applied)
2026-01-25	Investigation (no runs since fix)

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Research Workflow - Still failing after TAVILY_API_KEY fix (20% success rate, 17 days offline)</issue_title>
<issue_description>### Problem

Research workflow remains largely non-operational with only 20% success rate despite TAVILY_API_KEY secret being added. The workflow has been effectively offline for 17 days.

Current Status (2026-01-25)

• Success rate: 1/5 recent runs (20%)
• Latest failure: §21078189533 (2026-01-16)
• Last success: 2026-01-08 (17 days ago)
• Failed step: Suspected "Start MCP gateway" (similar to MCP Inspector)
• Previous issue: Fix Research workflow - Critical failure (90% failure rate) #11434 (auto-closed 2026-01-24)

Root Cause Analysis

Key insight: Daily News workflow recovered immediately after TAVILY_API_KEY was added (2026-01-22), but Research workflow did NOT recover. This suggests:

1. Hypothesis 1: Workflow needs recompilation

   • Secret was added AFTER last compilation
   • Lock file may not reference the new secret
   • Solution: make recompile

2. Hypothesis 2: Different MCP Gateway configuration

   • Research may use different MCP server setup than Daily News
   • May need additional configuration beyond TAVILY_API_KEY
   • Review frontmatter differences

3. Hypothesis 3: Intermittent MCP Gateway issues

   • 1/5 runs succeeded (20% rate)
   • May be timing/connectivity related
   • Could be transient MCP server availability

Comparison with Daily News and MCP Inspector

Aspect	Daily News (✅)	Research (⚠️)	MCP Inspector (❌)
TAVILY_API_KEY	Present	Present	Present
Recovery	Immediate	Partial (20%)	None (0%)
Success rate	40% recovering	20% low	0% failing
Last compiled	Unknown	Unknown	Unknown
MCP Gateway	Working	Intermittent	Failing

Recommended Investigation Steps

Step 1: Recompile Workflow

cd /path/to/repo
make recompile
git add .github/workflows/research.lock.yml
git commit -m "Recompile Research workflow after TAVILY_API_KEY fix"
git push

Step 2: Compare Frontmatter

Compare configurations:

• .github/workflows/daily-news.md (working, 40% success)
• .github/workflows/research.md (failing, 20% success)
• .github/workflows/mcp-inspector.md (failing, 0% success)

Look for differences in:

• MCP server configuration
• Tool permissions
• Timeout settings
• Environment variables

Step 3: Analyze Failed Run Logs

Download artifacts from run 21078189533:

• Check /tmp/gh-aw/mcp-logs/ for MCP Gateway errors
• Review agent stdio logs
• Look for timeout or connection issues

Step 4: Test Manually Multiple Times

# Run 3-5 times to check for intermittent issues
for i in {1..5}; do
  gh workflow run research.lock.yml
  sleep 60
done

Monitor success rate of manual runs.

Success Criteria

• Research workflow runs successfully
• Success rate returns to >80% over next 5 runs
• Research and knowledge work capabilities fully operational
• No intermittent failures

Priority: P1 (High)

Impact: Research capabilities severely limited for 17 days. This blocks automated research tasks, knowledge work, and investigation workflows.

Urgency: High - research functionality is critical for knowledge-based agents and analysis workflows.

Next steps:

1. Recompile workflow (5 min)
2. Test manually 3-5 times (30 min)
3. Analyze intermittent failure pattern (30 min)
4. Apply fix based on findings (variable)

References:

• Latest failed run §21078189533
• Workflow source
• Previous issue githubnext/gh-aw#11434 (auto-closed)
• Dashboard githubnext/gh-aw#11581

AI generated by Workflow Health Manager - Meta-Orchestrator

• expires on Jan 26, 2026, 3:08 AM UTC

Comments on the Issue (you are @copilot in this section)

• Fixes Research Workflow - Still failing after TAVILY_API_KEY fix (20% success rate, 17 days offline) #11722

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.