Common Errors and Fixes for STING-CE

1. “No honey jars available” in Bee Chat (Fixed January 2025)

Symptom

  • Bee Chat shows “No honey jars available” message
  • Message persists after logout/login
  • Honey Jar page works fine and shows available honey jars.

Root Cause

  • Chatbot was calling wrong API endpoint: /bee/context/public instead of /bee/context
  • Knowledge service connectivity issues between chatbot and knowledge service
  • Session/authentication issues after service restarts.

Fix Applied (v2025.1)

# The fix is already applied in the latest version
# If you're still seeing this issue, restart the chatbot:
./manage_sting.sh restart chatbot

Technical Details

  • Fixed endpoint: Chatbot now calls correct /bee/context endpoint.
  • Enhanced error logging: Specific error messages help identify the root cause.
  • Better error handling: Structured responses with status indicators.
  • Session persistence: Improved session management across service restarts.

Prevention

  • Keep services updated: ./manage_sting.sh update chatbot
  • Monitor chatbot logs: docker logs sting-ce-chatbot
  • If issue persists, check knowledge service health: ./manage_sting.sh status.

2. Zombie LLM Process Problem

Symptom

  • After reinstall, LLM service shows wrong model or old uptime
  • curl http://localhost:8086/health shows uptime of hours/days
  • Model changes don’t take effect.

Root Cause

  • Native LLM service process not killed during uninstall
  • PID file persists across reinstalls
  • Process continues running with old configuration.

Fix

# Find the process
lsof -i :8086

# Kill it (replace PID with actual process ID)
kill -9 PID

# Remove PID files
rm -f ~/.sting-ce/run/llm-gateway.pid
rm -f /opt/sting-ce/run/llm-gateway.pid

# Restart service
./sting-llm start

Prevention

  • Always run ./manage_sting.sh uninstall before reinstalling
  • The updated installer now kills processes automatically.

2. Repo ID vs Local Path Error

Symptom

{"detail":"Text generation failed: Repo id must be in the form 'repo_name' or 'namespace/repo_name': '/path/to/model'. Use `repo_type` argument if needed."}

Root Cause

  • Task routing configured to use non-existent model (e.g., phi3)
  • Model path being interpreted as HuggingFace repo ID
  • Path doesn’t exist or isn’t accessible
  • Wrong model being loaded due to config.

Fix

# 1. Check task routing configuration
grep -A 5 "chat:" ~/Documents/GitHub/STING-CE/STING/conf/config.yml

# 2. Fix config to use available models
# Edit config.yml to set tinyllama as primary for chat

# 3. Force load correct model
curl -X POST http://localhost:8086/models/tinyllama/load

# 4. Or specify model explicitly
curl -X POST http://localhost:8086/generate \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Hello", "model": "tinyllama", "max_length": 50}'

# 5. Restart service to reload config
./sting-llm restart

3. Missing Python Dependencies

Symptom

  • ModuleNotFoundError: No module named 'torch'
  • ModuleNotFoundError: No module named 'fastapi'.

Root Cause

  • Virtual environment missing dependencies
  • Incomplete installation.

Fix

cd /path/to/STING-CE/STING
source .venv/bin/activate
pip install -r llm_service/requirements.txt
pip install -r llm_service/requirements.common.txt
pip install -r llm_service/requirements.gateway.txt

4. Docker Compose Shows No Services

Symptom

  • docker compose ps shows empty
  • But docker ps shows containers running.

Root Cause

  • Running command from wrong directory
  • Docker Compose project name mismatch.

Fix

# Run from installation directory
cd ~/.sting-ce
docker compose ps

# Or specify compose file
cd /path/to/STING-CE/STING
docker compose -f docker-compose.yml -f docker-compose.mac.yml ps

5. Alpine Container as LLM Gateway

Symptom

  • docker ps shows alpine:latest as llm-gateway
  • Port forwarding not working.

Root Cause

  • macOS uses stub container to forward to native service
  • Native service not running.

Fix

# Start native service first
./sting-llm start

# Verify it's running
curl http://localhost:8086/health

# Restart docker services
docker compose restart llm-gateway

Quick Diagnostic Commands

# Check all services
docker ps | grep sting
./sting-llm status
curl http://localhost:8086/health
curl http://localhost:8081/health

# Check logs
docker compose logs --tail=50
tail -50 ~/.sting-ce/logs/llm-gateway.log

# Clean restart
./manage_sting.sh stop
pkill -f "python.*server.py"
./manage_sting.sh start

Last updated: