Enable intelligent querying of 300+ AI models via Poe API with automatic model selection, task-based routing, and cost-quality optimization.
Purpose: Enable OpenClaw to intelligently query multiple AI models via Poe API
Version: 1.0.0
Last Updated: 2026-03-03
This skill provides:
Use this skill when you need to:
Do NOT use this skill for:
CRITICAL: Before using this skill, read:
MODEL_SELECTION_GUIDE.md
This guide teaches you:
from skills.poe_api.scripts.poe_client import PoeClient
# Initialize
client = PoeClient()
# Simple query (auto-selects best model)
result = client.query_for_task(
task_type="programming",
message="Write a Python function to sort a list"
)
# Specific model
result = client.query("claude-sonnet-4.6", "Your prompt")
# Web search
result = client.web_search("Latest AI developments")
# Generate image
result = client.generate_image("A sunset over mountains")
# Generate video
result = client.generate_video("A cat playing piano")
| Task Type | Primary Model | When to Use |
|---|---|---|
| Programming | claude-sonnet-4.6 | General coding, algorithms |
| Complex Problems | claude-opus-4.6 | Deep reasoning, architecture |
| Fast/Cheap | claude-haiku-4.5 | Quick tasks, simple code |
| Huge Context | gemini-3.1-pro | >200K tokens, design systems |
| Code-Focused | gpt-5.3-codex | Debugging, code completion |
| UI/UX Design | gemini-3.1-pro | Design systems, UX research |
| Requirements | claude-sonnet-4.6 | Gathering, analysis |
| Data Analysis | claude-sonnet-4.6 | Data interpretation |
| Task Type | Model | When to Use |
|---|---|---|
| Simple Search | perplexity-search | Quick lookups |
| Complex Queries | perplexity-sonar-pro | In-depth research |
| Reasoning + Search | perplexity-sonar-rsn-pro | Analysis with sources |
| Deep Research | o3-deep-research | Extensive research |
| Budget Research | o4-mini-deep-research | Cost-conscious |
| Task Type | Model | When to Use |
|---|---|---|
| Best Quality | imagen-4-ultra | Professional graphics |
| Fast + Good | nano-banana-2 | Quick iterations |
| Text in Images | nano-banana-pro | Banners, signs |
| General Purpose | nano-banana | Standard generation |
| Professional Editing | gpt-image-1.5 | Complex edits |
| Asian Aesthetics | seedream-4.0 | Specific style |
| Task Type | Model | When to Use |
|---|---|---|
| Best + Audio | veo-3.1 | Cinematic with sound |
| Cinematic | sora-2-pro | High-fidelity |
| Versatile | kling-o3 | Multiple workflows |
| Standard | sora-2 | Good quality |
| Storytelling | wan-2.6 | Multi-scene |
| Fast | seedance-1.0-pro | Quick generation |
| Task Type | Model | When to Use |
|---|---|---|
| Realistic Speech | elevenlabs-v3 | Audiobooks, podcasts |
| Fast TTS | gemini-2.5-flash-tts | Quick conversion |
| Controlled Speech | hailuo-speech-02 | Fine-grained control |
| Music | hailuo-music-v1.5 | Song generation |
This guide contains:
When in doubt, use claude-sonnet-4.6:
Instead of manually selecting models, use:
# Automatic model selection
client.query_for_task(task_type="programming", message="...")
client.query_for_task(task_type="ui_design", message="...")
client.query_for_task(task_type="data_analysis", message="...")
Strengths:
Strengths:
Strengths:
Strengths:
# General coding
result = client.query_for_task(
task_type="programming",
message="Write a REST API in Python"
)
# Code review
result = client.query_for_task(
task_type="programming",
message=f"Review this code: {code}"
)
# Debugging
result = client.query_for_task(
task_type="programming",
message=f"Debug this error: {error}"
)
# Design system
result = client.query_for_task(
task_type="ui_design",
message="Create a design system for a fintech app"
)
# User research
result = client.query_for_task(
task_type="ui_design",
message="Analyze user flow for checkout process"
)
# Analyze data
result = client.query_for_task(
task_type="data_analysis",
message=f"Analyze this dataset: {data}"
)
# Generate insights
result = client.query_for_task(
task_type="data_analysis",
message="What trends do you see in this data?"
)
# Quick search
result = client.web_search("Latest AI developments 2026")
# Deep research
result = client.deep_search(
"Impact of AI on job markets",
model="o3-deep-research"
)
# Generate image
result = client.generate_image(
"Modern dashboard UI with dark theme"
)
# Generate video
result = client.generate_video(
"A drone shot of city skyline at sunset"
)
# Generate audio
result = client.generate_audio(
"[whispers] Welcome to our podcast",
voice_model="elevenlabs-v3"
)
Ask yourself:
programmingui_designdata_analysisweb_searchimage/video/audiogemini-3.1-pro# Let the skill choose the model
result = client.query_for_task(
task_type="programming",
message="Your task",
complexity="medium" # low, medium, high
)
max_calls_per_task to limit API callsresult["success"]See examples/ directory for:
Error: Model not found
Solution: Model names are case-sensitive. Use lowercase:
claude-sonnet-4.6Claude-Sonnet-4.6Error: Rate limit exceeded
Solution: Wait and retry, or use fallback model
Error: Context exceeds limit
Solution: Use gemini-3.1-pro (1M context)
MODEL_SELECTION_GUIDE.md for detailed model informationexamples/ for usage patternsquery_for_task() for automatic model selectionclaude-sonnet-4.6Remember: The key to using this skill effectively is understanding which model to use for which task. Read MODEL_SELECTION_GUIDE.md carefully! 🎯
ZIP package — ready to use