Update to correct Perplexity API models (2026)

Corrected model list based on official Perplexity API documentation:

REMOVED (not available via API):
- llama-3.1-sonar-small-128k-online
- llama-3.1-sonar-large-128k-online
- llama-3.1-sonar-huge-128k-online

ADDED (actual API models):
- sonar-reasoning-pro (powered by DeepSeek-R1)
- sonar-deep-research (long-form research)

AVAILABLE MODELS:
1. sonar-pro (recommended for chat bots)
2. sonar (lightweight, fast)
3. sonar-reasoning (real-time reasoning)
4. sonar-reasoning-pro (advanced reasoning)
5. sonar-deep-research (research reports)

Updated info popup with:
- Accurate model descriptions from official docs
- Built on Llama 3.3 70B (sonar/sonar-pro)
- DeepSeek-R1 for sonar-reasoning-pro
- Clear use case recommendations

Updated MODEL_CHANGE.md with correct information.

Source: Perplexity API official documentation (2026)

MODEL_CHANGE.md

@@ -19,42 +19,46 @@ Du kannst das Perplexity-Modell jederzeit ändern, ohne den Bot neu zu konfiguri
 3. Der Setup Wizard startet automatisch
 4. Wähle dein gewünschtes Modell aus der Dropdown-Liste
 
-## Verfügbare Modelle
+## Verfügbare Modelle (Perplexity API 2026)
 
 ### Empfohlen für Chat Bots:
 
-**sonar-pro** (Standard)
-- Beste Qualität und Reasoning
-- Ideal für Chat-Interaktionen
-- Gutes Preis-Leistungs-Verhältnis
+**sonar-pro** (Standard) ⭐
+- Tieferes Inhaltsverständnis
+- 2x mehr Suchergebnisse als sonar
+- Verbesserte Suchgenauigkeit
+- Ideal für komplexe, mehrstufige Anfragen
+- Basiert auf Llama 3.3 70B
+- **Beste Wahl für Twitch Chat Bots!**
 
 ### Alternative Modelle:
 
 **sonar**
-- Schneller, günstiger
-- Für einfache Anfragen
-- Weniger komplex
+- Leichtgewichtig und schnell
+- Niedrigere Kosten
+- Einfache Frage-Antwort-Features
+- Gut für geschwindigkeitsoptimierte Anwendungen
+- Basiert auf Llama 3.3 70B
 
 **sonar-reasoning**
-- Erweitertes Reasoning
-- Für komplexe Probleme
-- Langsamer, aber durchdachter
+- Echtzeit-Reasoning mit Suche
+- Zeigt Denkprozess
+- Gut für Problemlösung
+- Moderate Geschwindigkeit
 
-**llama-3.1-sonar-small-128k-online**
-- Kleinster, schnellster Llama
-- 128k Kontext
-- Beste für schnelle, einfache Antworten
-
-**llama-3.1-sonar-large-128k-online**
-- Größeres Llama-Modell
-- Bessere Qualität als small
-- Guter Kompromiss
-
-**llama-3.1-sonar-huge-128k-online**
-- Größtes Llama-Modell
-- Beste Qualität, aber langsamer
+**sonar-reasoning-pro**
+- Powered by DeepSeek-R1
+- Erweiterte Reasoning-Fähigkeiten
+- Sichtbarer Reasoning-Content via API
+- Beste für komplexe logische Aufgaben
 - Höhere Kosten
 
+**sonar-deep-research**
+- Lange, ausführliche Research-Reports
+- Quellenreiche Ausgabe
+- Beste für detaillierte Analysen
+- Langsamer, umfassende Antworten
+
 ## Beispiel .env Datei
 
 ```env

gui.py

@@ -223,13 +223,12 @@ class SetupWizard:
                  "sonar-pro",
                  "sonar",
                  "sonar-reasoning",
-                 "llama-3.1-sonar-small-128k-online",
-                 "llama-3.1-sonar-large-128k-online",
-                 "llama-3.1-sonar-huge-128k-online"
-             ], default_value="sonar-pro", key="-MODEL-", size=(35, 1)),
+                 "sonar-reasoning-pro",
+                 "sonar-deep-research"
+             ], default_value="sonar-pro", key="-MODEL-", size=(25, 1)),
              sg.Button("?", key="-MODEL-INFO-", size=(2, 1))],
             [sg.Text("", size=(20, 1)),
-             sg.Text("Recommended: sonar-pro (best quality)", font=("Arial", 8), text_color="gray")],
+             sg.Text("Recommended: sonar-pro (chat bots)", font=("Arial", 8), text_color="gray")],
             [sg.Text("Max Tokens:", size=(20, 1)), sg.Input("450", key="-TOKENS-")],
 
             [sg.HorizontalSeparator()],
@@ -253,39 +252,48 @@ class SetupWizard:
 
             # Show model info
             if event == "-MODEL-INFO-":
-                model_info = """PERPLEXITY MODEL OVERVIEW:
+                model_info = """PERPLEXITY API MODELS (2026):
 
-🔵 sonar-pro (RECOMMENDED)
-   - Best quality and reasoning capabilities
-   - Ideal for chat bots and complex queries
-   - Good balance of speed and intelligence
+🔵 sonar-pro (RECOMMENDED FOR CHAT BOTS)
+   - Deeper content understanding
+   - 2x more search results than sonar
+   - Enhanced search accuracy
+   - Best for complex, multi-step queries
+   - Built on Llama 3.3 70B
+   ➜ Use this for chat bots!
 
 🟢 sonar
-   - Faster, lower cost
-   - Good for simple queries
-   - Less complex reasoning
+   - Lightweight and fast
+   - Lower cost
+   - Simple question-answer features
+   - Good for speed-optimized applications
+   - Built on Llama 3.3 70B
 
 🧠 sonar-reasoning
-   - Enhanced reasoning for complex problems
-   - Slower but more thoughtful responses
+   - Real-time reasoning with search
+   - Shows reasoning process
+   - Good for problem-solving
+   - Moderate speed
 
-📚 llama-3.1-sonar-small-128k-online
-   - Smallest, fastest Llama model
-   - 128k context window
-   - Best for simple, quick responses
-
-📚 llama-3.1-sonar-large-128k-online
-   - Larger Llama model
-   - Better quality than small
-   - Good balance
-
-📚 llama-3.1-sonar-huge-128k-online
-   - Largest Llama model
-   - Best quality but slower
+⚡ sonar-reasoning-pro
+   - Powered by DeepSeek-R1
+   - Advanced reasoning capabilities
+   - Visible reasoning content via API
+   - Best for complex logical tasks
    - Higher cost
 
-For most chat bots: Use sonar-pro"""
-                sg.popup_scrolled(model_info, title="Model Information", size=(60, 25))
+📊 sonar-deep-research
+   - Long-form research reports
+   - Source-dense output
+   - Best for detailed analysis
+   - Slower, comprehensive responses
+
+RECOMMENDATION:
+• Chat Bots: sonar-pro
+• Quick answers: sonar
+• Complex reasoning: sonar-reasoning-pro
+• Research: sonar-deep-research"""
+                sg.popup_scrolled(model_info, title="Model Information", size=(70, 30))
                 continue
 
             if event == "-SAVE-":