Skip to content

Performance & Connectivity

This page covers common performance and connectivity issues and how to resolve them.

Slow Responses

Section titled “Slow Responses”

Symptom: Agents take a long time to respond to messages.

Solutions

Section titled “Solutions”

  1. Check the model — Larger, more powerful models (Opus, o1) take longer to generate responses than fast-tier models (Haiku, GPT-4o Mini, GLM-4.7 Flash). Switch to a faster model if speed is important.

  2. Check provider status — The AI provider may be experiencing high load:

  3. Reduce conversation length — Very long conversations include more history in each request, increasing response time. Start a new conversation if the current one is very long.

  4. Check tool invocations — If the agent is using tools (web search, code execution), each tool call adds latency. Check the audit log to see what tools are being invoked.

  5. Try a different provider — If one provider is slow, switch to another to determine if it is a provider-side issue.

Response Timeouts

Section titled “Response Timeouts”

Symptom: Messages produce a “timed out” error or the response stream stops.

Solutions

Section titled “Solutions”

  1. Retry the message — Transient network issues can cause timeouts. Simply resend the message.

  2. Use a simpler prompt — Very complex prompts that require extensive reasoning or multiple tool calls may exceed the timeout window.

  3. Switch to a faster model — Fast-tier models (Claude 3.5 Haiku, GPT-4o Mini) respond much more quickly than powerful-tier models.

  4. Check your internet connection — A slow or unstable connection can cause streaming to break.

Browser Issues

Section titled “Browser Issues”

Symptom: The console is unresponsive, not loading, or displaying errors.

Solutions

Section titled “Solutions”

  1. Hard refresh — Press Ctrl+Shift+R (Windows/Linux) or Cmd+Shift+R (macOS) to force a reload without cache.

  2. Clear browser cache — Clear cache and cookies for aigentic.life.

  3. Try a different browser — If the issue persists, try Chrome, Firefox, or Edge to rule out browser-specific problems.

  4. Check for browser extensions — Ad blockers or privacy extensions may interfere with WebSocket connections. Try disabling extensions temporarily.

  5. Check JavaScript console — Open the browser’s developer tools (F12) and check the Console tab for error messages.

WebSocket Disconnections

Section titled “WebSocket Disconnections”

Symptom: The chat interface shows a “Connecting to your agents” warning banner, or responses stop mid-stream.

Solutions

Section titled “Solutions”

  1. Refresh the page — The WebSocket connection will be re-established.

  2. Check your internet connection — WebSocket connections are persistent and sensitive to network interruptions. If you are on Wi-Fi, check signal strength.

  3. VPN interference — Some VPNs interfere with WebSocket connections. Try disconnecting the VPN temporarily.

  4. Corporate firewalls — Some corporate networks block WebSocket connections. Contact your IT department if you are behind a corporate firewall.

Model-Specific Issues

Section titled “Model-Specific Issues”

NVIDIA NIM Models Return 403

Section titled “NVIDIA NIM Models Return 403”

NVIDIA NIM models may return a 403 Forbidden error even with a valid API key. This usually means you need to accept the model’s End User License Agreement (EULA):

  1. Go to build.nvidia.com.
  2. Search for the model (e.g., DeepSeek V3.2, Kimi K2.5).
  3. Accept the license agreement on the model’s page.
  4. Retry in Aigentic.

Model Availability

Section titled “Model Availability”

NVIDIA NIM model availability may vary. If a model is temporarily unavailable, try:

  • A different NVIDIA NIM model
  • A free Z.AI model (GLM-4.7 Flash)
  • A different provider entirely

Channel-Specific Issues

Section titled “Channel-Specific Issues”

For troubleshooting specific channel integrations, see: