AI Setup & Configuration

ThreadTrak offers three AI analysis modes. Choose based on your needs for accuracy, privacy, and cost. All modes are bring-your-own-key (BYOK) — we never charge for AI.

Quick Reference

ANALYSIS MODES

  • Pattern Matching — Free, instant, offline
  • LM Studio — Free, local, smarter
  • Cloud AI — OpenAI, most accurate, BYOK

WHAT AI DETECTS

  • Intent: Question, Bug, Feature Request, Praise...
  • Sentiment: Positive, Neutral, Negative
  • Urgency: Low, Medium, High
  • Priority Score: 0-100

WHERE TO CONFIGURE

Settings → Analysis Mode → Choose your mode
For OpenAI: Add API key in AI Analysis section
For LM Studio: Enter server URL, select model

Choose Your Analysis Mode

Pattern Matching

FREE - No Setup Required

Built-in regex-based detection. Works offline, instant results. Detects questions, bugs, praise, complaints, and urgency markers.

Free forever
Works offline
Zero latency
100% private

Best for: Quick triage, basic classification

LM Studio

PRO - LOCAL AI

Connect to your local LLM server. Runs Llama, Mistral, or Phi models on your own hardware. Full privacy, no API costs.

Free to run
100% local
You pick the model
5-10s per batch

Best for: Privacy-conscious users with decent hardware

Cloud AI (OpenAI)

PRO - BEST ACCURACY

Bring your own OpenAI API key. Pay OpenAI directly — no monthly fees to us. GPT-4o recommended for best results.

Most accurate
Fastest processing
No hardware required
~$0.001-0.01/reply

Best for: Maximum accuracy and speed

Mode Comparison

FeaturePattern MatchingLM StudioCloud AI
TypeBuilt-inLocal LLMCloud API
SpeedInstant5-10sFast
CostFreeFree (local)~$0.001-0.01/reply
Privacy100% local100% localCloud
Intent Categories51919
Reply SuggestionsNoYesYes

Setup Instructions

Pattern Matching Setup

No setup required! Pattern Matching is enabled by default and works immediately.

  1. Open ThreadTrak on X.com
  2. Go to Settings (gear icon) → Analysis Mode
  3. Select "Pattern Matching" (should be selected by default)
  4. Open any thread — badges will appear automatically
Tip: Pattern Matching detects: Questions (?), Bug keywords (error, crash, bug), Praise (love, amazing, great), Complaints (hate, terrible, awful), and Urgency markers (urgent, ASAP, help).

Cloud AI (OpenAI) Setup

PRO FEATURE
  1. Go to platform.openai.com/api-keys and create a new API key.
  2. Open ThreadTrak on X.com and click the Settings (gear icon).
  3. Navigate to Analysis Mode → select "Cloud AI".
  4. Go to AI Analysis section.
  5. Paste your key (starts with sk-...) into the API Key field.
  6. Select your preferred model:
    • GPT-4o — Best accuracy, ~$0.005/reply
    • GPT-4o-mini — Good balance, ~$0.001/reply
  7. Click "Save". Your key is stored locally in your browser.
  8. Reload the thread to see AI analysis badges.
Cost Estimate: Analyzing 100 replies costs approximately $0.10-0.50 depending on reply length and model choice.

LM Studio Setup

PRO FEATURE
Prerequisite: Download and install LM Studio (free, available for Windows/Mac/Linux).
  1. Open LM Studio and download a model. Recommended:
    • Llama 3.2 3B — Fast, good for most hardware
    • Mistral 7B — Better quality, needs 8GB+ RAM
    • Phi-3 Mini — Lightweight, good balance
  2. Go to the Local Server tab (looks like ↔️).
  3. Load your model and click Start Server.
  4. Ensure the port is 1234 (default).
  5. In ThreadTrak, go to Settings → Analysis Mode → select "LM Studio".
  6. Go to LM Studio section.
  7. Server URL should be: http://localhost:1234
  8. Click Test Connection to verify.
  9. Reload the thread to see AI analysis badges.
Hardware Tip: 8GB RAM for 3B models, 16GB for 7B models. Processing takes 5-10 seconds per batch.

What AI Detects

Intent Categories (19 total)

QuestionAsking for info or help
Bug ReportReporting errors
Feature RequestSuggestions
PraiseAppreciation
CriticismNegative feedback

Plus: Support Request, Constructive Feedback, Agreement, Disagreement, Opinion, Advice, Joke, Sarcasm, Trolling, and more.

Other Signals

Sentiment

PositiveNeutralNegative

Urgency

LowMediumHigh

Priority Score

0-100 score based on intent, sentiment, urgency, and user status. High scores appear at top in Priority View.

AI Reply Suggestions

PRO

In Focus Mode, click "Suggest Reply" to generate response options based on context.

How It Works

  1. Select a reply in Focus Mode
  2. Click "Suggest Reply" or the AI generate button
  3. AI considers the reply, parent tweet, and your tone
  4. Generates 1-3 suggestions
  5. Click to insert, edit as needed, then send

Available Tones

ProfessionalFriendlyCasualWittySupportiveDirectEnthusiastic

Troubleshooting

"AI analysis not running"

Check: Settings → Analysis Mode → is a mode selected? For Cloud AI: Is your API key entered in AI Analysis section? Try reloading the thread.

"Pattern Matching badges not showing"

Check: Settings → AI Analysis → "Show badges in Tree" is enabled. Also: The reply text may not match any detection patterns.

"LM Studio not connecting"

1. Is LM Studio open with the server running?
2. Check the port (default 1234)
3. Enable CORS in LM Studio server settings
4. Make sure a model is loaded
5. Use "Test Connection" in ThreadTrak settings

"OpenAI quota exceeded"

Check your billing status at platform.openai.com. Even with a valid key, you need credits in your account. Add payment method if needed.

"Analysis is slow"

LM Studio processes ~5-10 seconds per batch. For threads with 50+ replies, consider using Pattern Matching or Cloud AI for speed.

Free vs. Pro AI Features

CapabilityFreePro
Pattern Matching✓✓
Pattern Matching badges✓✓
LM Studio integration—✓
Cloud AI (OpenAI)—✓
AI reply suggestions—✓
Priority scoringBasicAI-enhanced
Intent detectionPatterns onlyAI + Patterns