Skip to content

AI Chat

The AI Chat provides contextual conversations about your logs, issues, and system health.

Overview

SerOps AI Chat is powered by pluggable AI models (Claude, Gemini, DeepSeek, or any OpenAI-compatible provider, including a local Ollama). It understands:

  • Your application logs
  • Recent deployments and code changes
  • Issue history and root causes
  • Infrastructure metrics

Starting a Chat

New Conversation

  1. Click AI Chat in the sidebar
  2. Type your question
  3. The AI responds with context from your data

With Batch Context

From the Analyses page: 1. Select a batch 2. Click Chat about Batch 3. The batch context is automatically included

Chat Features

Session Management

Feature Description
New Session Start fresh conversation
Session History View past conversations
Delete Session Remove old conversations

Context Awareness

The AI has access to: - Last 24 hours of logs (configurable) - Recent code changes from GitLab/GitHub - Active issues and their severity - System health metrics

Sensitive Data Protection

Real-Time Scanning

All messages are scanned for sensitive data before sending to AI.

How It Works

  1. You type a message
  2. SerOps scans for PII, credentials, API keys
  3. If found → Warning Modal appears
  4. Choose to edit, send anyway, or cancel

What Gets Detected

  • API keys and tokens
  • Passwords and secrets
  • Email addresses
  • Credit card numbers
  • Connection strings

See Data Protection for full details.

Best Practices

Getting Better Responses

  1. Be Specific - "Why did service X fail at 2pm?" vs "What's wrong?"
  2. Reference Batches - Use "Chat about Batch" for context
  3. Ask Follow-ups - The AI remembers conversation history
  4. Request Actions - "Show me the logs for this error"

Example Questions

Question What AI Uses
"What caused the spike in errors yesterday?" Log analysis + deployment timeline
"Compare this batch with last week" Historical batch data
"Is the payment service healthy?" Metrics + recent issues
"What changed in the last deployment?" Code change tracking

Model Selection

Configure the AI model in Settings → AI Providers. SerOps supports a pluggable architecture. While Gemini and DeepSeek are defaults, you can register other providers (like Claude) in the database.

Model Best For
Gemini General analysis, faster responses (Default)
DeepSeek Deep technical analysis, code understanding
Custom Any compliant provider added to AIModelProvider

Fallback Configuration

If the primary model fails, SerOps automatically tries the fallback:

Primary (e.g., Gemini) → Fallback (e.g., DeepSeek)

Performance & Quota

Turbo Mode (Heuristics)

Turbo Mode is an automatic backend optimization -- it is not a user-configurable setting. SerOps uses Heuristic Intent Detection to save API quota and improve speed. If your question contains keywords like logs, metrics, issues, or code, the system automatically injects the data without making an extra API call.

Benefits: - 50% Less Quota: Skips the "Intent Classification" step. - Guaranteed Context: Data is always loaded for technical queries.

API Reference

Send Message

POST /api/chat/generate-text/
Content-Type: application/json

{
  "prompt": "What caused the error in batch_20260107?",
  "ai_model": "gemini",
  "chat_id": "optional-session-id"
}

Get Chat History

GET /api/chat/chats/{chat_id}/

List Sessions

GET /api/chat/chat-sessions/

Delete Session

DELETE /api/chat/chats/delete/{chat_id}